ProgrammableMarketing/liquidcompute.com (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 9
  • This quarter · 11
  • Big bets · 3
  • Wildcards · 1
  • How we can help
Generated Nov 10, 2026
  1. Programmable
  2. Playbooks
  3. Liquid Compute
Liquid Compute logo

growth playbook · Nov 10, 2026

liquidcompute.com (opens in a new tab)

Liquid Compute growth playbook

Diagnosis

Liquid Compute came out of stealth in September 2026 with a $15M seed, a $250M prepayment facility and a strong research voice: 13 Insights posts in seven weeks, a published offtaker credit score and a detailed CFTC comment letter. The website does not yet turn that authority into pipeline.

The index it calls a daily reference price is invisible to the public, the homepage hides its own announcements and proof, every request form asks lenders and traders for GPU quantities, the $250M facility has no page, and good tools (the Hedge Converter, GCAT, the 13-service registry) sit in the footer or behind unlinked URLs.

This playbook makes the index public and checkable first, fixes the forms, proof and layout shift, then turns the Credit Map, GCAT and the hedge tools into acquisition loops for the new Head of Go-to-Market and Head of Compute Supply, and ends with an independently assured index distributed where traders already look.

Growth Score

6/10

Biggest Opportunity

Become the compute price people quote: a public, transparent Liquid Compute Index, held to the standards the company's own CFTC letter sets, which brings buyers, sellers, traders and lenders to the desk.

Biggest Risk

Silicon Data (Bloomberg, Refinitiv, CME) and Ornn (ICE, Kalshi, hourly settled-trade index) are already distributing compute indices; if the LCI stays behind a login, the market standard will be set before PMEX Markets is designated.

01

Start here

The three moves to make first. Each one is written up in full below.

  1. 1

    Put the Liquid Compute Index on a public page: delayed daily prints, history and methodology, the standard your own CFTC letter sets

    Choose the delay (T+1) and write the methodology page from the CFTC letter's own recommendations.

    High impact First result in 1 month
  2. 2

    Show the $15M seed, the $250M facility and the CFTC letter on the homepage, which currently hides all three

    Agree the four proof lines and which partner names can appear.

    High impact First result in 2 weeks
  3. 3

    Replace the nine-field request forms with one short form that asks each role only what it needs

    Count form starts against submits on /buy-compute, /trade-indices and /finance-compute.

    High impact First result in 2 weeks

02

Your idea portfolio

24 ideas, grouped by when to run them. Open any idea for how to run it: the steps, who to approach, copy you can use, and how to measure it.

  • 8 ideas are marked Unexpected: plays you are unlikely to have tried.

    1 of these are things Programmable can build and run for you.

Every idea placed by the effort it takes against the impact it could have. Quick wins sit top left; select a number to open that idea. The filters below narrow the list by status, timing, category and what Programmable can build.

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 5Big bets · 7Fill-ins · 7Reconsider · 5
Effort →

Showing 24 of 24 ideas

This week 9

Quick wins with a clear first action.

  • 01

    The homepage calls the LCI "a daily reference price for A100, H100, H200, B200, and B300s" and the request forms call them "The US indices Liquid Compute publishes", yet no public page shows a single value, chart, history or methodology: "See current pricing" leads to the login page and /trade-indices is a form. Meanwhile the company's October 20 CFTC letter says the reference index is the main manipulation risk and asks for executed-transaction inputs, IOSCO-based governance and contributor thresholds. Silicon Data is distributed on Bloomberg and Refinitiv, and Ornn says its index updates hourly from settled trades. Right now a trader, lender or journalist can't check or quote the LCI without signing up, so it can't become the price people use.

    High impactMedium effort First result in 1 month Chief Product Officer

    First action

    Choose the delay (T+1) and write the methodology page from the CFTC letter's own recommendations.

    How to measure it

    Access requests from /index

    Now
    –
    Target
    40 a month
    By
    90 days

    Steps

    1. 1

      Set the policy

      Decide the public delay, which GPUs are public and what stays in paid data plans.

      CEO and Co-Founder 1 day

    2. 2

      Write the methodology

      Inputs, executed-transaction share, contributor minimums, weighting, governance and restatements, matching the CFTC letter.

      Senior Research Engineer 1 week

    3. 3

      Legal review

      Disclaimers and wording given the pending DCM and DCO applications.

      Legal counsel 3 days

    4. 4

      Build the pages

      /index and one page per GPU with chart, CSV and Dataset markup.

      Web engineer 1 week

    5. 5

      Link it everywhere

      Homepage Data block, Trade button, product page and LCI mentions in Insights.

      Chief Product Officer 1 day

    6. 6

      Pitch the print

      Send the page to reporters and analysts who cover compute prices, with citation rules.

      Head of Go-to-Market 2 weeks

    7. 7

      Measure

      Visits, citations, backlinks and access requests from /index.

      Chief Product Officer monthly

    First result
    1 month
    Cost
    Two to three weeks of research and engineering time
    Owner
    Chief Product Officer

    Smallest test

    Publish the H100 print with a one-day delay and methodology for two weeks, and count access requests and citations.

    Why it works

    Publish /index with the five GPU prints on a one-day delay, a 90-day chart per GPU, a methodology page (inputs, executed-transaction share, contributor minimums, weighting, governance, restatement and error policy), a CSV of daily history and simple citation rules ("Source: Liquid Compute Index, T+1"). Keep live and intraday data, API access and contract-level detail behind login for paid market-data plans. Link it from the homepage Data block, the Trade button and every Insights post that mentions the LCI, and give each GPU its own indexable URL (for example /index/h100).

    An index wins by being quoted. A public delayed print gives the press, analysts and AI assistants a number to cite with a link, lets counterparties check the desk's quotes before they call, and shows in public the standards the CFTC letter asks for. Delayed public prints with paid real-time access is the usual model for benchmarks, so it should not cost data revenue.

    Ready to use

    ChecklistPublic index page launch list

    [ ] /index with A100, H100, H200, B200 and B300 prints (T+1)
    [ ] One URL per GPU with a 90-day chart
    [ ] Methodology: inputs, executed-transaction share, contributor minimums, weights
    [ ] Governance, restatement and error policy
    [ ] Daily CSV download and citation rules
    [ ] Legal disclaimer reviewed
    [ ] Linked from homepage, Trade button, product page and Insights
    [ ] Access-request CTA tagged "from index"

    Risks and how to handle them

    • Free prints reduce paid data demand

      Keep the delay, intraday data, API and contract-level detail paid.

    • Thin transaction counts are exposed

      Publish contributor and transaction counts honestly; that is the standard the letter asks of everyone.

    Evidence

    • Your websiteHomepage (checked November 9, 2026): "Compute has a price. Now it has a market."; the Liquid Compute Index (LCI) is "a daily reference price for A100, H100, H200, B200, and B300s"; the exchange and clearinghouse are "coming soon" and hedges go through the OTC desk; Buy, Sell, Trade and Finance buttons; "What We're Seeing" shows posts from October 20, 14 and 8; no customer logos, funding or volume figures; Services and GPU testing are linked only in the footer.Source
    • Your websiteTrade indices page: a request form only; its Indices field calls A100, H100, H200, B200 and B300 "The US indices Liquid Compute publishes", but no index value, chart, methodology, history or download appears on any public page checked (November 9, 2026).Source
    • Your websiteProduct page: "Priced. Matched. Delivered."; standard contracts for A100, H100, H200, B200 and B300, "priced per GPU-hour, so offers compare across sellers", running month-to-month or for a fixed term; the "See current pricing" button leads to the login page; no fees or prices are shown.Source
    • Your websiteCFTC comment letter (dated October 20, 2026): treats the reference index as the main manipulation risk; recommends executed transactions as index inputs, IOSCO-based governance, minimum transaction and contributor thresholds, and no index administrator trading contracts on its own index; the authors are "not aware of any" compute price reporting agency under "an external assurance regime"; PMEX Markets, Inc. has applied for DCM designation and PMEX Clearing, Inc. for DCO registration, and PMEX "intends to operate under the name Liquid Compute".Source
    • Competitor researchSilicon Data's homepage lists H100, H200, A100, B200, B300 and MI300X indices "Built to financial-index standards and distributed on the Bloomberg and Refinitiv", claims "Over 3.5 million datapoints", sells benchmarking, forecasting and datasets, and offers a newsletter sign-up.Source
    • Competitor researchOrnn's homepage: "The market for compute. The index for intelligence."; "Settled trades from partners build the Ornn Compute Price Index. Updated hourly"; quotes from ICE and Kalshi on using its index; compute contracts "from an hour to multiple years".Source
  • 02

    The homepage has no customer logos, funding or volume figures, and its "What We're Seeing" block shows three research posts (October 20, 14 and 8) while skipping the two biggest announcements: the $250M facility with K8 Capital (October 6) and the CFTC comment letter (October 21). The seed post names partnerships with Susquehanna, BGC and Wintermute and a $15M round, but a first-time visitor sees none of it. For a market that asks institutions to trust a new venue, the homepage gives no reason to.

    High impactLow effort First result in 2 weeks Head of Go to Market

    First action

    Agree the four proof lines and which partner names can appear.

    How to measure it

    Homepage conversion to /contact

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Pick the proof

      Funding, facility, CFTC applications and approved partner names.

      Head of Go-to-Market 1 day

    2. 2

      Clear names

      Confirm partners and investors approve logo use.

      CEO and Co-Founder 2 days

    3. 3

      Design the strip

      One row under the hero, linking to each announcement.

      Web engineer 1 day

    4. 4

      Pin announcements

      Always show the newest Announcement in What We're Seeing.

      Web engineer half a day

    5. 5

      Add live-today lines

      One line per role: what can be done now.

      Head of Go-to-Market half a day

    6. 6

      Measure

      Clicks on the strip and access-request rate.

      Head of Go-to-Market monthly

    First result
    2 weeks
    Cost
    Staff time only
    Owner
    Head of Go to Market

    Smallest test

    Add the $250M and CFTC links to What We're Seeing today.

    Why it works

    Add a proof strip under the hero: "$15M seed (FirstMark, Chemistry, Y Combinator)", "$250M prepayment facility with K8 Capital", "Applied for CFTC DCM and DCO", and the partnership names already published in the seed post. Pin the latest Announcement next to the research posts in "What We're Seeing". Add one line per role showing what is live today.

    Institutional buyers and lenders look for capital, partners and regulatory progress before they request access. All of it is already public on the site, just not where they land.

    Risks and how to handle them

    • Implying partner endorsement

      Use the exact wording from the published announcements.

    Evidence

    • Your websiteHomepage (checked November 9, 2026): "Compute has a price. Now it has a market."; the Liquid Compute Index (LCI) is "a daily reference price for A100, H100, H200, B200, and B300s"; the exchange and clearinghouse are "coming soon" and hedges go through the OTC desk; Buy, Sell, Trade and Finance buttons; "What We're Seeing" shows posts from October 20, 14 and 8; no customer logos, funding or volume figures; Services and GPU testing are linked only in the footer.Source
    • Your websiteInsights listing: 13 posts dated September 2 to October 21, 2026, in Trading, Credit, Announcements and Substack (one entry links out to the CEO's Substack); the listing shows only date and category; no newsletter sign-up or RSS link on the page.Source
    • Your website"Announcing a $250M Facility to Fund Compute Buyers" (October 6, 2026): a prepayment facility with K8 Capital that funds upfront deposits on multi-year contracts, "available to qualifying buyers contracting capacity through Liquid Compute", with each draw secured by prepaid capacity that Liquid Compute verifies; the post gives no eligibility criteria or application path beyond "visit liquidcompute.com".Source
    • Your websiteSeed announcement (September 15, 2026), published at /insights/compute-is-a-grid-not-a-barrel: a $15 million seed round co-led by Chemistry and FirstMark, with K8 Capital, Night Capital, TrueBridge, Y Combinator and others; "a Y Combinator W25 company based in New York"; partnerships with Susquehanna, BGC and Wintermute; the PMEX Markets footnote is not explained.Source
  • 03

    The homepage Buy, Sell, Trade and Finance buttons lead to forms that mark nine fields as required, including GPU, Other GPU type, Indices, Quantity, Desired start date and Duration in months, even on /finance-compute, where the visitor is a lender. The /contact form already does it better: Name, Work email, Company, "I want to" with the four roles, and an optional note, followed by "a short call". The busiest buttons lead to the longest form.

    High impactLow effort First result in 2 weeks Chief Product Officer

    First action

    Count form starts against submits on /buy-compute, /trade-indices and /finance-compute.

    How to measure it

    Request form completion

    Now
    –
    Target
    +30%
    By
    45 days

    Steps

    1. 1

      Measure drop-off

      Starts and submits per form.

      Web engineer 2 days

    2. 2

      Define fields by role

      Two or three fields for buyers, sellers, traders and lenders.

      Head of Go-to-Market 1 day

    3. 3

      Build conditional fields

      On the /contact form, with the role preset from each button.

      Web engineer 3 days

    4. 4

      Redirect old forms

      Point /buy-compute, /trade-indices and /finance-compute to the new form with role set.

      Web engineer half a day

    5. 5

      Route by role

      Send supply to the Head of Compute Supply and demand, trading and lending to GTM.

      Head of Go-to-Market 1 day

    6. 6

      Compare

      Completion and qualified-call rate before and after.

      Chief Product Officer monthly

    First result
    2 weeks
    Cost
    About a week of engineering
    Owner
    Chief Product Officer

    Smallest test

    Make all but Name, Work email and Interest optional on /finance-compute for two weeks.

    Why it works

    Use the /contact form as the single entry point and add two or three fields that depend on the role: buyers give GPU, quantity and start date; sellers give GPU, quantity and availability; traders give index and direction; lenders give book size and deal stage. Make everything else optional and collect it on the call. Prefill the role from the button clicked.

    Fewer required fields means more first conversations, and the desk qualifies on the call anyway. Lenders and traders stop being asked questions that don't apply to them.

    Risks and how to handle them

    • Less detail before the call

      Ask the missing questions in the confirmation email.

    Evidence

    • Your websiteRequest forms: /finance-compute and /buy-compute mark as required Interest, Name, Work email, GPU, Other GPU type, Indices, Quantity, Desired start date and Duration in months, for lenders as well as buyers. The /contact form asks only Name, Work email, Company, "I want to" (Buy compute, Sell capacity, Take a position, Finance a buildout) and an optional note, then promises "a short call" and onboarding that includes API keys for market data on paid plans.Source
    • Your websiteHomepage (checked November 9, 2026): "Compute has a price. Now it has a market."; the Liquid Compute Index (LCI) is "a daily reference price for A100, H100, H200, B200, and B300s"; the exchange and clearinghouse are "coming soon" and hedges go through the OTC desk; Buy, Sell, Trade and Finance buttons; "What We're Seeing" shows posts from October 20, 14 and 8; no customer logos, funding or volume figures; Services and GPU testing are linked only in the footer.Source
  • 04

    Real Chrome visitors load the homepage's main content quickly (LCP 1.5s), but its layout shift score is 0.12, so the page fails Core Web Vitals on CLS, and the shifting element is the hero section with the headline. In Google's mobile lab test the hero poster image takes 8.4s to appear, the page weighs 2.4 MB and about 690 KB of image savings are available.

    Medium impactLow effort First result in 4 weeks Web engineer

    First action

    Give the hero art container a fixed aspect-ratio.

    How to measure it

    Homepage CLS (75th percentile)

    Now
    –
    Target
    Below 0.1
    By
    30 days

    Steps

    1. 1

      Reproduce

      Record the shift on a throttled phone profile.

      Web engineer 2 hours

    2. 2

      Reserve space

      Aspect-ratio on the hero art and poster.

      Web engineer 2 hours

    3. 3

      Stabilise fonts

      Preload and size-adjusted fallback for the headline font.

      Web engineer half a day

    4. 4

      Shrink the poster

      Responsive AVIF/WebP with priority loading.

      Web engineer half a day

    5. 5

      Trim JavaScript

      Defer below-the-fold animation code.

      Web engineer 1 day

    6. 6

      Fix contrast

      The one failing colour pair.

      Web engineer 1 hour

    7. 7

      Confirm

      Re-run PageSpeed and watch field CLS for four weeks.

      Web engineer 4 weeks

    First result
    4 weeks
    Cost
    Two to three days of engineering
    Owner
    Web engineer

    Smallest test

    Fix the hero aspect-ratio and re-run the lab test today.

    Why it works

    Reserve the hero art's space with fixed width, height or aspect-ratio, load the font with a matching fallback so the headline doesn't reflow, serve the poster image as AVIF/WebP at mobile sizes with priority loading, and defer the non-critical JavaScript. Fix the one colour-contrast issue at the same time.

    The first screen is where institutional visitors decide whether the venue looks solid; a jumping headline undermines that, and passing Core Web Vitals helps search visibility.

    Risks and how to handle them

    • Design changes to the hero

      Keep the visual; only reserve its space.

    Evidence

    • Google PageSpeedHomepage on mobile (November 9, 2026): performance 37 (desktop 69); lab Largest Contentful Paint 8.4s, from the hero poster image; 2.4 MB page with about 690 KB of image savings and 247 KB of unused JavaScript. Real Chrome visitors: LCP 1.5s (good) but layout shift 0.12, so the page fails Core Web Vitals on CLS; the shifting element is the hero section. One colour-contrast issue.Source
  • 05

    The $250M facility with K8 Capital is "available to qualifying buyers contracting capacity through Liquid Compute", but the announcement gives no criteria or application path beyond "visit liquidcompute.com", and the site's Finance button leads to a generic form that asks for GPU indices. The company already has a published way to judge buyers: the Offtaker Score, which finds 38 of 169 scored buyers able to support a multi-year offtake today.

    High impactLow effort First result in 1 month Head of Go to Market

    First action

    Write the eligibility section with the lending partner.

    How to measure it

    Qualified financing calls

    Now
    –
    Target
    15 a month
    By
    90 days

    Steps

    1. 1

      Agree criteria

      Plain-language eligibility with the lending partner.

      CEO and Co-Founder 3 days

    2. 2

      Write the page

      What it funds, how draws are verified, timeline, FAQs.

      Head of Go-to-Market 2 days

    3. 3

      Legal review

      Keep the existing not-an-offer disclosure.

      Legal counsel 2 days

    4. 4

      Build the pre-check

      Four questions and a booking link.

      Web engineer 2 days

    5. 5

      Link it

      Announcement, Finance button, Credit Map and buyer emails.

      Web engineer half a day

    6. 6

      Track

      Pre-checks, calls and draws.

      Head of Go-to-Market monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Head of Go to Market

    Smallest test

    Add an eligibility paragraph and booking link to the announcement post.

    Why it works

    Publish /financing/prepayment: what the facility pays for (upfront deposits on multi-year contracts), how a draw is secured and verified, who qualifies (in plain ranges), how long approval takes, and a four-question pre-check (contract size, term, operator chosen, funding stage) that books a call. Link it from the announcement, the Finance button and the Credit Map.

    A named facility with clear eligibility is a reason for AI companies to contract capacity through Liquid Compute rather than directly, and it gives every buyer conversation a concrete financing next step.

    Risks and how to handle them

    • Criteria read as a commitment to lend

      Present them as indicative and subject to underwriting.

    Builds on

    Evidence

    • Your website"Announcing a $250M Facility to Fund Compute Buyers" (October 6, 2026): a prepayment facility with K8 Capital that funds upfront deposits on multi-year contracts, "available to qualifying buyers contracting capacity through Liquid Compute", with each draw secured by prepaid capacity that Liquid Compute verifies; the post gives no eligibility criteria or application path beyond "visit liquidcompute.com".Source
    • Your websiteRequest forms: /finance-compute and /buy-compute mark as required Interest, Name, Work email, GPU, Other GPU type, Indices, Quantity, Desired start date and Duration in months, for lenders as well as buyers. The /contact form asks only Name, Work email, Company, "I want to" (Buy compute, Sell capacity, Take a position, Finance a buildout) and an optional note, then promises "a short call" and onboarding that includes API keys for market data on paid plans.Source
    • Your website"The GPU Offtaker Credit Map: 200 Private AI Compute Buyers, Ranked" (September 28, 2026): the Liquid Compute Offtaker Score (0-100, five weighted factors, published weights); 169 of 200 buyers scored; "38 of 169 scored companies can support a multi-year offtake today"; the company-level dataset is "available on request"; "We will update it each edition, and re-score any company that shares data with us under NDA."Source
  • 06

    Every page carries the footnote "PMEX Markets is pending designation as a DCM and registration as a DCO by the CFTC", but no page explains who PMEX Markets is; the seed announcement doesn't either. The answer is buried in the CFTC letter: PMEX Markets, Inc. applied for DCM designation, PMEX Clearing, Inc. for DCO registration, and PMEX "intends to operate under the name Liquid Compute". Meanwhile hedges are OTC transactions for eligible contract participants and the exchange is "coming soon". A compliance reviewer at a lender or trading firm has to piece this together.

    Medium impactLow effort First result in 2 weeks Legal counsel

    First action

    Turn the footnote into a link and draft the page from the CFTC letter.

    How to measure it

    Calls slowed by regulatory questions

    Now
    –
    Target
    -50%
    By
    90 days

    Steps

    1. 1

      Draft

      Entities, applications, live services and pending services.

      Legal counsel 2 days

    2. 2

      Check claims

      Match every line to filings and the CFTC letter.

      CEO and Co-Founder 1 day

    3. 3

      Build the page

      /regulatory with an updated date.

      Web engineer half a day

    4. 4

      Link the footnote

      On every page.

      Web engineer 1 hour

    5. 5

      Brief sales

      Use the page in first calls and diligence packs.

      Head of Go-to-Market 1 hour

    6. 6

      Keep current

      Update at each regulatory milestone.

      Legal counsel ongoing

    First result
    2 weeks
    Cost
    Staff time only
    Owner
    Legal counsel

    Smallest test

    Link the footnote to the CFTC letter today.

    Build the monitor. A live page fed by the public registries and re-checked every day. Usually live within days.

    Book a working session

    Why it works

    Publish /regulatory: the entities and what each has applied for, what is available today (index, OTC desk hedges for eligible contract participants, matched capacity, financing, GCAT) and what awaits designation (listed and cleared contracts), links to the CFTC comment letter, and a contact for compliance questions. Link the footnote on every page to it.

    Institutions buy only once compliance signs off. A clear status page shortens that review and stops visitors confusing "pending" with "not open for business".

    Risks and how to handle them

    • Statements read as predicting approval

      State only what has been filed and keep the forward-looking disclaimer.

    Evidence

    • Your websiteCFTC comment letter (dated October 20, 2026): treats the reference index as the main manipulation risk; recommends executed transactions as index inputs, IOSCO-based governance, minimum transaction and contributor thresholds, and no index administrator trading contracts on its own index; the authors are "not aware of any" compute price reporting agency under "an external assurance regime"; PMEX Markets, Inc. has applied for DCM designation and PMEX Clearing, Inc. for DCO registration, and PMEX "intends to operate under the name Liquid Compute".Source
    • Your websiteSeed announcement (September 15, 2026), published at /insights/compute-is-a-grid-not-a-barrel: a $15 million seed round co-led by Chemistry and FirstMark, with K8 Capital, Night Capital, TrueBridge, Y Combinator and others; "a Y Combinator W25 company based in New York"; partnerships with Susquehanna, BGC and Wintermute; the PMEX Markets footnote is not explained.Source
    • Your website"Compute credit is tightening" (October 20, 2026): a cash-settled put on the LCI as a residual value floor; in an illustrative $15M cluster the advance rises from 65% to 85%, and the floor saves the operator about $0.45M a year for about $0.17M a year; the options are OTC transactions for eligible contract participants, not exchange-listed or cleared; lenders are invited to request a mark and hedge.Source
    • Your websiteHomepage (checked November 9, 2026): "Compute has a price. Now it has a market."; the Liquid Compute Index (LCI) is "a daily reference price for A100, H100, H200, B200, and B300s"; the exchange and clearinghouse are "coming soon" and hedges go through the OTC desk; Buy, Sell, Trade and Finance buttons; "What We're Seeing" shows posts from October 20, 14 and 8; no customer logos, funding or volume figures; Services and GPU testing are linked only in the footer.Source
  • 08

    Liquid Compute has built a Hedge Converter that will "Translate your compute economics into an indicative H100 index hedge", with paths for beginners and traders. But the homepage Trade button goes to a request form, the September 8 primer "Hedging Compute Prices" works through a 128-GPU example without linking to the tool, and only the October 20 credit post mentions it (saying it needs a free account). The best self-serve demo of the product is reachable mainly by its URL.

    Medium impactLow effort First result in 2 weeks Chief Product Officer

    First action

    Add a converter link with preloaded numbers to the "Hedging Compute Prices" post.

    How to measure it

    Converter tickets

    Now
    –
    Target
    50 a month
    By
    60 days

    Steps

    1. 1

      Map entry points

      Trade button, product page, Trading posts.

      Chief Product Officer half a day

    2. 2

      Preload examples

      Pass each post's numbers in the link.

      Web engineer 2 days

    3. 3

      Open the result

      Show the indicative hedge before sign-up; ask for the account to save.

      Web engineer 2 days

    4. 4

      Add the links

      Buttons on every entry point.

      Web engineer half a day

    5. 5

      Route tickets

      New tickets alert the OTC desk.

      Head of Go-to-Market 1 day

    6. 6

      Track

      Visits, tickets and desk calls.

      Chief Product Officer monthly

    First result
    2 weeks
    Cost
    About a week of engineering
    Owner
    Chief Product Officer

    Smallest test

    Link the converter from the primer and count clicks for two weeks.

    Why it works

    Add "Price a hedge in two minutes" to the Trade button area, the product page's trading-firm path and every Trading post, with the post's own numbers preloaded (for example the primer's 128 GPUs at $4.40). Let visitors see the indicative result with example numbers before asking them to create the free account to save a ticket.

    Seeing their own hedge priced turns a curious reader into a qualified desk conversation, and the ticket gives the desk the deal's details before the call.

    Risks and how to handle them

    • Indicative numbers mistaken for quotes

      Label every output indicative and keep the OTC disclosure.

    Evidence

    • Your websiteHedge Converter (/trade/hedge-converter): "Translate your compute economics into an indicative H100 index hedge", with paths for "New to this", "I know the basics" and "I trade them"; the October 20 credit post says it needs a free account; the September 8 primer "Hedging Compute Prices" (128 GPUs for a year at $4.40 an hour, about $4.93M) does not link to it and ends with email CTAs for buyers, operators and lenders.Source
    • Your website"Compute credit is tightening" (October 20, 2026): a cash-settled put on the LCI as a residual value floor; in an illustrative $15M cluster the advance rises from 65% to 85%, and the floor saves the operator about $0.45M a year for about $0.17M a year; the options are OTC transactions for eligible contract participants, not exchange-listed or cleared; lenders are invited to request a mark and hedge.Source
    • Your websiteHomepage (checked November 9, 2026): "Compute has a price. Now it has a market."; the Liquid Compute Index (LCI) is "a daily reference price for A100, H100, H200, B200, and B300s"; the exchange and clearinghouse are "coming soon" and hedges go through the OTC desk; Buy, Sell, Trade and Finance buttons; "What We're Seeing" shows posts from October 20, 14 and 8; no customer logos, funding or volume figures; Services and GPU testing are linked only in the footer.Source
  • 09

    The Services page lists 13 concrete services by role (compute sourcing, a cash-settled price hedge, prepayment and full contract financing, unused capacity resale, GCAT, bankable offtake origination, tail-period price protection, remarketing, index and data licensing) with a five-step inquiry flow. It is linked only from the footer; the main navigation is Product, About, Insights and Careers. The product page shows four role cards, so most visitors never see the full menu.

    Medium impactLow effort First result in 2 weeks Head of Go to Market

    First action

    Add Services to the main navigation.

    How to measure it

    Service inquiries

    Now
    –
    Target
    +50%
    By
    60 days

    Steps

    1. 1

      Add to navigation

      Services between Product and About.

      Web engineer 1 hour

    2. 2

      Link role cards

      Product page cards open the filtered list.

      Web engineer half a day

    3. 3

      Write service pages

      Who it suits, how it works, typical size.

      Head of Go-to-Market 1 week

    4. 4

      Add inquiry buttons

      Prefilled with the service.

      Web engineer 1 day

    5. 5

      Tag inquiries

      Record which service started each request.

      Head of Go-to-Market 1 day

    6. 6

      Review demand

      Which services draw interest, for pricing and packaging.

      Head of Go-to-Market monthly

    First result
    2 weeks
    Cost
    Staff time only
    Owner
    Head of Go to Market

    Smallest test

    Add the nav link today and compare Services visits next week.

    Why it works

    Add Services to the main navigation, link each product-page role card to the matching filtered list ("I buy compute", "I operate compute", "I finance compute"), and give each service a short page with who it suits and a direct inquiry button.

    Visitors who arrive with a specific problem, such as a deposit bridge or tail-period protection, find the exact product name and inquire about it, which also tells the desk what they want.

    Risks and how to handle them

    • Some services are not ready to sell

      Mark them "by request" rather than hiding them.

    Evidence

    • Your websiteServices page (/product-registry), linked only from the footer: 13 services by role, including compute sourcing, a cash-settled price hedge, prepayment and full contract financing, unused capacity resale, GCAT, bankable offtake origination, tail-period price protection, remarketing and "Compute index and market-data licensing", with a five-step inquiry flow.Source
    • Your websiteHomepage (checked November 9, 2026): "Compute has a price. Now it has a market."; the Liquid Compute Index (LCI) is "a daily reference price for A100, H100, H200, B200, and B300s"; the exchange and clearinghouse are "coming soon" and hedges go through the OTC desk; Buy, Sell, Trade and Finance buttons; "What We're Seeing" shows posts from October 20, 14 and 8; no customer logos, funding or volume figures; Services and GPU testing are linked only in the footer.Source
  • 10

    The seed announcement, "Announcing Liquid Compute: A $15M Seed Round...", lives at /insights/compute-is-a-grid-not-a-barrel, so its URL says nothing about funding. In the sitemap, the $250M post published October 6 carries an October 31 lastmod, three posts share the same October 30 timestamp, and the 16 core pages (homepage, product, services forms) have no lastmod at all. Search engines use these signals to decide what to recrawl.

    Low impactLow effort First result in 2 weeks Web engineer

    First action

    Add the redirect and new slug for the seed post.

    How to measure it

    Brand-query impressions

    Now
    –
    Target
    +25%
    By
    60 days

    Steps

    1. 1

      Rename the slug

      Seed post to a descriptive URL.

      Web engineer 1 hour

    2. 2

      Redirect

      301 from the old URL.

      Web engineer 30 minutes

    3. 3

      Fix lastmod

      Use each page's real last edit.

      Web engineer 2 hours

    4. 4

      Cover core pages

      Add lastmod and any missing pages.

      Web engineer 1 hour

    5. 5

      Add article data

      datePublished and dateModified on posts.

      Web engineer 2 hours

    6. 6

      Resubmit

      Sitemap in Search Console; watch coverage.

      Web engineer 2 weeks

    First result
    2 weeks
    Cost
    $0
    Owner
    Web engineer

    Smallest test

    Rename the seed post URL and redirect it today.

    Why it works

    Move the seed post to /insights/announcing-15m-seed (with a permanent redirect from the old URL), set lastmod to each post's real last edit, add lastmod to core pages, and add the Services and GPU testing pages if any are missing. Add publish and updated dates in the posts' structured data.

    People searching for the company's funding or partners find the right page, and accurate dates help new research get crawled quickly.

    Risks and how to handle them

    • Broken inbound links

      Keep the permanent redirect indefinitely.

    Evidence

    • Your websiteSitemap (28 URLs, checked November 9, 2026): 12 Insights posts carry lastmod dates, but the $250M post published October 6 shows October 31 and three posts share one October 30 timestamp; the 16 core pages have no lastmod; the seed announcement lives at /insights/compute-is-a-grid-not-a-barrel.Source
    • Your websiteSeed announcement (September 15, 2026), published at /insights/compute-is-a-grid-not-a-barrel: a $15 million seed round co-led by Chemistry and FirstMark, with K8 Capital, Night Capital, TrueBridge, Y Combinator and others; "a Y Combinator W25 company based in New York"; partnerships with Susquehanna, BGC and Wintermute; the PMEX Markets footnote is not explained.Source
    • Your websiteInsights listing: 13 posts dated September 2 to October 21, 2026, in Trading, Credit, Announcements and Substack (one entry links out to the CEO's Substack); the listing shows only date and category; no newsletter sign-up or RSS link on the page.Source

This quarter 11

The core plays for the next three months.

  • 07

    Liquid Compute published 13 Insights posts between September 2 and October 21, 2026, but the listing has no newsletter sign-up or RSS link, and one entry sends readers off to the CEO's external Substack. Silicon Data offers a newsletter on its homepage. Every reader who likes a credit note today has to remember to come back.

    High impactLow effort First result in 1 month Chief Product Officer

    First action

    Add a research-notes sign-up box to the Insights listing.

    How to measure it

    Email subscribers

    Now
    –
    Target
    1,000
    By
    90 days

    Steps

    1. 1

      Pick the tool

      Email platform with double opt-in.

      Web engineer 1 day

    2. 2

      Add sign-ups

      Insights, every post and /index.

      Web engineer 2 days

    3. 3

      Automate the close

      Daily email from the public print.

      Senior Research Engineer 1 week

    4. 4

      Add RSS

      Feed for Insights.

      Web engineer half a day

    5. 5

      Bring Substack home

      Cross-post with a canonical link to the site.

      CEO and Co-Founder ongoing

    6. 6

      Measure

      Subscribers, opens and access requests from email.

      Chief Product Officer monthly

    First result
    1 month
    Cost
    Email tool subscription
    Owner
    Chief Product Officer

    Smallest test

    Add a research sign-up box and count sign-ups in two weeks.

    Why it works

    Add a two-option sign-up on Insights, every post and the public index page: "Daily LCI close" (one line per GPU with the change and a link) and "Research notes" (new posts only). Add an RSS feed. Bring Substack posts onto the site or cross-post them, so subscribers grow on the company's own list.

    A daily price email is a habit-forming touchpoint for traders and lenders, and each email is a reminder that the desk can quote against that number.

    Risks and how to handle them

    • Daily email fatigue

      Let subscribers choose daily or weekly.

    Builds on

    Evidence

    • Your websiteInsights listing: 13 posts dated September 2 to October 21, 2026, in Trading, Credit, Announcements and Substack (one entry links out to the CEO's Substack); the listing shows only date and category; no newsletter sign-up or RSS link on the page.Source
    • Competitor researchSilicon Data's homepage lists H100, H200, A100, B200, B300 and MI300X indices "Built to financial-index standards and distributed on the Bloomberg and Refinitiv", claims "Over 3.5 million datapoints", sells benchmarking, forecasting and datasets, and offers a newsletter sign-up.Source
    • Your websiteHomepage (checked November 9, 2026): "Compute has a price. Now it has a market."; the Liquid Compute Index (LCI) is "a daily reference price for A100, H100, H200, B200, and B300s"; the exchange and clearinghouse are "coming soon" and hedges go through the OTC desk; Buy, Sell, Trade and Finance buttons; "What We're Seeing" shows posts from October 20, 14 and 8; no customer logos, funding or volume figures; Services and GPU testing are linked only in the footer.Source
  • 11

    The GPU Offtaker Credit Map scores 169 of 200 private AI compute buyers, finds only 38 can support a multi-year offtake today, publishes its weights, keeps the company-level dataset "available on request", and promises to "re-score any company that shares data with us under NDA". Every company on that list is a potential compute buyer and facility borrower, and the ones scored low have a strong reason to share data. There is no button to do it.

    High impactMedium effort First result in 6 weeks Head of Go to Market

    First action

    Add a "Get scored under NDA" request button to the Credit Map post.

    How to measure it

    Companies requesting a score

    Now
    –
    Target
    20
    By
    1 quarter

    Steps

    1. 1

      Define the offer

      What a scored company receives and the NDA terms.

      Head of Go-to-Market 3 days

    2. 2

      Data checklist

      Inputs needed for each of the five factors.

      Senior Research Engineer 3 days

    3. 3

      Report template

      Score, tier, levers and suitable structures.

      Senior Research Engineer 1 week

    4. 4

      Add the buttons

      Score request for buyers, dataset request for lenders.

      Web engineer 1 day

    5. 5

      Reach out

      Invite listed companies without a score to share data.

      Head of Go-to-Market 4 weeks

    6. 6

      Hand off

      Each report ends with a capacity or financing call.

      Head of Go-to-Market ongoing

    First result
    6 weeks
    Cost
    Research time per report
    Owner
    Head of Go to Market

    Smallest test

    Email ten unscored companies an offer to be scored under NDA.

    Why it works

    Add "Get your company scored" to the Credit Map: a short NDA request, a secure data checklist (funding, commitments, revenue), and a private score report with what would move them up a tier and which structures (prepayment, hedge, guarantee) fit their tier. Offer the dataset to lenders and operators through a request form that captures who they are.

    A score is a personal, high-stakes output that companies will trade data for; the report naturally leads to capacity, financing and hedging conversations, which is the repeatable pipeline the GTM role is hired to build.

    Risks and how to handle them

    • Score read as a credit rating

      Keep the "It is not a rating" language on every report.

    Builds on

    Evidence

    • Your website"The GPU Offtaker Credit Map: 200 Private AI Compute Buyers, Ranked" (September 28, 2026): the Liquid Compute Offtaker Score (0-100, five weighted factors, published weights); 169 of 200 buyers scored; "38 of 169 scored companies can support a multi-year offtake today"; the company-level dataset is "available on request"; "We will update it each edition, and re-score any company that shares data with us under NDA."Source
    • Your website"Announcing a $250M Facility to Fund Compute Buyers" (October 6, 2026): a prepayment facility with K8 Capital that funds upfront deposits on multi-year contracts, "available to qualifying buyers contracting capacity through Liquid Compute", with each draw secured by prepaid capacity that Liquid Compute verifies; the post gives no eligibility criteria or application path beyond "visit liquidcompute.com".Source
    • HiringHead of Go-to-Market (New York City or San Francisco, $150,000-$200,000 base plus 0.10%-0.30% equity): owns the full commercial cycle from market mapping and outbound to expansion; "Make the market legible" without "hiding behind jargon"; builds outbound, referral, ecosystem, event and partnership motions, forecasting and win/loss review; success means referenceable customers and growing repeat transaction volume; helpful: building category presence through content, partnerships or events.Source
  • 12

    GCAT, Liquid Compute's acceptance test for NVIDIA A10, H100, H200, B200 and B300 nodes, runs with "No Liquid Compute account needed" and produces a pass/fail report kept in the browser for 30 days. Every operator who runs it has just proved they control idle, healthy GPUs, which is exactly the supply the new Head of Compute Supply must find. The page has no next step after the report.

    High impactMedium effort First result in 1 month Head of Compute Supply

    First action

    Add a "List this capacity" button to the report screen.

    How to measure it

    GCAT-sourced supply leads

    Now
    –
    Target
    25% of tests
    By
    1 quarter

    Steps

    1. 1

      Measure tests

      Completed tests and configurations so far.

      Head of Compute Supply 1 day

    2. 2

      Add next steps

      List capacity and indicative price buttons on the report.

      Web engineer 1 week

    3. 3

      Email the report

      Optional email copy beyond 30 days.

      Web engineer 3 days

    4. 4

      Prefill the form

      GPU model, count and pass status.

      Web engineer 2 days

    5. 5

      Promote GCAT

      To operators and data centers on the supply map.

      Head of Compute Supply 4 weeks

    6. 6

      Follow up

      Call every listing within one business day.

      Head of Compute Supply ongoing

    First result
    1 month
    Cost
    Two weeks of engineering
    Owner
    Head of Compute Supply

    Smallest test

    Add the button and email the last month's testers who left contact details.

    Why it works

    At the end of each report add "List this capacity" (prefilling GPU model, count and test result into a short seller form) and "Get an indicative price" from the index. Offer to email the report so it outlives the 30-day browser window, which also captures a contact. Promote GCAT to operators as a free pre-sale or pre-financing check.

    The test reaches operators at the moment they are preparing hardware for a customer or lender, and the report removes the first diligence step for both sides.

    Risks and how to handle them

    • Operators see the tool as a sales trap

      Keep the test free and the listing optional.

    Evidence

    • Your websiteGPU testing (GCAT): acceptance tests for NVIDIA A10, H100, H200, B200 and B300 nodes and clusters (DCGM diagnostics, GEMM, memory bandwidth, NCCL); "No Liquid Compute account needed"; "Full diagnostics may take 90 minutes or longer"; the pass/fail report stays in the same browser for 30 days and can be saved as a PDF. No price is stated.Source
    • HiringHead of Compute Supply (New York City or San Francisco, $150,000-$200,000 base): "Build the supply network behind a liquid market for GPU capacity"; a supply map by provider, accelerator, region, topology, availability date and contract term; assess counterparty quality and technical readiness; build pricing benchmarks, deal comparisons and supplier records.Source
    • Your websiteServices page (/product-registry), linked only from the footer: 13 services by role, including compute sourcing, a cash-settled price hedge, prepayment and full contract financing, unused capacity resale, GCAT, bankable offtake origination, tail-period price protection, remarketing and "Compute index and market-data licensing", with a five-step inquiry flow.Source
  • 13

    Contracts are "priced per GPU-hour, so offers compare across sellers", but price is only half the comparison: buyers also need to know the hardware performs. Liquid Compute already owns the test (GCAT) and already verifies capacity before each draw on the $250M facility. Making the test result part of every offer turns that diligence into a visible advantage.

    Medium impactMedium effort First result in 1 quarter Head of Compute Supply

    First action

    Add the GCAT result to the next ten quotes sent to buyers.

    How to measure it

    Quotes accepted

    Now
    –
    Target
    +15% for tested offers
    By
    2 quarters

    Steps

    1. 1

      Define the mark

      Which tests and how recent.

      Head of Compute Supply 3 days

    2. 2

      Add to quotes

      Result, date and configuration.

      Head of Compute Supply 1 week

    3. 3

      Match rule

      Tested capacity first at equal price.

      Chief Product Officer 1 week

    4. 4

      Tie to financing

      A recent pass for facility-funded capacity.

      CEO and Co-Founder 2 weeks

    5. 5

      Explain it

      A short section on the product page.

      Head of Go-to-Market 1 day

    6. 6

      Report

      Acceptance rate and delivery issues by tested vs untested.

      Head of Compute Supply quarterly

    First result
    1 quarter
    Cost
    Staff time and test runs
    Owner
    Head of Compute Supply

    Smallest test

    Show GCAT results on ten quotes and compare acceptance.

    Why it works

    Add a "GCAT passed" mark, with test date and configuration, to quotes the desk sends buyers, and match tested capacity first at equal price. Require a recent GCAT pass for capacity financed through the prepayment facility, and publish the share of delivered capacity that passed.

    Verified supply reduces delivery disputes and speeds buyer sign-off, and sellers have a reason to test (and list) through Liquid Compute.

    Risks and how to handle them

    • Fewer offers if testing is required

      Make it a preference, not a requirement, outside financing.

    Builds on

    Evidence

    • Your websiteGPU testing (GCAT): acceptance tests for NVIDIA A10, H100, H200, B200 and B300 nodes and clusters (DCGM diagnostics, GEMM, memory bandwidth, NCCL); "No Liquid Compute account needed"; "Full diagnostics may take 90 minutes or longer"; the pass/fail report stays in the same browser for 30 days and can be saved as a PDF. No price is stated.Source
    • Your websiteProduct page: "Priced. Matched. Delivered."; standard contracts for A100, H100, H200, B200 and B300, "priced per GPU-hour, so offers compare across sellers", running month-to-month or for a fixed term; the "See current pricing" button leads to the login page; no fees or prices are shown.Source
    • Your website"Announcing a $250M Facility to Fund Compute Buyers" (October 6, 2026): a prepayment facility with K8 Capital that funds upfront deposits on multi-year contracts, "available to qualifying buyers contracting capacity through Liquid Compute", with each draw secured by prepaid capacity that Liquid Compute verifies; the post gives no eligibility criteria or application path beyond "visit liquidcompute.com".Source
  • 14

    In July 2026 public H100 price trackers differed by about eight times ($0.94 against $7.40 an hour) because they measure different things: marketplace spot listings, hyperscaler on-demand benchmarks, different terms and windows. Silicon Data and Ornn each describe their own index, and Liquid Compute's CFTC letter lays out what a settlement index should be built from. Buyers and lenders searching for "H100 price" have nobody explaining the differences.

    Medium impactLow effort First result in 1 month Chief Product Officer

    First action

    Gather each provider's own published description of its index.

    How to measure it

    Explainer visits

    Now
    –
    Target
    2,000 a month
    By
    1 quarter

    Steps

    1. 1

      Collect sources

      Each provider's own methodology text.

      Senior Research Engineer 3 days

    2. 2

      Build the table

      Inputs, contract type, frequency, distribution.

      Senior Research Engineer 3 days

    3. 3

      Write the guide

      Why prices differ and which index fits which use.

      Chief Product Officer 1 week

    4. 4

      Legal review

      Trademarks and factual claims.

      Legal counsel 2 days

    5. 5

      Publish and pitch

      To reporters covering compute futures.

      Head of Go-to-Market 1 week

    6. 6

      Refresh

      Quarterly, with corrections log.

      Senior Research Engineer quarterly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Chief Product Officer

    Smallest test

    Publish the comparison table alone as an Insights post.

    Why it works

    A fair, dated explainer that sets out what each public index or tracker measures (inputs, contract type, term, update frequency, distribution), why the numbers differ, and which suits budgeting, hedging or lending. Show where the LCI sits, link its methodology, and invite corrections from the other providers.

    Explaining the category is the GTM brief ("Make the market legible"), and a neutral guide earns links from the press and AI assistants answering "what does an H100 cost".

    Risks and how to handle them

    • Seen as attacking rivals

      Quote each provider's own description and invite corrections.

    Builds on

    Evidence

    • PressImplicator (July 29, 2026): public H100 price trackers differ by about eight times ($0.94 against $7.40 an hour) because they measure different things; CME announced compute futures that would settle against Silicon Data's indices, subject to regulatory review.Source
    • Competitor researchSilicon Data's homepage lists H100, H200, A100, B200, B300 and MI300X indices "Built to financial-index standards and distributed on the Bloomberg and Refinitiv", claims "Over 3.5 million datapoints", sells benchmarking, forecasting and datasets, and offers a newsletter sign-up.Source
    • Competitor researchOrnn's homepage: "The market for compute. The index for intelligence."; "Settled trades from partners build the Ornn Compute Price Index. Updated hourly"; quotes from ICE and Kalshi on using its index; compute contracts "from an hour to multiple years".Source
    • Your websiteCFTC comment letter (dated October 20, 2026): treats the reference index as the main manipulation risk; recommends executed transactions as index inputs, IOSCO-based governance, minimum transaction and contributor thresholds, and no index administrator trading contracts on its own index; the authors are "not aware of any" compute price reporting agency under "an external assurance regime"; PMEX Markets, Inc. has applied for DCM designation and PMEX Clearing, Inc. for DCO registration, and PMEX "intends to operate under the name Liquid Compute".Source
  • 15

    The contact page promises onboarding with API keys for market data "on paid plans", and the Services registry lists "Compute index and market-data licensing", but no page says what the plans include or who they are for. Silicon Data sells named datasets, forecasting and benchmarking openly. Trading firms and lenders who want data, not a desk relationship, have nowhere to start.

    Medium impactMedium effort First result in 6 weeks Head of Go to Market

    First action

    List what current paid users actually receive.

    How to measure it

    Data trials

    Now
    –
    Target
    15
    By
    1 quarter

    Steps

    1. 1

      Inventory

      Data, history and API access already provided.

      Chief Product Officer 2 days

    2. 2

      Define plans

      Two or three tiers by use.

      Head of Go-to-Market 1 week

    3. 3

      Licensing terms

      Display, internal use, redistribution, settlement.

      Legal counsel 2 weeks

    4. 4

      Build /data

      Plans, samples and trial request.

      Web engineer 1 week

    5. 5

      Offer trials

      Two-week API trials for qualified requests.

      Head of Go-to-Market ongoing

    6. 6

      Review

      Trials, conversions and price feedback.

      Head of Go-to-Market monthly

    First result
    6 weeks
    Cost
    Staff and legal time
    Owner
    Head of Go to Market

    Smallest test

    Add a plan comparison table to the Services licensing entry.

    Why it works

    Publish /data with two or three plans (for example Public: delayed daily prints; Professional: real-time prints, history and API; Enterprise: redistribution and settlement licensing), what each includes and how to start a trial. Prices can be "from" figures or "talk to us", but the contents should be explicit.

    A clear data product gives a revenue line before the exchange launches, and data subscribers become the natural first users of listed contracts.

    Risks and how to handle them

    • Pricing anchored too low early

      Publish contents and "from" prices only.

    Builds on

    Evidence

    • Your websiteRequest forms: /finance-compute and /buy-compute mark as required Interest, Name, Work email, GPU, Other GPU type, Indices, Quantity, Desired start date and Duration in months, for lenders as well as buyers. The /contact form asks only Name, Work email, Company, "I want to" (Buy compute, Sell capacity, Take a position, Finance a buildout) and an optional note, then promises "a short call" and onboarding that includes API keys for market data on paid plans.Source
    • Your websiteServices page (/product-registry), linked only from the footer: 13 services by role, including compute sourcing, a cash-settled price hedge, prepayment and full contract financing, unused capacity resale, GCAT, bankable offtake origination, tail-period price protection, remarketing and "Compute index and market-data licensing", with a five-step inquiry flow.Source
    • Competitor researchSilicon Data's homepage lists H100, H200, A100, B200, B300 and MI300X indices "Built to financial-index standards and distributed on the Bloomberg and Refinitiv", claims "Over 3.5 million datapoints", sells benchmarking, forecasting and datasets, and offers a newsletter sign-up.Source
  • 16

    The Head of Go-to-Market is hired to build outbound, referral, ecosystem, event and partnership motions with a lean team. GPU deals already pass through brokers, data-center advisers and debt arrangers, and Liquid Compute offers products they can't, such as prepayment financing from a $250M facility, tail-period price protection and bankable offtake origination. There is no referral or partner page today.

    Medium impactMedium effort First result in 1 quarter Head of Go to Market

    First action

    List the ten brokers and advisers who have already sent deals or questions.

    How to measure it

    Introduced qualified deals

    Now
    –
    Target
    10 a quarter
    By
    2 quarters

    Steps

    1. 1

      Map introducers

      Brokers, advisers and arrangers already in contact.

      Head of Go-to-Market 1 week

    2. 2

      Set fees

      By service, on closed transactions.

      CEO and Co-Founder 1 week

    3. 3

      Legal terms

      Introducer agreement and disclosures.

      Legal counsel 2 weeks

    4. 4

      Partner page

      Programme, fees and registration.

      Web engineer 3 days

    5. 5

      Deal registration

      Simple tracking so introducers are protected.

      Head of Go-to-Market 1 week

    6. 6

      Monthly wants list

      What the desk needs now.

      Head of Compute Supply monthly

    First result
    1 quarter
    Cost
    Fees on closed deals only
    Owner
    Head of Go to Market

    Smallest test

    Offer five brokers a fee on capacity deals for one quarter.

    Why it works

    Launch a simple introducer programme: a partner page, a registration form, a published fee for closed transactions (by service), deal registration so introducers are protected, and a monthly note on what the desk is looking for (GPU types, terms, borrowers).

    Introducers bring qualified deals at low cost, and the fee is paid only when a transaction closes, which suits a lean team building pipeline.

    Risks and how to handle them

    • Regulatory limits on paying introducers for derivatives

      Have counsel decide which services can carry introducer fees.

    Builds on

    Evidence

    • HiringHead of Go-to-Market (New York City or San Francisco, $150,000-$200,000 base plus 0.10%-0.30% equity): owns the full commercial cycle from market mapping and outbound to expansion; "Make the market legible" without "hiding behind jargon"; builds outbound, referral, ecosystem, event and partnership motions, forecasting and win/loss review; success means referenceable customers and growing repeat transaction volume; helpful: building category presence through content, partnerships or events.Source
    • Your websiteServices page (/product-registry), linked only from the footer: 13 services by role, including compute sourcing, a cash-settled price hedge, prepayment and full contract financing, unused capacity resale, GCAT, bankable offtake origination, tail-period price protection, remarketing and "Compute index and market-data licensing", with a five-step inquiry flow.Source
    • Your website"Announcing a $250M Facility to Fund Compute Buyers" (October 6, 2026): a prepayment facility with K8 Capital that funds upfront deposits on multi-year contracts, "available to qualifying buyers contracting capacity through Liquid Compute", with each draw secured by prepaid capacity that Liquid Compute verifies; the post gives no eligibility criteria or application path beyond "visit liquidcompute.com".Source
  • 17

    Capacity contracts run month-to-month or for a fixed term, and success for the GTM role includes growing repeat transaction volume. The Services registry already lists unused capacity resale and tail-period price protection, and SF Compute has made "Sell back compute you don't use" its headline product. The moment a contract nears its end, or a buyer's usage drops, is when they decide whether to come back.

    High impactLow effort First result in 1 quarter Head of Go to Market

    First action

    List every live contract with its end date.

    How to measure it

    Buyers transacting again

    Now
    –
    Target
    60%
    By
    2 quarters

    Steps

    1. 1

      Build the calendar

      All contracts and end dates.

      Head of Go-to-Market 2 days

    2. 2

      Review template

      Price against index, usage and options.

      Chief Product Officer 3 days

    3. 3

      Resale path

      Match unused capacity to new buyers.

      Head of Compute Supply 2 weeks

    4. 4

      Hedge offer

      Indicative price for the next term.

      Head of Go-to-Market per contract

    5. 5

      Mid-term check

      Usage review and resale offer.

      Head of Go-to-Market per contract

    6. 6

      Track

      Renewals, resales and churn reasons.

      Head of Go-to-Market monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Head of Go to Market

    Smallest test

    Send the review to the next three contracts ending.

    Why it works

    Set up a term calendar: 60 days before each term ends, send the buyer a one-page review with the delivered price against the index, a renewal quote, a resale option for unused capacity and a hedge for the next term. Check usage at mid-term and offer resale if it is low.

    Repeat transactions cost far less to win than new ones, and offering resale keeps a buyer who would otherwise sit on unused capacity and leave unhappy.

    Risks and how to handle them

    • Resale undercuts the desk's own supply

      Price resold capacity off the index like any other offer.

    Evidence

    • Your websiteProduct page: "Priced. Matched. Delivered."; standard contracts for A100, H100, H200, B200 and B300, "priced per GPU-hour, so offers compare across sellers", running month-to-month or for a fixed term; the "See current pricing" button leads to the login page; no fees or prices are shown.Source
    • Your websiteServices page (/product-registry), linked only from the footer: 13 services by role, including compute sourcing, a cash-settled price hedge, prepayment and full contract financing, unused capacity resale, GCAT, bankable offtake origination, tail-period price protection, remarketing and "Compute index and market-data licensing", with a five-step inquiry flow.Source
    • Competitor researchSF Compute's homepage: "Supercomputers you can sublease"; a Resell product to "Sell back compute you don't use", with example reserved and realized per-GPU-hour prices shown without a login.Source
    • HiringHead of Go-to-Market (New York City or San Francisco, $150,000-$200,000 base plus 0.10%-0.30% equity): owns the full commercial cycle from market mapping and outbound to expansion; "Make the market legible" without "hiding behind jargon"; builds outbound, referral, ecosystem, event and partnership motions, forecasting and win/loss review; success means referenceable customers and growing repeat transaction volume; helpful: building category presence through content, partnerships or events.Source
  • 18

    The October 20 credit note shows a cash-settled put on the LCI lifting a $15M cluster's advance from 65% to 85%, with the floor saving the operator about $0.45M a year for about $0.17M a year, and invites lenders with GPU books to request "a mark and hedge". That invitation is a line in a post, not a product: there is no page, sample or way to ask beyond a general email.

    High impactMedium effort First result in 6 weeks Head of Go to Market

    First action

    Build a sample mark report from the post's $15M illustration.

    How to measure it

    Lender books marked

    Now
    –
    Target
    8
    By
    1 quarter

    Steps

    1. 1

      Define the mark

      Inputs, outputs and confidentiality.

      Senior Research Engineer 1 week

    2. 2

      Sample report

      Anonymised, from the $15M illustration.

      Senior Research Engineer 3 days

    3. 3

      Lender page

      Offer, sample, checklist and booking.

      Web engineer 3 days

    4. 4

      Compliance check

      Eligible contract participant language.

      Legal counsel 3 days

    5. 5

      Outreach

      Private credit funds and banks with GPU exposure.

      Head of Go-to-Market 4 weeks

    6. 6

      Price floors

      For the riskiest positions in each mark.

      Head of Go-to-Market per mark

    First result
    6 weeks
    Cost
    Research time per mark
    Owner
    Head of Go to Market

    Smallest test

    Offer three lenders a mark of one cluster each.

    Why it works

    Package it: a page for lenders explaining the free mark (each financed cluster valued against the relevant LCI and term, with residual exposure), a sample anonymised mark report, a secure upload checklist and a booking link. Each mark ends with a priced floor for the riskiest positions.

    A free, specific valuation of a lender's own book is hard to ignore when credit is tightening, and it leads directly to the hedge and to lenders using the LCI in their models.

    Risks and how to handle them

    • Lenders unwilling to share position data

      Accept anonymised cluster-level data under NDA.

    Evidence

    • Your website"Compute credit is tightening" (October 20, 2026): a cash-settled put on the LCI as a residual value floor; in an illustrative $15M cluster the advance rises from 65% to 85%, and the floor saves the operator about $0.45M a year for about $0.17M a year; the options are OTC transactions for eligible contract participants, not exchange-listed or cleared; lenders are invited to request a mark and hedge.Source
    • Your website"The GPU Offtaker Credit Map: 200 Private AI Compute Buyers, Ranked" (September 28, 2026): the Liquid Compute Offtaker Score (0-100, five weighted factors, published weights); 169 of 200 buyers scored; "38 of 169 scored companies can support a multi-year offtake today"; the company-level dataset is "available on request"; "We will update it each edition, and re-score any company that shares data with us under NDA."Source
    • Your website"Announcing a $250M Facility to Fund Compute Buyers" (October 6, 2026): a prepayment facility with K8 Capital that funds upfront deposits on multi-year contracts, "available to qualifying buyers contracting capacity through Liquid Compute", with each draw secured by prepaid capacity that Liquid Compute verifies; the post gives no eligibility criteria or application path beyond "visit liquidcompute.com".Source
  • 19

    Liquid Compute's CFTC letter recommends a staff roundtable on compute derivatives, and the company already convenes all four sides of the market. The GTM role lists events and category presence among its tools. The team is in New York and San Francisco, where most compute financiers and buyers are, but there is no events page.

    Medium impactMedium effort First result in 1 quarter CEO and Co Founder

    First action

    Invite 15 people for the first New York session.

    How to measure it

    Attendee-sourced opportunities

    Now
    –
    Target
    5 a quarter
    By
    2 quarters

    Steps

    1. 1

      Set the format

      Agenda, rules on attribution, size.

      CEO and Co-Founder 2 days

    2. 2

      Build the list

      Balanced across the four sides.

      Head of Go-to-Market 1 week

    3. 3

      Prepare the data

      Index moves and credit review.

      Senior Research Engineer monthly

    4. 4

      Host

      Alternate New York and San Francisco.

      CEO and Co-Founder monthly

    5. 5

      Publish the note

      Unattributed summary on Insights.

      Chief Product Officer monthly

    6. 6

      Follow up

      Within a week with each attendee.

      Head of Go-to-Market monthly

    First result
    1 quarter
    Cost
    Venue and catering, a few thousand dollars a session
    Owner
    CEO and Co Founder

    Smallest test

    Host one dinner for 12 in New York.

    Why it works

    A monthly, invite-only roundtable alternating between New York and San Francisco (with a remote option): a 15-minute review of the month's index moves and credit conditions, one guest from each side of the market, then open discussion. Publish a short note afterwards, without attributing comments.

    Putting operators, buyers and lenders in one room is the market Liquid Compute wants to run; each session produces relationships, deal flow and content.

    Risks and how to handle them

    • Competitors sharing sensitive information

      Have counsel set discussion rules and keep prices historical.

    Evidence

    • HiringHead of Go-to-Market (New York City or San Francisco, $150,000-$200,000 base plus 0.10%-0.30% equity): owns the full commercial cycle from market mapping and outbound to expansion; "Make the market legible" without "hiding behind jargon"; builds outbound, referral, ecosystem, event and partnership motions, forecasting and win/loss review; success means referenceable customers and growing repeat transaction volume; helpful: building category presence through content, partnerships or events.Source
    • Your websiteCFTC comment letter (dated October 20, 2026): treats the reference index as the main manipulation risk; recommends executed transactions as index inputs, IOSCO-based governance, minimum transaction and contributor thresholds, and no index administrator trading contracts on its own index; the authors are "not aware of any" compute price reporting agency under "an external assurance regime"; PMEX Markets, Inc. has applied for DCM designation and PMEX Clearing, Inc. for DCO registration, and PMEX "intends to operate under the name Liquid Compute".Source
    • Your websiteSeed announcement (September 15, 2026), published at /insights/compute-is-a-grid-not-a-barrel: a $15 million seed round co-led by Chemistry and FirstMark, with K8 Capital, Night Capital, TrueBridge, Y Combinator and others; "a Y Combinator W25 company based in New York"; partnerships with Susquehanna, BGC and Wintermute; the PMEX Markets footnote is not explained.Source
  • 20

    The Insights listing shows only a date and category; authors appear only inside each post's disclosure or contact line, and calls to action end in individual staff email addresses. One listing entry sends readers to the CEO's external Substack. The research is the company's strongest marketing, but readers can't follow a person or a theme (credit, hedging, market structure).

    Medium impactLow effort First result in 1 month Chief Product Officer

    First action

    Add bylines to the listing and each post header.

    How to measure it

    Insights pages per session

    Now
    –
    Target
    +30%
    By
    60 days

    Steps

    1. 1

      Bylines

      On the listing and post headers.

      Web engineer 1 day

    2. 2

      Author pages

      Role, focus and posts.

      Chief Product Officer 2 days

    3. 3

      Topic hubs

      Credit, Trading, Market Structure.

      Chief Product Officer 3 days

    4. 4

      Related research

      Three links at the end of each post.

      Web engineer 1 day

    5. 5

      Route CTAs

      Role-based request buttons instead of personal emails.

      Web engineer 1 day

    6. 6

      Measure

      Pages per session and requests from posts.

      Chief Product Officer monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Chief Product Officer

    Smallest test

    Add bylines and related links to the five most-read posts.

    Why it works

    Add bylines and short author pages (role, focus, posts, LinkedIn), topic hubs for Credit, Trading and Market Structure with a one-paragraph introduction and the key posts, and "Related research" at the end of each post. Replace personal email CTAs with role-routed request buttons.

    Named experts earn more trust and shares than an anonymous company blog, topic hubs rank for category terms, and routed CTAs stop leads sitting in one inbox.

    Risks and how to handle them

    • Staff turnover leaves orphaned author pages

      Keep "Liquid Compute Research" as the publisher on every post.

    Evidence

    • Your websiteInsights listing: 13 posts dated September 2 to October 21, 2026, in Trading, Credit, Announcements and Substack (one entry links out to the CEO's Substack); the listing shows only date and category; no newsletter sign-up or RSS link on the page.Source
    • Your websiteHedge Converter (/trade/hedge-converter): "Translate your compute economics into an indicative H100 index hedge", with paths for "New to this", "I know the basics" and "I trade them"; the October 20 credit post says it needs a free account; the September 8 primer "Hedging Compute Prices" (128 GPUs for a year at $4.40 an hour, about $4.93M) does not link to it and ends with email CTAs for buyers, operators and lenders.Source
    • Your website"Compute credit is tightening" (October 20, 2026): a cash-settled put on the LCI as a residual value floor; in an illustrative $15M cluster the advance rises from 65% to 85%, and the floor saves the operator about $0.45M a year for about $0.17M a year; the options are OTC transactions for eligible contract participants, not exchange-listed or cleared; lenders are invited to request a mark and hedge.Source

Big bets 3

Larger plays that can change the trajectory.

  • 21

    Liquid Compute's own CFTC letter says the authors are "not aware of any" compute price reporting agency subject to "an external assurance regime", and recommends IOSCO-based governance for settlement indices. With public H100 prices differing eightfold and CME and ICE contracts referencing other providers' indices, the index that can prove its integrity has the strongest claim to settle regulated contracts.

    High impactHigh effort First result in 2 quarters CEO and Co Founder

    First action

    Ask two audit or assurance firms for a scope and quote.

    How to measure it

    Independent assurance completed

    Now
    –
    Target
    Published report
    By
    2 quarters

    Steps

    1. 1

      Gap review

      LCI governance against the IOSCO principles.

      Senior Research Engineer 3 weeks

    2. 2

      Fix gaps

      Policies, controls, contributor checks and records.

      Chief Product Officer 6 weeks

    3. 3

      Oversight committee

      Include independent members.

      CEO and Co-Founder 1 month

    4. 4

      Engage reviewer

      Independent assurance firm.

      CEO and Co-Founder 2 weeks

    5. 5

      Review

      Design and operating effectiveness.

      Legal counsel 2 months

    6. 6

      Publish

      Statement of compliance and report on /index.

      Chief Product Officer 1 week

    7. 7

      Use it

      In lender, trading firm and exchange conversations.

      Head of Go-to-Market ongoing

    First result
    2 quarters
    Cost
    Assurance fees, likely tens of thousands of dollars, plus staff time
    Owner
    CEO and Co Founder

    Smallest test

    Publish a self-assessed statement of compliance first.

    Why it works

    Commission an independent assurance review of the LCI's design and operation against the IOSCO Principles for Financial Benchmarks, publish the administrator's statement of compliance and the reviewer's report, set up an oversight committee with outside members, and repeat the review yearly.

    It turns the company's regulatory argument into a fact no rival can currently claim, supports the PMEX designation process, and gives lenders and trading firms a reason to adopt the LCI as their reference.

    Risks and how to handle them

    • Review finds material gaps

      Do the gap review first and fix before engaging the reviewer.

    Builds on

    Evidence

    • Your websiteCFTC comment letter (dated October 20, 2026): treats the reference index as the main manipulation risk; recommends executed transactions as index inputs, IOSCO-based governance, minimum transaction and contributor thresholds, and no index administrator trading contracts on its own index; the authors are "not aware of any" compute price reporting agency under "an external assurance regime"; PMEX Markets, Inc. has applied for DCM designation and PMEX Clearing, Inc. for DCO registration, and PMEX "intends to operate under the name Liquid Compute".Source
    • PressImplicator (July 29, 2026): public H100 price trackers differ by about eight times ($0.94 against $7.40 an hour) because they measure different things; CME announced compute futures that would settle against Silicon Data's indices, subject to regulatory review.Source
    • Competitor researchSilicon Data's homepage lists H100, H200, A100, B200, B300 and MI300X indices "Built to financial-index standards and distributed on the Bloomberg and Refinitiv", claims "Over 3.5 million datapoints", sells benchmarking, forecasting and datasets, and offers a newsletter sign-up.Source
  • 22

    Silicon Data says its indices are distributed on Bloomberg and Refinitiv, and Ornn quotes ICE and Kalshi on using its index. The LCI is published only inside Liquid Compute's own login. Liquid Compute already lists "Compute index and market-data licensing" as a service, and its investors include trading firms; the index needs to be on the screens where hedging and lending decisions get made.

    High impactHigh effort First result in 2 quarters Head of Go to Market

    First action

    Ask the data vendors for their index-listing requirements.

    How to measure it

    Third-party platforms carrying the LCI

    Now
    –
    Target
    2
    By
    2 quarters

    Steps

    1. 1

      Requirements

      Listing criteria for each vendor.

      Head of Go-to-Market 2 weeks

    2. 2

      Data feed

      Reliable delivery format and history.

      Senior Research Engineer 1 month

    3. 3

      Licensing terms

      Display, redistribution and settlement use.

      Legal counsel 1 month

    4. 4

      Apply

      Two vendors, delayed close first.

      Head of Go-to-Market 1 month

    5. 5

      Launch

      Announce with tickers on /index.

      Chief Product Officer 1 week

    6. 6

      License

      Offer to venues and lenders.

      CEO and Co-Founder ongoing

    First result
    2 quarters
    Cost
    Vendor fees, legal time and engineering
    Owner
    Head of Go to Market

    Smallest test

    Put the delayed daily close on one data platform.

    Why it works

    Apply to list the LCI on one or two major market-data terminals and data platforms, starting with the delayed daily close, then license real-time data. Offer index licences to venues and lenders that need a settlement or valuation reference, under the governance in idea 21.

    Distribution makes the LCI a habit for the people who will trade and finance compute, and every terminal user who sees it is a potential desk client.

    Risks and how to handle them

    • Vendors require a longer history or more contributors

      Start with a data-platform listing while the history builds.

    Builds on

    Evidence

    • Competitor researchSilicon Data's homepage lists H100, H200, A100, B200, B300 and MI300X indices "Built to financial-index standards and distributed on the Bloomberg and Refinitiv", claims "Over 3.5 million datapoints", sells benchmarking, forecasting and datasets, and offers a newsletter sign-up.Source
    • Competitor researchOrnn's homepage: "The market for compute. The index for intelligence."; "Settled trades from partners build the Ornn Compute Price Index. Updated hourly"; quotes from ICE and Kalshi on using its index; compute contracts "from an hour to multiple years".Source
    • Your websiteServices page (/product-registry), linked only from the footer: 13 services by role, including compute sourcing, a cash-settled price hedge, prepayment and full contract financing, unused capacity resale, GCAT, bankable offtake origination, tail-period price protection, remarketing and "Compute index and market-data licensing", with a five-step inquiry flow.Source
  • 23

    Liquid Compute's contracts have a set start date and run month-to-month or for a fixed term, and its hedging posts price six-month and multi-year exposure. The About page's premise is that compute is traded "with no market price behind it". A spot index answers "what does an H100 cost today"; lenders and buyers signing three-year deals need to know what it costs over time, and the desk sees term prices nobody else publishes.

    High impactHigh effort First result in 2 quarters Chief Product Officer

    First action

    Count executed term contracts per GPU and tenor to see which points are publishable.

    How to measure it

    Lenders and buyers using the curve

    Now
    –
    Target
    25 organisations
    By
    3 quarters

    Steps

    1. 1

      Data audit

      Term contracts and quotes by GPU and tenor.

      Senior Research Engineer 2 weeks

    2. 2

      Method

      Interpolation, minimum inputs and disclosure.

      Senior Research Engineer 1 month

    3. 3

      Governance

      Same oversight as the LCI.

      CEO and Co-Founder 1 month

    4. 4

      Pilot

      Share privately with five lenders for feedback.

      Head of Go-to-Market 1 month

    5. 5

      Publish

      Delayed curve on /index, full curve in data plans.

      Chief Product Officer 2 weeks

    6. 6

      Use in products

      Reference for floors and prepayment pricing.

      Head of Go-to-Market ongoing

    First result
    2 quarters
    Cost
    Research and engineering time
    Owner
    Chief Product Officer

    Smallest test

    Publish an H100 6- and 12-month indicative point in one research note.

    Why it works

    From executed term contracts and desk quotes, publish a monthly indicative curve per GPU (spot, 6, 12, 24 and 36 months), with the number of inputs behind each point, under the same governance as the LCI. Put a delayed version on /index and the full curve in paid data plans.

    A term curve is the reference lenders need for residual value and buyers need for budgeting, and it is exactly what the residual-value floor and prepayment products are priced on.

    Risks and how to handle them

    • Too few long-dated trades

      Publish only tenors with enough inputs and show the count.

    Builds on

    Evidence

    • Your websiteProduct page: "Priced. Matched. Delivered."; standard contracts for A100, H100, H200, B200 and B300, "priced per GPU-hour, so offers compare across sellers", running month-to-month or for a fixed term; the "See current pricing" button leads to the login page; no fees or prices are shown.Source
    • Your websiteAbout page: "Compute gets bought and sold every day with no market price behind it."; the plan is a reference index, physical delivery, credit against contracted capacity and eventually an exchange; $15M seed co-led by FirstMark and Chemistry; no office location stated.Source
    • Your website"Compute credit is tightening" (October 20, 2026): a cash-settled put on the LCI as a residual value floor; in an illustrative $15M cluster the advance rises from 65% to 85%, and the floor saves the operator about $0.45M a year for about $0.17M a year; the options are OTC transactions for eligible contract participants, not exchange-listed or cleared; lenders are invited to request a mark and hedge.Source

Wildcards 1

Unexpected experiments, each with a cheap way to test it.

  • 24

    The Head of Compute Supply will build a supply map of capacity by provider, accelerator, region, topology, availability date and contract term. Aggregated, that is the data buyers and lenders most want and can't find: how much H100, H200 and B200 capacity becomes available each quarter. The Credit Map has already shown that a ranked, published dataset gets attention and invites companies to share data.

    Medium impactMedium effort First result in 2 quarters Head of Compute Supply

    First action

    Decide the minimum coverage needed before the first edition.

    How to measure it

    Operators in the snapshot

    Now
    –
    Target
    30
    By
    3 quarters

    Steps

    1. 1

      Build the map

      Provider, GPU, region, availability and term.

      Head of Compute Supply 2 months

    2. 2

      Set rules

      Aggregation so no provider is identifiable.

      Legal counsel 2 weeks

    3. 3

      Draft the snapshot

      Charts by GPU, region and quarter.

      Senior Research Engineer 2 weeks

    4. 4

      Invite operators

      Confidential contribution form.

      Head of Compute Supply ongoing

    5. 5

      Publish

      On Insights with a subscribe prompt.

      Chief Product Officer quarterly

    6. 6

      Match

      Connect contributed capacity with buyers.

      Head of Compute Supply ongoing

    First result
    2 quarters
    Cost
    Staff time only
    Owner
    Head of Compute Supply

    Smallest test

    Publish one chart of H100 availability by quarter from current desk data.

    Why it works

    Each quarter, publish an aggregated view (no provider names) of capacity becoming available by GPU type, region and quarter, with the methodology and share of the market covered. Invite operators to add their capacity confidentially to be included and matched.

    It gives buyers a reason to plan purchases with Liquid Compute, gives operators a reason to share supply data, and turns the supply team's internal work into category leadership.

    Risks and how to handle them

    • Operators fear revealing their position

      Publish only aggregates with minimum group sizes.

    Evidence

    • HiringHead of Compute Supply (New York City or San Francisco, $150,000-$200,000 base): "Build the supply network behind a liquid market for GPU capacity"; a supply map by provider, accelerator, region, topology, availability date and contract term; assess counterparty quality and technical readiness; build pricing benchmarks, deal comparisons and supplier records.Source
    • Your website"The GPU Offtaker Credit Map: 200 Private AI Compute Buyers, Ranked" (September 28, 2026): the Liquid Compute Offtaker Score (0-100, five weighted factors, published weights); 169 of 200 buyers scored; "38 of 169 scored companies can support a multi-year offtake today"; the company-level dataset is "available on request"; "We will update it each edition, and re-score any company that shares data with us under NDA."Source

03

How Programmable can help

Where an idea in this playbook is something Programmable builds or runs, it's listed here with what we'd do. Everything else is yours to run.

Have it built

ProgrammableAgency

1 of the 24 ideas are things Programmable builds and runs for its customers: pages, outreach, prospect lists, booking flows and monitors. Pick one, or send the whole list.

Monitors1A live page fed by the public registries and re-checked every day.

A 30-minute session with the engineer who would build it, not a sales call. You leave with a plan whether or not you hire us.

Book a working session

Usually live within days.

Tools for the rest

ProgrammableTalk

Fill the events in this plan

These ideas are events. Talk gets registrants for them and tells you who's coming and what their company does before the day.

Plan an event (opens programmable.talk in a new tab)
ProgrammableDigital

Show it, don't just describe it

These ideas need something to show: a demo, a video, an animation, graphics or a deck. Digital turns screenshots or the site into all of them, in the company's own brand.

Make a demo (opens programmable.digital in a new tab)
More example playbooks and reviews

More growth playbooks

Other companies we have researched, each read from what they are hiring for.

  • vCluster Labs growth strategyA platform that lets AI cloud providers, AI factories and enterprises turn GPU and CPU infrastructure into isolated tenant clusters…22 growth ideas
  • AgentShelf growth strategyA no-code platform to build, deploy and govern AI agents, with an AI Front Desk for business websites, dedicated agent workspaces, an…22 growth ideas
  • Later growth strategyAn influencer marketing company with three lines: full-service and self-serve influencer campaigns for enterprise brands (Later…21 growth ideas
  • Fenrock growth strategyAn AI workspace for the banking back office. Agents sit on top of a bank's existing systems and gather evidence, triage alerts, draft…22 growth ideas
  • Vanta growth strategyAn AI-powered trust management platform that automates compliance (SOC 2, ISO 27001, HIPAA and 35+ frameworks), continuous monitoring…21 growth ideas
  • Confido growth strategyAn AI platform that runs a CPG brand's cash application, deductions and disputes, trade promotion management, sales forecasting and demand…21 growth ideas
  • HG Insights growth strategyA Contextual Intelligence Platform for go-to-market teams and AI agents: technographic, IT spend, intent and contact data (Fabric), HG…22 growth ideas
  • Big Island Coffee Roasters growth strategyAn award-winning Hilo roaster and micro-mill selling the widest range of 100% Hawaiian coffees online (Kona, Ka'u, Puna, Maui), with…21 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 10, 2026, and cites them throughout. Liquid Compute did not commission, review or endorse it, and Programmable is not affiliated with Liquid Compute. Company and product names are trademarks of their owners, used only to identify them. The ideas are suggestions and opinion, not professional, legal or financial advice, and not statements about how Liquid Compute runs its business.

Something wrong, or do you represent Liquid Compute? Email [email protected] and we will correct it or take it down.

30/60/90 roadmap

What to focus on in each month, and the ideas that carry it.

  1. Days 1 to 309 ideas

    Public index page, homepage proof and announcements, one short role-based form, the hero layout fix, a $250M facility page, a regulatory status page, hedge converter links, Services in the nav and sitemap cleanup.

    Programmable can ship 1 of these in days.

  2. Days 31 to 609 ideas

    Daily index email, Credit Map scoring path, GCAT-to-listing loop, index explainer, market-data plans, broker referrals, renewal and resale outreach, lender marks and bylines.

  3. Days 61 to 906 ideas

    GCAT-verified matching and the monthly roundtable; commit to the big bets: independent index assurance, terminal distribution and a public forward curve, plus the supply snapshot.

The research behind the plan

3 research areas, each summarised in a few lines. Open any of them for the full detail.

Customers

What customers say

Customers value A price they can check against the market before committing. The most cited problem: compute is bought and sold with no market price behind it.

  • Signed terms before anything changes hands
  • Public GPU price trackers disagree by multiples
  • In their words: “GPU-hour”

Market

Competitive landscape

3 competitors compared: Silicon Data, Ornn and SF Compute.

  • Gap to exploit: A compute index under external assurance (the CFTC letter says none exists)
  • Silicon Data: Data only: no physical delivery, matching or financing
  • Ornn: No published credit framework for offtakers

Hiring signals

Liquid Compute is hiring its first two commercial leaders, both in New York or San Francisco and both working directly with the founders.

  • Head of Go-to-Market: First commercial leader: pipeline, referrals, events, repeat volume. Ideas 2, 3, 11, 16, 17, 19.
  • Head of Compute Supply: Supply map, benchmarks and supplier diligence. Ideas 12, 13, 24.

From the team that wrote this playbook

All products
  • ProgrammableAgencyWorking AI systems in days, not quarters (opens programmable.agency in a new tab)
  • ProgrammableInfoAn agent that finds, researches and asks first (opens programmable.info in a new tab)
  • ProgrammableDigitalScreenshots in, an animated product demo out (opens programmable.digital in a new tab)
  • ProgrammableWatchKnow the day a supplier falls out of compliance (opens programmable.watch in a new tab)
  • ProgrammableTalkGet registrants for your events, and know who’s coming (opens programmable.talk in a new tab)
  • ProgrammablePageText what you need, your customer gets a page that works (opens programmable.page in a new tab)
  • ProgrammableCommunityForums with labelled AI members, never empty (opens programmable.community in a new tab)
ProgrammableMarketing
ExamplesLeaderboardsProductsTermsPrivacyCookiesDPALegalContact

© 2026 AI-Powered Growth Intelligence