ProgrammableMarketing/vessl.ai (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 6
  • This quarter · 10
  • Big bets · 4
  • Wildcards · 3
  • How we can help
Generated Oct 23, 2026
  1. Programmable
  2. Playbooks
  3. VESSL AI
VESSL AI logo

growth playbook · Oct 23, 2026

vessl.ai (opens in a new tab)

VESSL AI growth playbook

Diagnosis

VESSL AI is growing fast (revenue up 3.4x, GPU capacity tripling, half of revenue from the US) and is building its US team: a first Head of Marketing, a Head of Sales and a founding AE in San Francisco, with sales operations and brand in Seoul.

Its best proof is in Korean only: Upstage Solar Pro 4 inference, SK Biopharm's 128-GPU B200 cluster and the WEF selection have no English versions, and the English blog has a third of the Korean posts.

This playbook brings that proof to US buyers first, fixes the sales form and routing, launches a research credit program, connects self-serve usage to sales, and adds comparison pages and published lead times, then builds English guides, community, field events, standard terms and a funnel dashboard, with bets on a GPU price index, campus agreements, region pages and life sciences.

Growth Score

7/10

Biggest Opportunity

Give US buyers the proof VESSL already has, in English, and connect the self-serve researchers to the sales team as they grow into cluster customers.

Biggest Risk

GPU supply drives everything: if capacity lags demand, marketing creates waiting lists, and rivals such as Lambda, CoreWeave, Nebius and Runpod compete on the same GPUs.

01

Start here

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

  1. 1

    Publish the Korean customer proof in English, starting with Upstage's Solar Pro 4 inference

    Translate the Upstage Solar Pro 4 post this week and send it to the open US inference deals.

    High impact First result in 1 month
  2. 2

    Add the missing paths to "Talk to sales" and route each segment to its owner

    Add the two options and a GPU count field.

    High impact First result in 2 weeks
  3. 3

    Launch a research compute credit program, built on the papers that already thank VESSL

    Write the eligibility rules and the credit tiers.

    High impact First result in 1 month

02

Your idea portfolio

23 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.

  • 6 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 · 2Big bets · 15Fill-ins · 3Reconsider · 3
Effort →

Showing 23 of 23 ideas

This week 6

Quick wins with a clear first action.

  • 01

    VESSL is hiring its first Head of Marketing to give US sellers "case studies and competitor comparisons", and its Head of Sales says demand already outruns selling capacity. The proof those sellers need exists, but only in Korean: VESSL ran global inference for Upstage's Solar Pro 4 while it was the third most used model on OpenRouter, doubling average speed and cutting cost per token by up to half; it supplies SK Biopharm with 128 B200 GPUs; it was named a WEF Technology Pioneer. The English blog has 39 posts to the Korean blog's 113 and one customer story, and the English address of each of these stories redirects to the blog index. Half of VESSL's revenue comes from the US.

    High impactMedium effort First result in 1 month Head of Marketing

    First action

    Translate the Upstage Solar Pro 4 post this week and send it to the open US inference deals.

    How to measure it

    Late-stage US deals with a case study sent

    Now
    –
    Target
    80%
    By
    60 days

    Steps

    1. 1

      Pick five stories

      Upstage, SK Biopharm, a university, a physical AI team, WEF.

      Head of Marketing 1 day

    2. 2

      Check permission

      Confirm each customer is happy to be named in English.

      Head of Sales 3 days

    3. 3

      Adapt, not just translate

      Write for US buyers: workload, GPUs, result, quote.

      Head of Marketing 1 week

    4. 4

      Publish at /en/blog

      Real English addresses, no redirect to the index.

      Growth Marketer 2 days

    5. 5

      Link them

      Homepage, use-case pages and the Provisioned Throughput page.

      Growth Marketer 2 days

    6. 6

      Hand to sales

      One-line summary of each for AE outreach and follow-ups.

      Account Executive 1 day

    7. 7

      Set a rule

      Every new Korean customer post gets an English version within two weeks.

      Head of Marketing ongoing

    8. 8

      Measure

      Views, sales-sourced shares and deal stage movement in HubSpot.

      Growth Marketer monthly

    First result
    1 month
    Cost
    Translation and writing time
    Owner
    Head of Marketing

    Smallest test

    Publish the Solar Pro 4 story in English and track replies from five inference prospects.

    Why it works

    Translate and adapt the five strongest Korean proof posts into English case studies for US buyers: Upstage Solar Pro 4 inference (speed, cost per token, regions), SK Biopharm's 128-GPU B200 cluster, a university customer, a physical AI customer and the WEF selection. Give each a short summary, the workload, the GPUs and the result, link them from the homepage, the use-case pages and the inference page, and make the English addresses resolve instead of redirecting.

    US buyers comparing neoclouds look for proof from teams like theirs, and these results are stronger than most rivals publish. Every new seller can send them the day they start.

    Ready to use

    ChecklistEnglish case study checklist

    [ ] Customer and segment (physical AI, lab, academia, enterprise)
    [ ] Workload: training, fine-tuning or inference
    [ ] GPUs, node count and term
    [ ] Measured result (speed, cost per token, time saved)
    [ ] Customer quote
    [ ] English URL that does not redirect
    [ ] Linked from the matching use-case page
    [ ] Summary line for AE emails

    Risks and how to handle them

    • Customer approval for English publication

      Ask before publishing; use an unnamed version if needed.

    • Literal translation reads oddly

      Have a US-based writer edit every post.

    Evidence

    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
    • Your websiteSitemap (checked October 22, 2026): 39 English blog posts against 113 Korean ones. Only one English customer story (ScatterLab). Korean-only posts include VESSL running global inference for Upstage's Solar Pro 4, supplying 128 B200 GPUs to SK Biopharm, the WEF Technology Pioneer 2026 selection, 3.4x revenue growth, a 2026 GPU cloud buying guide, a Google Colab alternatives guide, a five-year GPU TCO guide and a VLA fine-tuning guide for Isaac GR00T. The English addresses of these posts redirect to the blog index.Source
    • Your websiteKorean blog, October 8, 2026: VESSL ran global inference for Upstage's Solar Pro 4 from GPU servers in Korea, Europe and the US while it was the third most used model on OpenRouter around September; cache and quantization work raised processing speed "2x on average, up to 3x" and cut cost per token by up to half.Source
    • Your websiteKorean blog, September 2, 2026: VESSL supplies SK Biopharm with 128 NVIDIA B200 GPUs on 16 nodes, running since August 2026, for drug research and pretraining foundation models on medical data, extending VESSL from AI companies and research institutes into pharma and biotech.Source
    • Your websiteKorean blog, August 23, 2026 (from an Aju Business interview with the CEO): about 7,000 GPUs today, about 10,000 by year end and a target of 50,000 next year; "half of total revenue comes from the US", with about 20 US customers waiting for GPU supply. A February 2026 post reports revenue up 3.4x year on year, monthly users up 3.7x, and paying university customers such as UW-Madison and the University of Minnesota.Source
  • 02

    The sales form offers only "Reserved GPU instance" or "Inference". Academic labs, teams wanting on-demand credits and physical AI teams have no option, and the form asks nothing about GPU count or segment. The Head of Sales plans one owner per segment.

    High impactLow effort First result in 2 weeks Head of Sales

    First action

    Add the two options and a GPU count field.

    How to measure it

    Sales requests that become meetings

    Now
    –
    Target
    +25%
    By
    60 days

    Steps

    1. 1

      Add options and fields

      Segment, GPUs, region.

      Growth Marketer 1 day

    2. 2

      Set routing

      Owner by segment in HubSpot.

      Sales Operations Specialist 2 days

    3. 3

      Auto-reply

      Matching case study and next step.

      Growth Marketer 1 day

    4. 4

      Capacity check

      GPU request goes to capacity planning.

      Head of Sales ongoing

    5. 5

      Track

      Meetings by segment.

      Sales Operations Specialist weekly

    6. 6

      Review

      Adjust fields after a month.

      Head of Sales 1 month

    First result
    2 weeks
    Cost
    $0
    Owner
    Head of Sales

    Smallest test

    Add the academic option today.

    Why it works

    Add "Academic lab or research credits" and "On-demand for a team" options, plus segment, GPU type and count, and region. Route each request in HubSpot to its segment owner and send an automatic reply with the matching case study.

    Better-qualified requests get faster, more relevant replies, and the GPU details let sales check capacity before the first call.

    Risks and how to handle them

    • More fields lower completion

      Make GPU count a simple menu.

    Evidence

    • Your websiteTalk to sales page: "One form for every product" with two choices only, "Reserved GPU instance" and "Inference", then name, work email, company, role and start date; "We reply within 1 business day".Source
    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
  • 03

    The Head of Marketing is asked to build "compute credit programs, and student ambassadors", and the academia page already shows six papers that thank VESSL for compute. There is no page where a lab can apply.

    High impactMedium effort First result in 1 month Head of Marketing

    First action

    Write the eligibility rules and the credit tiers.

    How to measure it

    Labs with credits

    Now
    –
    Target
    50
    By
    1 quarter

    Steps

    1. 1

      Set rules

      Eligibility and tiers.

      Head of Marketing 3 days

    2. 2

      Build page

      With the application form.

      Growth Marketer 1 week

    3. 3

      Show past papers

      Acknowledged research.

      Growth Marketer 1 day

    4. 4

      Review weekly

      Approve and onboard.

      Growth Marketer weekly

    5. 5

      Hand off growth

      Labs past the credit to sales.

      Account Executive ongoing

    6. 6

      Measure

      Paid conversion by cohort.

      Growth Marketer monthly

    First result
    1 month
    Cost
    Credit budget
    Owner
    Head of Marketing

    Smallest test

    Offer credits to ten labs at one university.

    Why it works

    Publish a credits page with eligibility (university labs, PhD students), credit tiers, an application form, the papers already supported, and a simple ask: acknowledge VESSL in the paper. Review applications weekly and hand labs that grow to the academic seller.

    Credits bring in researchers who later move to labs and companies, which is the path the marketing post describes.

    Risks and how to handle them

    • Credits used and abandoned

      Release credits in stages based on use.

    Evidence

    • HiringThe same Head of Marketing post: "the researcher running a first job on a single GPU today is the lab reserving a cluster tomorrow"; partner with product on "onboarding, activation, in-product growth loops, and a clean handoff of high-intent users to sales"; build "penetration into academic institutions through university labs, research groups, compute credit programs, and student ambassadors"; run "meetups, workshops, and hackathons, plus executive dinners".Source
    • Your websiteAcademia use-case page: "No quotas, spin up a GPU, no approvals", per-second billing, and six research papers that thank VESSL AI for compute (two of them alongside Berkeley's Center for Human-Compatible AI); no compute credit program, application or ambassador page on the site.Source
    • Your websiteKorean blog, August 23, 2026 (from an Aju Business interview with the CEO): about 7,000 GPUs today, about 10,000 by year end and a target of 50,000 next year; "half of total revenue comes from the US", with about 20 US customers waiting for GPU supply. A February 2026 post reports revenue up 3.4x year on year, monthly users up 3.7x, and paying university customers such as UW-Madison and the University of Minnesota.Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
  • 04

    The AE post asks for working with the Growth Marketer to find "high-potential accounts on cloud.vessl.ai" and to reach out "at the right moment". Self-serve runs on per-second billing with no quotas, so usage is the best buying signal.

    High impactLow effort First result in 1 month Account Executive

    First action

    List accounts that ran eight or more GPUs last month.

    How to measure it

    Self-serve accounts moving to reserved terms

    Now
    –
    Target
    5 a month
    By
    1 quarter

    Steps

    1. 1

      Pick signals

      From accounts that already converted.

      Growth Marketer 1 week

    2. 2

      Send to HubSpot

      Usage alerts.

      Sales Operations Specialist 1 week

    3. 3

      Write outreach

      Offer tied to their usage.

      Account Executive 2 days

    4. 4

      Contact

      Within one business day.

      Account Executive ongoing

    5. 5

      Track

      Conversion by signal.

      Sales Operations Specialist weekly

    6. 6

      Tune

      Drop weak signals.

      Growth Marketer monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Account Executive

    Smallest test

    Email ten high-usage accounts this week.

    Why it works

    Define three signals (for example multi-GPU jobs, spend over a threshold in a week, several team members on one account) and send them to HubSpot as alerts with the account's usage summary. The AE contacts each one within a day with a reserved-capacity offer.

    Reaching out when usage climbs turns self-serve growth into reserved contracts before the team shops elsewhere.

    Risks and how to handle them

    • Users dislike sales contact

      Offer help with the workload first.

    Evidence

    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
    • HiringThe same Head of Marketing post: "the researcher running a first job on a single GPU today is the lab reserving a cluster tomorrow"; partner with product on "onboarding, activation, in-product growth loops, and a clean handoff of high-intent users to sales"; build "penetration into academic institutions through university labs, research groups, compute credit programs, and student ambassadors"; run "meetups, workshops, and hackathons, plus executive dinners".Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
  • 05

    The Head of Marketing must own "one core story on why AI researchers choose us over hyperscalers and neoclouds" and write competitor comparisons. VESSL already has a blog post on hyperscalers versus neoclouds, but no page compares VESSL with a named rival, where buyers search before they choose.

    High impactMedium effort First result in 1 month Head of Marketing

    First action

    Write the Lambda comparison first.

    How to measure it

    Sign-ups and sales requests from comparisons

    Now
    –
    Target
    100 a month
    By
    1 quarter

    Steps

    1. 1

      Pick rivals

      Those met most in deals.

      Head of Sales 1 day

    2. 2

      Check facts

      Current public prices and features.

      Growth Marketer 1 week

    3. 3

      Write pages

      Fair tables and a switching guide.

      Head of Marketing 2 weeks

    4. 4

      Review

      Legal and sales.

      Head of Sales 3 days

    5. 5

      Publish

      English and Korean.

      Growth Marketer 2 days

    6. 6

      Refresh

      Prices change often.

      Growth Marketer monthly

    First result
    1 month
    Cost
    Writing time
    Owner
    Head of Marketing

    Smallest test

    Publish one comparison and track sign-ups.

    Why it works

    Build fair comparison pages: on-demand prices, reserved options, per-second billing, quotas, multi-node InfiniBand, storage, inference terms, support and multi-provider supply. Use the internal competitor comparisons for sales and keep the public pages factual and dated.

    Comparison searches are high intent, and AI assistants use these pages when people ask which GPU cloud to choose.

    Risks and how to handle them

    • Out-of-date prices

      Date every table and recheck monthly.

    Evidence

    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
    • Competitor researchRivals' own homepages: Lambda, "Supercomputers for training and inference" with on-demand instances or a reserved cluster; Runpod, "The AI Developer Cloud"; Nebius, "The Ultimate AI Cloud ... engineered from silicon to API"; CoreWeave, "The essential cloud for AI".Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
    • Your websiteSitemap (checked October 22, 2026): 39 English blog posts against 113 Korean ones. Only one English customer story (ScatterLab). Korean-only posts include VESSL running global inference for Upstage's Solar Pro 4, supplying 128 B200 GPUs to SK Biopharm, the WEF Technology Pioneer 2026 selection, 3.4x revenue growth, a 2026 GPU cloud buying guide, a Google Colab alternatives guide, a five-year GPU TCO guide and a VLA fine-tuning guide for Isaac GR00T. The English addresses of these posts redirect to the blog index.Source
  • 06

    Every Blackwell and H200 option says "Quoted by term" or "Reserved only", and VESSL's own Korean coverage says about 20 US customers are waiting for GPU supply. Buyers cannot tell whether a B200 cluster is available next week or next quarter until they talk to sales.

    High impactMedium effort First result in 1 month Head of Sales

    First action

    Agree who updates the table each week.

    How to measure it

    Reserved requests that match available capacity

    Now
    –
    Target
    70%
    By
    1 quarter

    Steps

    1. 1

      Agree the data

      Start dates and sizes by GPU.

      Head of Sales 1 week

    2. 2

      Design table

      Pricing and GPU pages.

      Brand Designer (Senior) 1 week

    3. 3

      Build

      Editable by sales.

      Growth Marketer 1 week

    4. 4

      Weekly update

      Capacity and sales review.

      Head of Sales weekly

    5. 5

      Add hold request

      Form with GPU and dates.

      Growth Marketer 2 days

    6. 6

      Measure

      Requests and close rate.

      Sales Operations Specialist monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Head of Sales

    Smallest test

    Publish lead times for two GPU types.

    Why it works

    Add an availability table to the pricing and GPU pages: for each GPU type, the earliest start date and typical cluster sizes, updated weekly by the sales and capacity team, with a "hold this slot" request.

    In a supply-limited market, a clear start date is often more persuasive than a price, and it filters requests to what VESSL can deliver.

    Risks and how to handle them

    • Competitors read it

      Show ranges, not exact counts.

    Evidence

    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
    • Your websiteKorean blog, August 23, 2026 (from an Aju Business interview with the CEO): about 7,000 GPUs today, about 10,000 by year end and a target of 50,000 next year; "half of total revenue comes from the US", with about 20 US customers waiting for GPU supply. A February 2026 post reports revenue up 3.4x year on year, monthly users up 3.7x, and paying university customers such as UW-Madison and the University of Minnesota.Source

This quarter 10

The core plays for the next three months.

  • 07

    The Head of Marketing will own SEO and AEO. Korean posts already answer the questions buyers ask (which GPU cloud to choose in 2026, when to move off Google Colab, how to plan a five-year GPU budget), but none exists in English.

    High impactMedium effort First result in 1 quarter Growth Marketer

    First action

    Pick ten Korean guides by search demand in English.

    How to measure it

    Sign-ups from English guides

    Now
    –
    Target
    200 a month
    By
    2 quarters

    Steps

    1. 1

      Pick guides

      By English search demand.

      Growth Marketer 1 week

    2. 2

      Adapt

      US prices and examples.

      Growth Marketer 1 month

    3. 3

      Edit

      Native English editor.

      Head of Marketing ongoing

    4. 4

      Publish

      With estimator links.

      Growth Marketer weekly

    5. 5

      Track

      Rankings and AI citations.

      Growth Marketer monthly

    6. 6

      Update

      Prices quarterly.

      Growth Marketer quarterly

    First result
    1 quarter
    Cost
    Writing time
    Owner
    Growth Marketer

    Smallest test

    Publish the Colab alternatives guide in English.

    Why it works

    Adapt the ten best Korean guides for US readers, with current prices and VESSL's on-demand rates, a clear answer in the first paragraph, and a link to the cost estimator or sign-up.

    Guides that answer common questions bring researchers in at the start of their search, and AI assistants cite clear answers.

    Risks and how to handle them

    • Thin translations

      Rewrite with US context, not word for word.

    Builds on

    Evidence

    • Your websiteSitemap (checked October 22, 2026): 39 English blog posts against 113 Korean ones. Only one English customer story (ScatterLab). Korean-only posts include VESSL running global inference for Upstage's Solar Pro 4, supplying 128 B200 GPUs to SK Biopharm, the WEF Technology Pioneer 2026 selection, 3.4x revenue growth, a 2026 GPU cloud buying guide, a Google Colab alternatives guide, a five-year GPU TCO guide and a VLA fine-tuning guide for Isaac GR00T. The English addresses of these posts redirect to the blog index.Source
    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
  • 08

    The marketing post names Discord, Reddit and Hacker News as the places researchers spend time, but VESSL links only GitHub, YouTube, LinkedIn and X. Researchers on self-serve GPUs have nowhere to ask each other questions.

    Medium impactMedium effort First result in 1 quarter Head of Marketing

    First action

    Set up the server and invite current active users.

    How to measure it

    Weekly active members

    Now
    –
    Target
    500
    By
    2 quarters

    Steps

    1. 1

      Set up

      Channels and rules.

      Growth Marketer 1 week

    2. 2

      Invite

      Active users and credit labs.

      Growth Marketer 1 week

    3. 3

      Office hours

      Weekly with engineers.

      Forward Deployed Engineer (Senior) weekly

    4. 4

      Early access

      For active members.

      Head of Marketing monthly

    5. 5

      Moderate

      Daily.

      Growth Marketer ongoing

    6. 6

      Measure

      Activity and conversions.

      Growth Marketer monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Head of Marketing

    Smallest test

    Run one office hour on X first.

    Why it works

    Open a Discord with channels by workload (fine-tuning, RL, physical AI, inference), weekly office hours with engineers, and early access to new GPUs and features for active members.

    A community lowers support load, keeps self-serve users active and surfaces the teams likely to need clusters.

    Risks and how to handle them

    • Quiet server

      Seed it with office hours and credit-program labs.

    Evidence

    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
    • Public channelsChannels linked from the site: GitHub (48 public repositories, 153 stars in total), YouTube (2,440 subscribers, 83 videos), LinkedIn and X; no Discord or Reddit community. Documentation and changelog on docs.cloud.vessl.ai; a status page; a Vanta trust center.Source
    • HiringThe same Head of Marketing post: "the researcher running a first job on a single GPU today is the lab reserving a cluster tomorrow"; partner with product on "onboarding, activation, in-product growth loops, and a clean handoff of high-intent users to sales"; build "penetration into academic institutions through university labs, research groups, compute credit programs, and student ambassadors"; run "meetups, workshops, and hackathons, plus executive dinners".Source
  • 09

    Six papers on the academia page thank VESSL for compute. Each acknowledgement is a researcher who trusted the platform and a lab where others will notice.

    Medium impactLow effort First result in 1 quarter Growth Marketer

    First action

    Search recent arXiv papers that mention VESSL.

    How to measure it

    Active student ambassadors

    Now
    –
    Target
    25
    By
    2 quarters

    Steps

    1. 1

      Find papers

      arXiv and conference proceedings.

      Growth Marketer 1 week

    2. 2

      Build page

      Papers, labs, GPUs.

      Growth Marketer 1 week

    3. 3

      Invite authors

      Ambassador offer.

      Growth Marketer 2 weeks

    4. 4

      Give ambassadors tools

      Credits and talks kit.

      Head of Marketing 1 month

    5. 5

      Share papers

      Social posts.

      Growth Marketer ongoing

    6. 6

      Measure

      Sign-ups from ambassador links.

      Growth Marketer monthly

    First result
    1 quarter
    Cost
    Credit budget
    Owner
    Growth Marketer

    Smallest test

    Invite the six acknowledged authors.

    Why it works

    Track every paper that acknowledges VESSL, list them on a public page with the lab and the GPUs used, invite the authors to be student ambassadors with extra credits, and share each new paper on X and LinkedIn.

    Researchers follow peers, and a visible list of papers shows the platform works for serious research.

    Risks and how to handle them

    • Ambassadors go quiet

      Tie extra credits to activity.

    Builds on

    Evidence

    • Your websiteAcademia use-case page: "No quotas, spin up a GPU, no approvals", per-second billing, and six research papers that thank VESSL AI for compute (two of them alongside Berkeley's Center for Human-Compatible AI); no compute credit program, application or ambassador page on the site.Source
    • HiringThe same Head of Marketing post: "the researcher running a first job on a single GPU today is the lab reserving a cluster tomorrow"; partner with product on "onboarding, activation, in-product growth loops, and a clean handoff of high-intent users to sales"; build "penetration into academic institutions through university labs, research groups, compute credit programs, and student ambassadors"; run "meetups, workshops, and hackathons, plus executive dinners".Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
  • 10

    The marketing post asks for a global events calendar, workshops, hackathons and executive dinners, and VESSL already ran hackathons around ICML in Korea. US AEs say event contacts are part of today's pipeline.

    High impactMedium effort First result in 2 quarters Head of Marketing

    First action

    Choose the next two US events.

    How to measure it

    Event-sourced pipeline

    Now
    –
    Target
    25% of new US pipeline
    By
    2 quarters

    Steps

    1. 1

      Pick events

      Four a year.

      Head of Marketing 1 week

    2. 2

      Plan workshops

      Hands-on on VESSL.

      Forward Deployed Engineer (Senior) 1 month

    3. 3

      Dinners

      Invite lab leaders.

      Head of Sales per event

    4. 4

      Follow up

      Within two days.

      Account Executive per event

    5. 5

      Track

      Contacts to pipeline.

      Sales Operations Specialist monthly

    6. 6

      Review

      Keep the best events.

      Head of Marketing quarterly

    First result
    2 quarters
    Cost
    Event budget
    Owner
    Head of Marketing

    Smallest test

    Host one dinner around the next event.

    Why it works

    Pick four US events a year, run a hands-on workshop at each (for example fine-tuning on B200), host a dinner for 10-15 lab leaders, and follow up every contact within two days with credits or a capacity call.

    Researchers decide with their hands, and dinners build the relationships behind multi-year cluster deals.

    Risks and how to handle them

    • Costly events with few leads

      Measure pipeline per event and drop weak ones.

    Evidence

    • HiringThe same Head of Marketing post: "the researcher running a first job on a single GPU today is the lab reserving a cluster tomorrow"; partner with product on "onboarding, activation, in-product growth loops, and a clean handoff of high-intent users to sales"; build "penetration into academic institutions through university labs, research groups, compute credit programs, and student ambassadors"; run "meetups, workshops, and hackathons, plus executive dinners".Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
  • 11

    The Head of Sales will design terms from "short-term on-demand credits through multi-year reserved capacity commitments" and a deal desk that moves quotes "in days". Today the site says only that terms follow volume and length.

    High impactMedium effort First result in 1 quarter Head of Sales

    First action

    Draft the packages from the last ten reserved deals.

    How to measure it

    Days to quote

    Now
    –
    Target
    Under 3
    By
    1 quarter

    Steps

    1. 1

      Review deals

      Last ten reserved contracts.

      Head of Sales 1 week

    2. 2

      Define packages

      Terms and inclusions.

      Head of Sales 2 weeks

    3. 3

      Finance check

      Margins.

      Finance Manager 1 week

    4. 4

      Publish

      On pricing.

      Growth Marketer 1 week

    5. 5

      Train sales

      Use packages first.

      Head of Sales 1 week

    6. 6

      Track

      Quote time and win rate.

      Sales Operations Specialist monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Head of Sales

    Smallest test

    Use the packages in the next five quotes.

    Why it works

    Publish three or four standard reserved packages (for example one month, three months, one year) with what each includes (support, SLA, storage, ramp-up), indicative discounts against on-demand, and a quote request that preselects the package.

    Standard terms speed up quotes and set expectations, so sales time goes to the deals that need custom terms.

    Risks and how to handle them

    • Price anchoring

      Show indicative ranges, not fixed prices.

    Evidence

    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
    • Your websiteTalk to sales page: "One form for every product" with two choices only, "Reserved GPU instance" and "Inference", then name, work email, company, role and start date; "We reply within 1 business day".Source
  • 12

    Physical AI is one of the four segments the Head of Sales will staff, and customers include physical AI startups. The Korean blog has a VLA fine-tuning guide for Isaac GR00T and a post on why GPU infrastructure is the bottleneck for physical AI; the English site has the use-case page but not these.

    High impactMedium effort First result in 1 quarter Head of Marketing

    First action

    Adapt the GR00T guide into English.

    How to measure it

    Sign-ups tagged physical AI

    Now
    –
    Target
    100 a month
    By
    2 quarters

    Steps

    1. 1

      Adapt GR00T guide

      English, tested.

      Forward Deployed Engineer (Senior) 2 weeks

    2. 2

      RL tutorial

      Simulation on VESSL.

      Forward Deployed Engineer (Senior) 1 month

    3. 3

      Customer story

      One physical AI team.

      Head of Marketing 1 month

    4. 4

      Template

      One-click environment.

      Forward Deployed Engineer (Senior) 2 weeks

    5. 5

      Promote

      Robotics communities.

      Growth Marketer ongoing

    6. 6

      Track

      Sign-ups and pipeline.

      Sales Operations Specialist monthly

    First result
    1 quarter
    Cost
    Engineering time
    Owner
    Head of Marketing

    Smallest test

    Publish the GR00T guide.

    Why it works

    Publish English tutorials (fine-tune a VLA model, run RL in simulation on VESSL), a physical AI customer story and a robotics-ready template, and promote them in robotics communities.

    Practical tutorials reach robotics engineers before they choose a cloud, and they support the physical AI seller.

    Risks and how to handle them

    • Tutorials go stale

      Retest each quarter.

    Evidence

    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
    • Your websiteSitemap (checked October 22, 2026): 39 English blog posts against 113 Korean ones. Only one English customer story (ScatterLab). Korean-only posts include VESSL running global inference for Upstage's Solar Pro 4, supplying 128 B200 GPUs to SK Biopharm, the WEF Technology Pioneer 2026 selection, 3.4x revenue growth, a 2026 GPU cloud buying guide, a Google Colab alternatives guide, a five-year GPU TCO guide and a VLA fine-tuning guide for Isaac GR00T. The English addresses of these posts redirect to the blog index.Source
    • Your websiteUse cases: Physical AI, Academia, and NLP & foundation models ("MoE and 100B+ parameter runs, scaled to 1,000+ GPUs"). The Provisioned Throughput page offers OpenAI- and Anthropic-compatible APIs with throughput and TPOT targets in the contract, and "Need a different model? Tell us which model you need".Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
  • 13

    The marketing post owns lifecycle and activation, and "the researcher running a first job on a single GPU today is the lab reserving a cluster tomorrow". Self-serve has no quotas, so the main barrier is setup.

    High impactMedium effort First result in 1 month Growth Marketer

    First action

    Measure today's time from sign-up to first job.

    How to measure it

    First job within 24 hours

    Now
    –
    Target
    50%
    By
    1 quarter

    Steps

    1. 1

      Measure

      Current activation.

      Growth Marketer 1 week

    2. 2

      Write emails

      Four steps.

      Growth Marketer 1 week

    3. 3

      Templates

      By workload.

      Forward Deployed Engineer (Senior) 2 weeks

    4. 4

      Set triggers

      Product events.

      Growth Marketer 1 week

    5. 5

      Test

      A/B subject lines.

      Growth Marketer 1 month

    6. 6

      Review

      Activation by cohort.

      Growth Marketer monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Growth Marketer

    Smallest test

    Send the 24-hour nudge by hand for a week.

    Why it works

    Send a short sequence: a starter template for the user's stated workload, the vesslctl CLI quick start, a nudge if no job ran in 24 hours, and an invite to office hours. Measure time to first job.

    Users who run a job early stay, and their usage creates the signals sales uses.

    Risks and how to handle them

    • Too many emails

      Stop the sequence once a job runs.

    Evidence

    • HiringThe same Head of Marketing post: "the researcher running a first job on a single GPU today is the lab reserving a cluster tomorrow"; partner with product on "onboarding, activation, in-product growth loops, and a clean handoff of high-intent users to sales"; build "penetration into academic institutions through university labs, research groups, compute credit programs, and student ambassadors"; run "meetups, workshops, and hackathons, plus executive dinners".Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
    • Public channelsChannels linked from the site: GitHub (48 public repositories, 153 stars in total), YouTube (2,440 subscribers, 83 videos), LinkedIn and X; no Discord or Reddit community. Documentation and changelog on docs.cloud.vessl.ai; a status page; a Vanta trust center.Source
  • 14

    The Head of Marketing needs "a dashboard VESSL leadership can plan against", and the Sales Operations Specialist is rebuilding pipeline and forecasting in HubSpot. Self-serve, sales requests and events now flow into different places.

    High impactMedium effort First result in 1 quarter Sales Operations Specialist

    First action

    Agree the funnel stages.

    How to measure it

    Deals with a source

    Now
    –
    Target
    90%
    By
    1 quarter

    Steps

    1. 1

      Agree stages

      Marketing and sales.

      Head of Marketing 1 week

    2. 2

      Source field

      Everywhere.

      Sales Operations Specialist 2 weeks

    3. 3

      Connect usage

      Product to HubSpot.

      Sales Operations Specialist 1 month

    4. 4

      Build dashboard

      By segment and region.

      Sales Operations Specialist 2 weeks

    5. 5

      Review

      Weekly.

      Head of Sales weekly

    6. 6

      Act

      Shift budget.

      Head of Marketing monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Sales Operations Specialist

    Smallest test

    Build the funnel by hand for last month.

    Why it works

    Connect site, product sign-ups, usage alerts and HubSpot deals with one source field, and build one dashboard by segment and region: visits, sign-ups, activations, usage signals, sales requests, pipeline and closed revenue.

    One view shows which channels create reserved revenue, so budget follows results.

    Risks and how to handle them

    • Messy old data

      Start clean from this quarter.

    Builds on

    Evidence

    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
    • HiringIn Seoul: a Sales Operations Specialist (posted September 18, 2026) to rebuild pipeline and forecasting in HubSpot and design operations "for regional expansion"; a Senior Brand Designer (posted August 26) working with marketing, sales, PR and IR; a Project Manager (posted October 16) building "Customer RFI, Deal Readiness Review, Go-live Checklist"; and Forward Deployed Engineers.Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
  • 15

    The inference page asks visitors to "Tell us which model you need". Requests sent privately do not show demand to other buyers or let VESSL rank them.

    Medium impactMedium effort First result in 1 quarter Head of Sales

    First action

    List the models requested so far.

    How to measure it

    PTU trials started

    Now
    –
    Target
    10 a quarter
    By
    2 quarters

    Steps

    1. 1

      Collect requests

      From sales and the form.

      Head of Sales 1 week

    2. 2

      Build board

      Votes and use cases.

      Growth Marketer 2 weeks

    3. 3

      Rank

      Demand and cost.

      Head of Sales monthly

    4. 4

      Launch models

      Top requests.

      Forward Deployed Engineer (Senior) monthly

    5. 5

      Notify voters

      With a trial.

      Growth Marketer per launch

    6. 6

      Measure

      Trials to contracts.

      Sales Operations Specialist monthly

    First result
    1 quarter
    Cost
    Engineering time
    Owner
    Head of Sales

    Smallest test

    Publish the board with five models.

    Why it works

    Publish a board of requested open models with votes and the use case behind each, add the most-voted models first, and email voters when their model goes live with a PTU trial.

    Votes show demand before VESSL spends GPU time, and each launch comes with a ready list of buyers.

    Risks and how to handle them

    • Requests for models VESSL cannot serve

      Explain limits on the board.

    Evidence

    • Your websiteUse cases: Physical AI, Academia, and NLP & foundation models ("MoE and 100B+ parameter runs, scaled to 1,000+ GPUs"). The Provisioned Throughput page offers OpenAI- and Anthropic-compatible APIs with throughput and TPOT targets in the contract, and "Need a different model? Tell us which model you need".Source
    • Your websiteKorean blog, October 8, 2026: VESSL ran global inference for Upstage's Solar Pro 4 from GPU servers in Korea, Europe and the US while it was the third most used model on OpenRouter around September; cache and quantization work raised processing speed "2x on average, up to 3x" and cut cost per token by up to half.Source
    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
  • 16

    VESSL ran Solar Pro 4's global inference while it was among OpenRouter's most used models, planning capacity with Upstage around launch and promotion dates. Other model labs face the same spike at launch.

    High impactHigh effort First result in 2 quarters Head of Sales

    First action

    List ten labs planning open model releases.

    How to measure it

    Model labs served at launch

    Now
    –
    Target
    3
    By
    3 quarters

    Steps

    1. 1

      Define offer

      Capacity, regions, targets.

      Head of Sales 2 weeks

    2. 2

      Publish story

      Solar Pro 4 in English.

      Head of Marketing 2 weeks

    3. 3

      Target labs

      Open model releases.

      Account Executive 1 month

    4. 4

      Plan capacity

      With each lab.

      Project Manager per deal

    5. 5

      Serve launch

      Multi-region.

      Forward Deployed Engineer (Senior) per deal

    6. 6

      Publish result

      Case study.

      Head of Marketing per deal

    First result
    2 quarters
    Cost
    Reserved GPU capacity
    Owner
    Head of Sales

    Smallest test

    Pitch the offer to two labs.

    Why it works

    Package a launch offer for model labs: capacity planned with the lab before launch, multi-region serving, cost per token targets, and a short case study after. Pitch it to labs releasing open models.

    Launches are when labs most need extra capacity and are least able to build it themselves, and each one leaves a public reference.

    Risks and how to handle them

    • Launch traffic misjudged

      Agree burst limits in the contract.

    Builds on

    Evidence

    • Your websiteKorean blog, October 8, 2026: VESSL ran global inference for Upstage's Solar Pro 4 from GPU servers in Korea, Europe and the US while it was the third most used model on OpenRouter around September; cache and quantization work raised processing speed "2x on average, up to 3x" and cut cost per token by up to half.Source
    • Your websiteUse cases: Physical AI, Academia, and NLP & foundation models ("MoE and 100B+ parameter runs, scaled to 1,000+ GPUs"). The Provisioned Throughput page offers OpenAI- and Anthropic-compatible APIs with throughput and TPOT targets in the contract, and "Need a different model? Tell us which model you need".Source
    • Your websiteKorean WEF post (June 10, 2026): VESSL works with data centers in the US, Israel, Finland, Japan and Korea, and was named a WEF Technology Pioneer 2026, one of 100 companies chosen each year.Source

Big bets 4

Larger plays that can change the trajectory.

  • 17

    VESSL describes itself as routing capacity "across multiple clusters and providers" and as "the GPU liquidity layer", so it sees supply and prices across the market. Buyers lack a neutral view of where GPUs are available and at what price.

    High impactHigh effort First result in 3 quarters Head of Marketing

    First action

    Build the first index from internal sourcing data.

    How to measure it

    Index subscribers

    Now
    –
    Target
    5,000
    By
    1 year

    Steps

    1. 1

      Define method

      Data sources and rules.

      Head of Marketing 1 month

    2. 2

      Legal check

      Supplier data use.

      Legal 2 weeks

    3. 3

      Build report

      Charts and notes.

      Growth Marketer 1 month

    4. 4

      Launch

      Press and social.

      Head of Marketing 1 month

    5. 5

      Monthly updates

      Same day each month.

      Growth Marketer monthly

    6. 6

      Use in sales

      Market context.

      Head of Sales ongoing

    First result
    3 quarters
    Cost
    Analyst time
    Owner
    Head of Marketing

    Smallest test

    Publish one chart of B200 lead times.

    Why it works

    Publish a monthly index of on-demand and reserved prices and availability by GPU type and region, built from VESSL's own sourcing data and public price lists, with commentary and a sign-up for alerts.

    A trusted index earns press, links and AI citations, and positions VESSL as the expert on GPU supply.

    Risks and how to handle them

    • Suppliers object

      Aggregate and never name supplier prices.

    Evidence

    • Your websitePositioning differs by page: the homepage says "The AI Cloud for All Models"; the careers page says "the GPU liquidity layer" and "the orchestration layer for AI infrastructure", "$16M+ Total Funding", "100+ Customers", "3.4x Revenue Growth (YoY)", partners Hyundai Motor and Tmap Mobility, and "VESSL Cloud in February 2026".Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
    • Competitor researchRivals' own homepages: Lambda, "Supercomputers for training and inference" with on-demand instances or a reserved cluster; Runpod, "The AI Developer Cloud"; Nebius, "The Ultimate AI Cloud ... engineered from silicon to API"; CoreWeave, "The essential cloud for AI".Source
    • Your websiteKorean blog, August 23, 2026 (from an Aju Business interview with the CEO): about 7,000 GPUs today, about 10,000 by year end and a target of 50,000 next year; "half of total revenue comes from the US", with about 20 US customers waiting for GPU supply. A February 2026 post reports revenue up 3.4x year on year, monthly users up 3.7x, and paying university customers such as UW-Madison and the University of Minnesota.Source
  • 18

    Universities such as UW-Madison and the University of Minnesota already pay for VESSL Cloud, and researchers at Stanford, Berkeley, UCLA and NYU use it. Single-lab deals are small; research computing offices buy for the whole campus.

    High impactHigh effort First result in 4 quarters Head of Sales

    First action

    Ask the two paying universities who buys research computing centrally.

    How to measure it

    Campus-wide agreements

    Now
    –
    Target
    3
    By
    4 quarters

    Steps

    1. 1

      Interview offices

      Needs and budgets.

      Head of Sales 1 month

    2. 2

      Design offer

      Pooled credits and billing.

      Head of Sales 1 month

    3. 3

      Build admin view

      Lab usage.

      Product team 1 quarter

    4. 4

      Pilot

      One campus.

      Account Executive 2 quarters

    5. 5

      Case study

      Results.

      Head of Marketing 1 month

    6. 6

      Expand

      More campuses.

      Head of Sales 2 quarters

    First result
    4 quarters
    Cost
    Product and sales time
    Owner
    Head of Sales

    Smallest test

    Pitch a pooled-credit pilot to one university.

    Why it works

    Offer research computing offices a campus agreement: pooled credits, central billing, a lab admin console and usage reports, priced on committed spend. Start with the two paying universities.

    One campus agreement replaces dozens of small deals and makes VESSL the default for new labs.

    Risks and how to handle them

    • Slow university procurement

      Start with a pilot under existing lab contracts.

    Builds on

    Evidence

    • Your websiteKorean blog, August 23, 2026 (from an Aju Business interview with the CEO): about 7,000 GPUs today, about 10,000 by year end and a target of 50,000 next year; "half of total revenue comes from the US", with about 20 US customers waiting for GPU supply. A February 2026 post reports revenue up 3.4x year on year, monthly users up 3.7x, and paying university customers such as UW-Madison and the University of Minnesota.Source
    • HiringAccount Executive (San Francisco, posted August 12, 2026), "the first dedicated sales hire in the US": VESSL is "Series A, ~40 people, scaling from 3,000 to 10,000 GPUs by Q4 2026"; customers include "an Artificial Analysis Top 15 foundation model team, Physical AI startups, and research groups at Stanford, UC Berkeley, UMN, UCLA, NYU"; the AE works "with the Growth Marketer to identify high-potential accounts on cloud.vessl.ai" and owns the CRM (HubSpot).Source
    • Your websiteAcademia use-case page: "No quotas, spin up a GPU, no approvals", per-second billing, and six research papers that thank VESSL AI for compute (two of them alongside Berkeley's Center for Human-Compatible AI); no compute credit program, application or ambassador page on the site.Source
    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
  • 19

    VESSL works with data centers in the US, Israel, Finland, Japan and Korea, and served Solar Pro 4 from Korea, Europe and the US. European and Japanese buyers often require data to stay in-region, and the site has no region pages.

    High impactHigh effort First result in 4 quarters Head of Sales

    First action

    Write the European region page.

    How to measure it

    Non-US, non-Korea requests

    Now
    –
    Target
    +50%
    By
    4 quarters

    Steps

    1. 1

      Confirm regions

      GPUs and terms.

      Head of Sales 2 weeks

    2. 2

      Write pages

      Region facts.

      Head of Marketing 1 month

    3. 3

      Compliance review

      Data handling claims.

      Legal 2 weeks

    4. 4

      Publish

      With region routing.

      Growth Marketer 1 week

    5. 5

      Regional owners

      Sales coverage.

      Head of Sales 1 quarter

    6. 6

      Measure

      Requests by region.

      Sales Operations Specialist monthly

    First result
    4 quarters
    Cost
    Staff time only
    Owner
    Head of Sales

    Smallest test

    Publish one region page.

    Why it works

    Publish pages per region with available GPUs, data location, compliance (SOC 2 Type II) and latency, and assign regional owners as the marketing post plans regional teams.

    Region pages open buyers who must keep data local and give regional teams their own entry point.

    Risks and how to handle them

    • Claims about residency

      State only what contracts guarantee.

    Evidence

    • Your websiteKorean WEF post (June 10, 2026): VESSL works with data centers in the US, Israel, Finland, Japan and Korea, and was named a WEF Technology Pioneer 2026, one of 100 companies chosen each year.Source
    • Your websiteKorean blog, October 8, 2026: VESSL ran global inference for Upstage's Solar Pro 4 from GPU servers in Korea, Europe and the US while it was the third most used model on OpenRouter around September; cache and quantization work raised processing speed "2x on average, up to 3x" and cut cost per token by up to half.Source
    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source
  • 20

    SK Biopharm runs 128 B200 GPUs on VESSL for drug research and pretraining on medical data, which takes VESSL into pharma. US biotech teams have the same need for large, private training runs.

    High impactHigh effort First result in 4 quarters Head of Sales

    First action

    List 30 US biotech companies training their own models.

    How to measure it

    Life sciences opportunities

    Now
    –
    Target
    10
    By
    4 quarters

    Steps

    1. 1

      Publish story

      SK Biopharm in English.

      Head of Marketing 2 weeks

    2. 2

      Define offer

      Clusters, isolation, storage.

      Head of Sales 1 month

    3. 3

      Reference architecture

      Drug discovery models.

      Forward Deployed Engineer (Senior) 1 month

    4. 4

      Target list

      US biotech AI teams.

      Account Executive 1 month

    5. 5

      Events

      One biotech AI event.

      Head of Marketing 1 quarter

    6. 6

      Review

      Pipeline and fit.

      Head of Sales quarterly

    First result
    4 quarters
    Cost
    Sales and engineering time
    Owner
    Head of Sales

    Smallest test

    Send the story to ten biotech AI leads.

    Build the prospect list. The list compiled from public sources, verified, deduplicated and delivered to your CRM. Usually live within days.

    Book a working session

    Why it works

    Package a life sciences offer: dedicated clusters, data isolation, storage for large datasets, and a reference architecture for drug discovery models. Publish the SK Biopharm story in English and target US biotech AI teams.

    An industry offer with a named reference opens a well-funded segment beyond AI labs.

    Risks and how to handle them

    • Regulated data needs

      Agree data handling in each contract.

    Builds on

    Evidence

    • Your websiteKorean blog, September 2, 2026: VESSL supplies SK Biopharm with 128 NVIDIA B200 GPUs on 16 nodes, running since August 2026, for drug research and pretraining foundation models on medical data, extending VESSL from AI companies and research institutes into pharma and biotech.Source
    • HiringHead of Sales (San Francisco, posted September 14, 2026): "our founding AE is closing, and demand is outrunning our selling capacity"; one owner "per segment across physical AI, frontier and AI native labs, academia, and enterprise"; "term structures spanning short-term on-demand credits through multi-year reserved capacity commitments"; "In a GPU cloud business, selling and capacity allocation are the same conversation".Source
    • Your websitePricing page: on-demand H100 SXM at $2.98/hr, A100 SXM at $1.48/hr and L40S at $1.80/hr, per-second billing, "No quotas or usage limits"; B300, B200, GB300, H200 and H100 VM Clusters are "Quoted by term" with "no self-serve rate"; Provisioned Throughput at $0.05 per PTU per minute; SOC 2 Type II; a monthly cost estimator.Source

Wildcards 3

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

  • 21

    VESSL's blog runs multi-part technical series, and its YouTube channel has 83 videos and 2,440 subscribers. Researchers often learn methods from video.

    Medium impactLow effort First result in 1 quarter Head of Marketing

    First action

    Record the speculative decoding explainer.

    How to measure it

    Sign-ups from YouTube

    Now
    –
    Target
    100 a month
    By
    2 quarters

    Steps

    1. 1

      Pick series

      Speculative decoding first.

      Head of Marketing 1 day

    2. 2

      Script

      With the author.

      Forward Deployed Engineer (Senior) 1 week

    3. 3

      Record

      Code on VESSL Cloud.

      Forward Deployed Engineer (Senior) 1 week

    4. 4

      Edit

      Long and short.

      Brand Designer (Senior) 1 week

    5. 5

      Publish

      With notebook links.

      Growth Marketer 1 day

    6. 6

      Measure

      Views and sign-ups.

      Growth Marketer monthly

    First result
    1 quarter
    Cost
    Production time
    Owner
    Head of Marketing

    Smallest test

    Publish one explainer.

    Why it works

    Record a 10-15 minute explainer for each series part, with code running on VESSL Cloud, link the notebook, and cut short clips for X.

    Technical video builds trust with researchers and shows the product in use without a sales pitch.

    Risks and how to handle them

    • Low views at first

      Embed videos in the blog posts.

    Evidence

    • Public channelsChannels linked from the site: GitHub (48 public repositories, 153 stars in total), YouTube (2,440 subscribers, 83 videos), LinkedIn and X; no Discord or Reddit community. Documentation and changelog on docs.cloud.vessl.ai; a status page; a Vanta trust center.Source
    • Your websiteSitemap (checked October 22, 2026): 39 English blog posts against 113 Korean ones. Only one English customer story (ScatterLab). Korean-only posts include VESSL running global inference for Upstage's Solar Pro 4, supplying 128 B200 GPUs to SK Biopharm, the WEF Technology Pioneer 2026 selection, 3.4x revenue growth, a 2026 GPU cloud buying guide, a Google Colab alternatives guide, a five-year GPU TCO guide and a VLA fine-tuning guide for Isaac GR00T. The English addresses of these posts redirect to the blog index.Source
    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
  • 22

    Real Chrome visitors pass Core Web Vitals (LCP 1.6 s, INP 81 ms), but Google's mobile lab test scores the homepage 31. The homepage repeats the full GPU card set several times in the page.

    Low impactLow effort First result in 2 weeks Growth Marketer

    First action

    Run PageSpeed and list the heaviest elements.

    How to measure it

    Homepage mobile lab score

    Now
    –
    Target
    60+
    By
    1 quarter

    Steps

    1. 1

      Measure

      PageSpeed details.

      Growth Marketer 1 day

    2. 2

      Fix repeats

      One GPU card set.

      Growth Marketer 2 days

    3. 3

      Defer scripts

      Non-critical tags.

      Growth Marketer 2 days

    4. 4

      Budget

      For new pages.

      Growth Marketer 1 day

    5. 5

      Retest

      Lab and field.

      Growth Marketer 1 week

    6. 6

      Monitor

      Monthly.

      Growth Marketer monthly

    First result
    2 weeks
    Cost
    Developer time
    Owner
    Growth Marketer

    Smallest test

    Remove one duplicate GPU card set and retest.

    Why it works

    Render the GPU cards once, defer non-critical scripts, and set a performance budget for new pages such as case studies and comparisons.

    A lighter page helps visitors on slower connections and keeps new pages fast.

    Risks and how to handle them

    • Breaking the carousel

      Test on mobile before release.

    Evidence

    • PageSpeedGoogle PageSpeed for the English homepage: mobile lab performance score 31 (desktop 62), but real Chrome users pass Core Web Vitals, with Largest Contentful Paint 1.6 s at the 75th percentile, INP 81 ms and CLS 0.Source
  • 23

    The homepage calls VESSL "The AI Cloud for All Models", the careers page "the GPU liquidity layer" and "the orchestration layer for AI infrastructure", and the product is VESSL Cloud from VESSL AI. The new Head of Marketing must set "one core story", and a senior brand designer is being hired.

    Medium impactMedium effort First result in 1 quarter Brand Designer

    First action

    List every tagline used on the site and in decks.

    How to measure it

    Pages using the core line

    Now
    –
    Target
    100%
    By
    1 quarter

    Steps

    1. 1

      Collect lines

      Site, decks, profiles.

      Brand Designer (Senior) 1 week

    2. 2

      Decide

      One line and naming.

      Head of Marketing 2 weeks

    3. 3

      Update site

      English and Korean.

      Growth Marketer 2 weeks

    4. 4

      Update decks

      Sales and investors.

      Brand Designer (Senior) 2 weeks

    5. 5

      Brief teams

      Regional guide.

      Head of Marketing 1 week

    6. 6

      Check

      Quarterly audit.

      Brand Designer (Senior) quarterly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Brand Designer (Senior)

    Smallest test

    Align the careers page with the homepage.

    Why it works

    Choose one category line and one name for the product and the company, apply it to every page in English and Korean, the careers page, decks and the trust center, and brief the regional teams.

    Buyers and AI assistants describe a company the way its pages do, and mixed lines blur what VESSL is.

    Risks and how to handle them

    • Losing a known phrase

      Keep "AI cloud" if it drives search.

    Evidence

    • Your websitePositioning differs by page: the homepage says "The AI Cloud for All Models"; the careers page says "the GPU liquidity layer" and "the orchestration layer for AI infrastructure", "$16M+ Total Funding", "100+ Customers", "3.4x Revenue Growth (YoY)", partners Hyundai Motor and Tmap Mobility, and "VESSL Cloud in February 2026".Source
    • HiringHead of Marketing (San Francisco, posted October 9, 2026): "What we don't have is a marketing leader whose job is to own how the market finds, understands, and chooses VESSL." The role will "Own all global messaging and positioning ... one core story on why AI researchers choose us over hyperscalers and neoclouds, localized with regional teams", "arm sales teams in every region with the narrative, case studies, and competitor comparisons", and run "paid, SEO and AEO, content, lifecycle" and the channels researchers use ("X, Reddit, Hacker News, GitHub, Discord").Source
    • HiringIn Seoul: a Sales Operations Specialist (posted September 18, 2026) to rebuild pipeline and forecasting in HubSpot and design operations "for regional expansion"; a Senior Brand Designer (posted August 26) working with marketing, sales, PR and IR; a Project Manager (posted October 16) building "Customer RFI, Deal Readiness Review, Go-live Checklist"; and Forward Deployed Engineers.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 23 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.

Prospect lists1The list compiled from public sources, verified, deduplicated and delivered to your CRM.

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.

  • Hudl growth strategySports video, data and fan technology: AI cameras (Focus), video and analysis software (Hudl, Sportscode, Assist, Instat, Fastmodel…22 growth ideas
  • Townsquare Media growth strategyA digital-first local media company: Townsquare Ignite programmatic and owned digital advertising, Townsquare Interactive subscription…21 growth ideas
  • The Joinary growth strategyAn Australian automation and integration consultancy (formerly CBO) that builds, fixes and maintains Zapier, Make and n8n workflows…24 growth ideas
  • Freya growth strategyVoice AI agents that answer and make customer calls for banks and insurers, built on Freya's own speech-to-text and text-to-speech models…23 growth ideas
  • traceroot growth strategyAn open-source platform that traces AI agents in production, flags bad runs with detectors, groups them into Signals with a root cause…23 growth ideas
  • Pursuit growth strategyA New York nonprofit and public benefit corporation that trains adults from low-income backgrounds, mostly without degrees, for AI-economy…22 growth ideas
  • Stigg growth strategyThe usage runtime for AI products: real-time entitlements, AI credits, usage metering, spend governance, contracts and invoicing in one…22 growth ideas
  • CourseStorm growth strategyClass and camp registration and marketing software for informal education: catalog websites, registration and payments, rosters and…24 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on October 23, 2026, and cites them throughout. VESSL AI did not commission, review or endorse it, and Programmable is not affiliated with VESSL AI. 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 VESSL AI runs its business.

Something wrong, or do you represent VESSL AI? 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 307 ideas

    English case studies, the sales form, research credits, usage alerts, comparison pages and published lead times; start the brand line.

  2. Days 31 to 6012 ideas

    English guides, Discord, paper ambassadors, field events, standard terms, physical AI content, lifecycle emails, the funnel dashboard, the model request board and launch-day inference; video and site speed.

  3. Days 61 to 904 ideas

    Commit to the GPU index, campus agreements, region pages and the life sciences offer.

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 no quotas and per-second billing. The most cited problem: waiting for GPU quotas and capacity.

  • Access to Blackwell and multi-node clusters
  • Rigid contracts with one provider
  • In their words: “no quotas”

Market

Competitive landscape

5 competitors compared: Lambda, CoreWeave, Nebius, Runpod and 1 more.

  • Gap to exploit: Clear GPU availability and lead times
  • Lambda: Single-provider supply
  • CoreWeave: Built for the largest buyers

Hiring signals

VESSL is building its US go-to-market from almost zero while GPU capacity triples.

  • Head of Marketing: Positioning, case studies, SEO/AEO, lifecycle, community and credits. Ideas 1, 3, 5, 7, 8, 9, 10.
  • Head of Sales: Segment owners, reserved terms, capacity allocation. Ideas 2, 6, 11, 16.
  • Account Executive: First US seller; convert cloud.vessl.ai usage. Ideas 4, 14.

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 firstNew accounts get 100,000 free enrichment credits with code BETA100K (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)
  • ProgrammableMediaAn AI note taker, interpreter and webinar host for your calls (opens programmable.media 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