fal is hiring a hunter AE for Compute because customers are "coming to us for raw compute", plus an infrastructure product marketer and a content manager for "landing pages" and "solution pages". But Compute has no product page: every Compute link on the homepage, and fal.ai/compute itself, lands on a documentation introduction. That page lists only 1x and 8x H100 instances, while the homepage promises "interconnected clusters of H100, H200, and B200 GPUs" across "hundreds or thousands of GPUs". The pricing page puts Compute in one table with Serverless and sends custom deals to [email protected]. A neolab founder comparing GPU clouds finds specs and an email address, not a way to ask for a cluster.
First action
Agree with the Compute team which GPU types and cluster sizes can be promised on the page.
How to measure it
Qualified Compute quote requests
- Now
- –
- Target
- 40 a month
- By
- 90 days
Steps
Agree the offer
GPU types, cluster sizes, terms that can be sold today.
Product Marketer, Infrastructure 3 days
Write the page
Specs, terms, use cases, serving path and proof.
Technical Content Marketing Manager, Infrastructure 1 week
Design and build
Product page in the Serverless style.
Senior Product Designer 1 week
Build the quote form
GPU type, count, start date, term; route to the Compute AE.
GTM Operations Engineer 2 days
Repoint links
Homepage, pricing and nav to /compute; docs stay for setup.
Senior Product Designer 1 day
Add proof
One training customer story or quote.
Technical Content Marketing Manager, Infrastructure 3 weeks
Measure
Quotes, meetings and pipeline from the page.
GTM Operations Engineer weekly
Why it works
Build fal.ai/compute as a product page: the GPU generations and cluster sizes actually offered, on-demand vs reserved terms, interconnect and storage, what training and fine-tuning teams run on it, how Compute connects to Serverless for serving the trained model, and one customer proof point. End it with a short quote form (GPU type, count, start date, term) that goes straight to the Compute AE, not support. Point the homepage Compute links and the pricing page at it, and keep the docs page for setup.
Buyers of reserved GPU capacity compare providers on a single page and then ask for a quote. A page that says what is available and takes the request turns the traffic fal already has into Compute pipeline for the new AE, and gives the infrastructure team the landing page its SEO and content plans need.
Ready to use
ChecklistCompute page outline
[ ] Headline for training and fine-tuning teams [ ] GPU generations and cluster sizes offered now [ ] On-demand vs reserved terms [ ] Interconnect, storage and regions [ ] Train on Compute, serve on Serverless [ ] One customer proof point [ ] Quote form: GPU, count, start date, term [ ] Link to docs for setup
Risks and how to handle them
Promising capacity that is not available
List only configurations the Compute team signs off, with "contact us" for larger clusters.
Cannibalising Serverless
Explain clearly which workloads fit each product.
Evidence
- HiringAI Native Account Executive, Compute (posted October 19, 2026): "More and more of them are coming to us for raw compute as well as inference and model APIs." The AE sells "dedicated GPU clusters, reserved and on-demand capacity, and the inference platform", owns "all net-new logo acquisition across a defined set of AI-native startups, neolabs and research-driven model builders", and builds pipeline "through founder and investor networks, targeted outbound, community and events". "This is a pure hunter role. You won't inherit a book."Source
- Your websiteHomepage (checked October 22, 2026): "Trusted by over 2,500,000 developers"; "Access 1,500+ generative media models"; the Compute section says "Access large, interconnected clusters of H100, H200, and B200 GPUs" to "Scale workloads across hundreds or thousands of GPUs". All four Compute links on the homepage go to the documentation (fal.ai/compute redirects to /docs/documentation/compute); Serverless has its own product page. Quotes from Black Forest Labs, Canva and Quora ("fal powers 40% of Poe's official image and video generation bots").Source
- Your websiteCompute documentation: "Dedicated GPU instances for training, fine-tuning"; "Compute runs on NVIDIA H100 SXM GPUs"; two instance types, 1xH100-SXM and 8xH100-SXM, with InfiniBand between 8-GPU instances in the same sector; "billed per-hour at fixed rates". The page is a technical introduction with no contact or quote form.Source
- Your websitePricing page: a single "Serverless & Compute Pricing" table with list and "as low as" hourly prices (B300 $12.99/h list, as low as $5.99/h; B200 $7.99/h, as low as $5.49/h; H200 $6.00/h, as low as $2.99/h; H100 $4.50/h, as low as $2.49/h); for custom deployments it says "Contact [email protected] to get started". There is no separate section for reserved clusters, multi-node training or commitment terms.Source
- HiringProduct Marketer, Infrastructure (posted October 5, 2026): owns GTM for "Serverless for deploying and scaling custom AI models across thousands of GPUs, to Compute for training on large, interconnected GPU clusters"; must "Develop our competitive positioning across inference platforms, GPU clouds, hyperscalers, and in-house infrastructure", "Own SEO and AEO strategy for infrastructure, capturing developers actively searching for inference, training, and compute solutions" and create "benchmarks, product launches, solution pages" and "case studies".Source
- HiringTechnical Content Marketing Manager, Infrastructure (posted October 5, 2026): "Some of the most interesting work happening at fal is deep inside the infrastructure, and most of the market never sees it." Writes "technical blogs, landing pages, guides, tutorials, solution pages" for Serverless and Compute, "benchmarks and competitive content", and "technical case studies".Source