Who is buying GPU compute

The Neolab Directory tracks 500+ neolabs and AI labs that rent GPU clusters, with the cloud and GPUs they rent, their funding and who to contact.

Last updated November 3, 2026

Compute, funding and contacts are in the full list.

Who buys GPU compute?

Mostly AI labs that train their own models. Neolabs and model companies rent large GPU clusters on multi-month and multi-year contracts, which makes them the main buyers of neocloud capacity. Enterprises and inference companies buy too, usually in smaller blocks.

What is a neolab?

A neolab is a young AI lab that trains its own models instead of building on someone else's. Many are founded by researchers who left OpenAI, Google DeepMind or Meta, and most rent large GPU clusters to train. This list also covers AI companies that train models in house, from video and voice to biology and robotics.

Neolabs by what they train

Language and reasoning
Thinking Machines LabSafe SuperintelligenceReflection AIMistral AICohereHumans&Ineffable IntelligenceEssential AILiquid AIInception LabsSakana AIDeep Cogito
Code
PoolsideMagicImbueRelace
Image, video and world models
Black Forest LabsLumaRunwayPikaGenmoDecartOdysseyGeneral Intuition
Voice and music
ElevenLabsCartesiaSunoKyutaiDeepgramFish Audio
Robotics
Physical IntelligenceSkild AIFigure AI1X TechnologiesGeneralist AIGenesis AIField AI
Biology
Chai DiscoveryXaira TherapeuticsGenerate BiomedicinesProfluentCradleLatent LabsBoltz
Science and materials
Periodic LabsLila SciencesRadical AICuspAIOrbital Materials
Autonomous driving
WayveWaabi

And the rest of the 500+ in the full list.

Who uses it

Neolabs raise large rounds and spend them on compute, data, tools and people. The list shows you which labs to sell to and who to talk to.

Neoclouds and GPU owners
Find the labs that rent clusters like yours.
Data and evaluation vendors
Reach the labs training models in your field.
Tooling and infrastructure
Get to the founders and infrastructure leads.
Recruiters
See which labs are growing and who runs them.
Investors
Follow new labs, new rounds and new compute deals.

The full list

We help neoclouds find offtakers, so we track who every lab rents compute from. The list adds what each lab trains, how it is funded and who to talk to.

What they train
Models, team size and where they are based.
Compute today
Which cloud they rent from, which GPUs and how many, where it is public.
Funding
Total raised and the last round.
Contacts
Founders and infrastructure leads, where we have them.
Monthly change log
New labs, new rounds and new compute deals.

Full list

$199per month

Every lab, and who they train on.

  • Every lab we track, as a spreadsheet
  • Compute provider and GPU footprint, where public
  • Funding and contacts, where we have them
  • Monthly refresh and change log

Full list + intros

Custom

The list, plus we find the offtaker.

  • Everything in Full list
  • Six introductions a month
  • Equipment financing arranged
  • Custom cuts by model type or region

Need help getting offtake?

We help neoclouds and GPU owners find offtakers. Tell us the GPUs, cluster size, region and timing, and we'll introduce you to the labs that fit.

Questions

Who buys GPU compute?

Mostly AI labs that train their own models. Neolabs and model companies rent large GPU clusters on multi-month and multi-year contracts, which makes them the main buyers of neocloud capacity. Enterprises and inference companies buy too, usually in smaller blocks.

Who is looking to buy GPU compute right now?

Labs that just raised or are scaling up training. A new round is often followed by a new compute contract, so the full list tracks each lab's latest round, the cloud and GPUs it rents today, and a monthly change log of new rounds and compute deals.

Which AI labs are renting B300s?

Only a few labs say so publicly, since most compute contracts are private. The full list records the cloud and GPUs each lab rents where they are public, from H100 to B300 and GB300, and the change log adds new compute deals as they are announced.

What is a neolab?

A neolab is a young AI lab that trains its own models instead of building on someone else's. Many are founded by researchers who left OpenAI, Google DeepMind or Meta, and most rent large GPU clusters to train. This list also covers AI companies that train models in house, from video and voice to biology and robotics.

How many neolabs are there?

We track 500+ AI labs and model companies that train their own models. Most public neolab lists cover the 50 to 100 best-funded labs. Ours also covers the smaller labs training video, voice, robotics, biology and science models.

Which AI labs rent GPU clusters?

Most neolabs rent rather than own, on contracts with neoclouds and hyperscalers for H100, H200, B200 and B300 clusters. The full list shows which cloud each lab trains on and how many GPUs, where it is public.

How do I get offtake for my GPU cluster?

Start with the labs whose training fits your GPUs, cluster size and region, then reach the person who buys their compute. The full list has the compute and contacts for that. If you'd rather we find the offtaker, contact us and we'll make the introductions.

How do I sell to AI labs?

Start with the labs whose models fit what you sell, then reach the founder or the lead who owns your area. The full list has what each lab trains, its team size, funding and contacts.

What is in the full list?

Every lab we track as a spreadsheet, with what they train, headquarters, team size, compute provider and GPU footprint where public, funding, contacts where we have them, and a monthly change log.

How often is the list updated?

Monthly. Each issue has a change log of new labs, new rounds and new compute deals.

Who is the Neolab Directory for?

Anyone who sells to AI labs or invests in them. Neoclouds and GPU owners use it to find offtakers, data and tooling companies to find the labs training models in their field, recruiters to see which labs are growing, and investors to follow new rounds and compute deals.

Looking for capacity instead? Neocloud Directory

Have a question? Contact us