Compute clouds
Renting AI computing by the hour or on multi-year contracts, from the big clouds and from GPU-only clouds (neoclouds).
Who is making money
The companies earning or raising real money in this part of AI. Suppliers are marked: they sell to the builders rather than to end users.
Google Cloud rents its TPU AI chips and Gemini models to businesses.
25% of AI chip supply (Epoch AI)
Open profileIts Azure cloud rents AI computing to OpenAI and others, and it buys extra capacity from GPU clouds.
20M+ users (2025)
Open profileAWS is the biggest cloud, and it rents out its Trainium chips alongside Nvidia GPUs.
5% of AI chip supply (Epoch AI)
Open profileGPU cloud
Up to 1.2 GW campus under way in Missouri
Open profileRuns large cloud data centres and rents AI computing, with OpenAI as its biggest customer.
Capital spending up 162% to $55.7B in fiscal 2026. Free cash flow minus $23.7B
Open profilethe biggest GPU cloud
Quarterly revenue $2.6B, up 112%. $104B backlog. Still loss-making
Open profileBuilds AI servers from Nvidia and AMD chips for enterprises, clouds and governments.
Full-year fiscal 2027 guidance raised to $192B (Sep 2026)
Open profileGPU cloud
About $760M annualised in 2025 (Sacra estimate)
Open profileHosts and fine-tunes open AI models on rented GPUs for companies.
About $1B annualised (Sacra estimate, Feb 2026)
Open profileAI data centres and cloud
Raised $3.9B at a $30.9B valuation (Sep 2026)
Open profileBuilds AI servers for big cloud firms, mainly Nvidia GB300 systems.
NT$809.2B in Q1 2026, up 66.6%
Open profileBuilds AI servers and racks from Nvidia and AMD parts, fast and to order.
$10.2B in the quarter to March 2026, up 122.7%. Full-year outlook at least $40B (Feb 2026)
Open profileDesigns and builds AI server racks for large cloud customers, part of the Wistron group.
NT$276.5B in Q1 2026, up 62%
Open profileThe biggest contract electronics maker, and the main assembler of Nvidia's AI server racks.
Cloud and networking (mostly AI servers) was 51% of Q2 2026 revenue, the first time above half
Open profileBuilds AI servers for enterprises and clouds, including Nvidia-based systems.
Open profileBuilds AI servers for enterprises and some cloud customers.
Open profileWhere the money flows
AI labs and companies pay clouds for computing. GPU clouds sign take-or-pay contracts with a few labs, then borrow against them to buy chips.
rents GPU capacity from CoreWeave on long contracts
chartmill.combuys dedicated AI capacity from Nebius's Vineland, New Jersey site
convergedigest.combuys AI capacity from Nebius, with a larger firm contract from early 2027
convergedigest.compays Oracle for cloud computing under the Stargate project, starting in 2027
datacenterdynamics.comcommits to Trainium AI capacity through AWS
letsdatascience.comuses tens of thousands of Lambda's Nvidia GB300 systems through a leasing deal
sacra.com
The openings
Gaps still open in this area, from ones a small business could take to ones only giants can play.
- Small business Cost tracking and allocation tools for companies renting AI compute
- Startup money Marketplaces for spare GPU capacity, including local NZ and Australian capacity
- Giants only A regional GPU cloud at scale
CoreWeave's debt now trades at distressed levels (some bonds at 15% to 16%), and its growth depends on a few customers while it still loses money.
Sources: datacenterdynamics.com, pulse2.com, zawya.com, letsdatascience.com
The map updates three times a week.
241 companies across 25 areas, 62 money flows and 89 openings so far. New research lands every Tuesday, Thursday and Saturday.
Back to the full mapFree daily email
Get the map updates in your inbox
A short read every morning on what is moving in AI, including new areas of the map as they go live. Free.
You are in
Your first AI Pulse lands tomorrow morning. Keep an eye on your inbox.