TL;DR
A new Thorsten Meyer AI analysis says the AI compute market has become a circular rental system linking AI labs, neoclouds, chip suppliers and investors. The confirmed deals include major SpaceX/xAI compute leases with Anthropic and Google, while the broader cartel label is an interpretation of market concentration, not a legal finding.
A new Thorsten Meyer AI report says the AI industry’s race for compute has shifted into a circular rental market, with OpenAI, Anthropic, SpaceX/xAI, CoreWeave, Nvidia, AMD and Microsoft tied through leases, chip orders and equity-linked financing. The finding matters because GPU access now shapes which AI products can scale and who carries the financial risk if demand slows.
The report’s central development is the rise of the neocloud: AI-focused cloud companies that rent large GPU clusters to labs that cannot or do not want to own every machine they use. CoreWeave is described as the largest of that group, with a contracted backlog above $55 billion; MarketWatch reported CoreWeave’s backlog rose to $55.6 billion in late 2025.
The strongest recent example is SpaceX/xAI turning Colossus capacity into outside revenue. Business Insider reported, citing SpaceX’s S-1 and an Anthropic spokesperson, that Anthropic agreed to pay SpaceX $1.25 billion a month for compute capacity through May 2029, with termination rights. Business Insider also reported that Google agreed to pay $920 million a month for SpaceX compute capacity beginning in October 2026.
The report also points to supplier financing as a second loop. Nvidia announced plans in 2025 to invest up to $100 billion in OpenAI as part of a 10-gigawatt infrastructure buildout, while OpenAI’s AMD deal included warrants for up to 160 million AMD shares, according to reports on the companies’ agreement. These are reported commitments, many spread over years; they are not the same as cash already spent.
The Neocloud Cartel
Almost no one racing to build AI owns the machine it runs on. They rent — increasingly from each other — and the money loops back to one chip maker that’s also an investor in nearly everyone at the table.
The cartel isn’t a conspiracy — it’s the endpoint of extreme capital intensity, real scarcity, and one dominant supplier. But the same circularity that makes it powerful makes it a fuse: each cancelled order is someone else’s missing revenue. Don’t be a price-taker at the bottom of a loop you don’t control — own your inference, keep an open-weight fallback, diversify silicon.
Compute Risk Hits Every AI User
The issue for readers is not only who has the best model. It is who controls the machines needed to run those models, who can afford them, and what happens when the same firms are customers, suppliers, landlords and investors.
If a small set of companies controls access to scarce GPUs, AI builders may face higher prices, tighter contract terms and less room to switch providers. If demand weakens, a canceled lease or delayed chip order can become missing revenue for another company in the same loop.
The report uses the phrase neocloud cartel to describe market concentration and circular dependence. It does not establish unlawful coordination. The confirmed record supports a picture of intense capital needs, scarce hardware and heavy reliance on Nvidia-centered supply chains.
GPU cloud computing services
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Neoclouds Became Landlords
The neocloud category grew out of the 2024-25 GPU shortage, when AI labs faced long waits for Nvidia hardware and needed faster ways to train and serve models. Renting capacity let labs scale without building every data center themselves.
The source material lists CoreWeave, Nebius, Crusoe, Lambda, Together, Fireworks, Nscale and IREN among the firms renting AI compute. It also says CoreWeave drew 77% of revenue from two customers, a claim that points to customer concentration risk if large buyers slow orders.
By 2026, the pattern had widened beyond specialist GPU clouds. SpaceX/xAI, built around its own Grok ambitions, began renting Colossus capacity to other AI companies. That made the market’s shift plain: ownership of a cluster and use of a cluster are no longer the same thing.
“If it seems circular, it is.”
— Thorsten Meyer AI, The Control Series
AI training GPU rental
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Lease Terms And Cash Flows
Several points remain unresolved. It is not yet clear how much of the reported trillion-dollar-scale compute commitments will turn into actual spending, how long short-cancelable leases will last, or whether AI revenue can grow fast enough to support the buildout.
It is also unclear how regulators will treat supplier-funded customer deals. The available record shows circular financing and market concentration, but no confirmed legal finding of cartel conduct.
high performance GPU server
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Contracts Face Their First Stress Test
The next signals will come from SpaceX filings, CoreWeave earnings, Nvidia and OpenAI’s partnership status, AMD warrant milestones, and any regulator inquiries into circular AI financing. The main test is whether GPU demand keeps absorbing new capacity or whether falling rental rates expose weak links in the loop.
cloud GPU for AI development
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news here?
The news is a new analysis tying recent AI compute deals into one market pattern: major AI companies are renting GPU capacity from neoclouds, suppliers and competitors while suppliers also finance customers.
What is a neocloud?
A neocloud is an AI-focused cloud provider that rents GPU capacity, usually Nvidia-based, without operating like a full legacy cloud platform. CoreWeave is the best-known example cited in the report.
Does this mean the companies formed an illegal cartel?
No legal finding is confirmed. The report uses cartel as a market-structure description for concentration, circular financing and shared dependence on scarce compute.
Why do AI labs rent instead of own?
Renting can give labs faster access to scarce GPUs and lets them scale before building their own data centers. The tradeoff is dependence on another company’s prices, capacity and contract terms.
What should AI customers watch now?
Customers should watch compute pricing, model availability, contract lock-ins, silicon options beyond Nvidia, and whether providers can keep service stable if a major lease or chip order changes.
Source: Thorsten Meyer AI
