The Compute Cartel: How Microsoft, Amazon, and Google Control Every Major AI Lab

TLDR:

• Amazon, Microsoft, and Google invest billions in AI labs, then collect that money back as locked-in cloud revenue.
  
• Within 16 days, three megadeals handed Anthropic compute from Amazon, Google, and SpaceX simultaneously.
  
• xAI failed to operate independently and was forced to lease its GPU data center directly to rival Anthropic.
  
• Regulators in the US, EU, and UK are now probing whether these compute deals amount to backdoor monopolies.

Major AI labs — OpenAI, Anthropic, and xAI — are often presented as fierce competitors racing to dominate artificial intelligence.

However, a closer look at recent financial deals tells a different story. All three companies rely on the same small group of cloud infrastructure providers: Microsoft, Amazon, and Google.

These tech giants invest billions into AI labs, then collect that money back as cloud revenue. The arrangement is raising serious antitrust concerns around the world.

The Deals That Changed the AI Landscape

Within just 16 days in April and May 2026, three major compute deals reshaped the AI industry. Amazon expanded its Anthropic commitment to $13 billion, tied to $100 billion in AWS cloud spend over ten years.

Google followed with up to $40 billion in equity plus five gigawatts of TPU capacity for Anthropic. Then SpaceX, which had absorbed xAI, handed Anthropic its entire Colossus 1 data center — over 220,000 Nvidia GPUs and 300 megawatts of power.

The financial structure behind these deals follows a repeating pattern. Tech giants invest in AI labs, then require those labs to spend the money back on their own cloud services.

Microsoft’s total OpenAI investment reportedly exceeds $100 billion when Azure infrastructure is counted. In return, OpenAI committed $250 billion in cloud spend back to Azure over a decade.

Amazon also invested $50 billion into OpenAI in February 2026, making it the only hyperscaler holding equity stakes in both Anthropic and OpenAI simultaneously.

As @coinbureau noted, “OpenAI and Anthropic are not rivals, they are financially bound tenants to the same hyperscaler landlords.”

The financial results of this arrangement became clear in Q1 2026. Alphabet reported $37.7 billion in other income, largely from unrealized gains on its Anthropic and SpaceX stakes.

Amazon reported $16.8 billion in non-operating income, also driven by its Anthropic stake. These are paper gains—not cash—and they could reverse if Anthropic’s next funding round prices lower.

When Independence Becomes Too Expensive

xAI’s story offers a cautionary example of operating outside this structure. Grok’s model utilization was reported at around 11%, far below the 40% achieved by competitors.

Unable to sustain infrastructure costs independently, xAI was absorbed by SpaceX, which then leased the Colossus 1 data center directly to Anthropic — the company xAI was built to compete against.

Meta stands apart as the only major frontier AI player operating without cloud dependency. The company is running $125 billion to $145 billion in capital expenditure for 2026, building its own data centers and holding no equity ties to a single cloud partner.

Regulators are now responding to this concentrated structure. The FTC is investigating Amazon and Microsoft’s deals as potential backdoor mergers. The EU is enforcing antitrust rules against exclusive cloud contracts.

The UK has flagged over 90 cross-pollinating partnerships among Big Tech companies. Even if the AI compute stack is broken apart, the same four companies — Microsoft, Amazon, Google, and Meta — are positioned to remain dominant at the foundational layer.