Corporate filings reveal that Microsoft's $13 billion investment in OpenAI returned to Microsoft as cloud credits — which OpenAI spent back on Azure compute — and Microsoft then recorded those same dollars as fresh revenue. Google's Q1 2026 profit included $28.7 billion in unrealized gains tied to its Anthropic stake. Oracle's $553 billion pipeline is 54% dependent on a single AI startup: OpenAI.

This is the AI revenue loop — and it's raising uncomfortable questions for CFOs evaluating cloud vendor deals.

The Circular Flow Mechanics

Here's how the round-trip revenue model works:

Step 1: A cloud giant invests billions in an AI startup (Microsoft → OpenAI, Google → Anthropic, Oracle → OpenAI).

Step 2: The investment comes with a requirement: the startup must spend those funds on the investor's cloud infrastructure.

Step 3: The startup spends the money back on compute credits from the same company that funded them.

Step 4: The cloud giant records those expenditures as revenue growth in their cloud division.

The same money gets counted twice — once as an investment, once as revenue.

BullTheoryio summarized it plainly on Twitter: "A tech giant gives billions of dollars to an AI startup as an 'investment'. But hidden in the contract is a strict rule forcing the startup to hand that exact same money straight back to the tech giant to rent their computer servers."

The $2 Trillion Cloud Backlog Problem

OpenAI and Anthropic together account for over 50% of the $2 trillion cloud backlog held by Microsoft, Oracle, Google, and Amazon combined.

Let's break down the dependency:

• Microsoft: $627 billion in future cloud commitments tied to OpenAI
• Oracle: 54% of its entire $553 billion pipeline relying on OpenAI alone
• Google: $28.7 billion in unrealized paper gains from Anthropic (Q1 2026)
• CoreWeave: 62% of revenue from Microsoft in 2024 (up from 35% in 2023)

The red flag: These aren't diversified revenue streams. They're concentrated bets on a handful of AI startups that can only spend money their investors gave them.

What This Means for CFOs

If you're evaluating a cloud vendor deal — especially for AI workloads — here are the questions your finance team should be asking:

1. Is this cloud credit or cash?

If the vendor is offering cloud credits tied to their own infrastructure, you're locked into their ecosystem. That's vendor lock-in disguised as investment.

2. What percentage of the vendor's AI revenue comes from their own funded startups?

If Microsoft's Azure AI revenue is heavily dependent on OpenAI spend — and OpenAI can only spend what Microsoft gave them — how sustainable is that growth?

3. Are you comparing real TCO or circular pricing?

When a vendor subsidizes their own customer's infrastructure spend, their public pricing doesn't reflect what you'll actually pay as an outside customer.

4. What happens when the funding rounds stop?

Anthropic is planning to consume one million TPUs and over a gigawatt of capacity by 2026 while projecting $11 billion in losses for the same year. If the venture capital dries up, can they keep buying compute at the same scale?

The CoreWeave Case Study

CoreWeave is the poster child for circular revenue concerns.

In 2024, Microsoft accounted for 62% of CoreWeave's revenue — up from 35% in 2023. Sacra estimated CoreWeave's 2024 revenue at $2 billion, partly driven by a new $10 billion contract with Microsoft/OpenAI.

But in March 2025, Microsoft began scaling back agreements with CoreWeave due to delivery issues and missed deadlines. CoreWeave denied "contract cancellations," but the concentration risk was undeniable.

By early 2026, securities class-action lAWSuits alleged that CoreWeave overstated its ability to meet customer demand and concealed construction delays from investors.

The lesson: When your largest customer is also your largest investor, the dependency cuts both ways.

Why This Matters for Technical Leaders

For CIOs and CTOs, the circular revenue problem creates three operational risks:

1. Vendor Viability Risk

If Oracle's pipeline is 54% dependent on OpenAI, what happens if OpenAI switches cloud providers? (They already renegotiated their Microsoft deal in April 2026 to allow multi-cloud flexibility.)

2. Pricing Transparency Risk

When a vendor's largest customer gets subsidized pricing (via cloud credits), you have no benchmark for what fair market pricing looks like.

3. Innovation Risk

If a cloud provider's AI roadmap is driven by one or two startup customers, their product priorities may not align with broader enterprise needs.

The SEC Question Nobody's Asking

In traditional industries, round-trip transactions — where money flows from Company A to Company B and back to Company A — require careful disclosure.

The question: Should cloud giants be required to disclose what percentage of their AI cloud revenue comes from startups they themselves funded?

Right now, that number is buried in corporate filings. Most investors (and most CFOs) don't know that over half of Big Tech's $2 trillion AI cloud backlog is tied to two companies: OpenAI and Anthropic.

What Real Growth Looks Like

To be clear: Not all AI cloud growth is circular.

AWS's AI revenue comes from thousands of independent customers. Microsoft's GitHub Copilot revenue ($1.5 billion annualized as of Q4 2025) comes from developers who chose the product, not from cloud credits tied to an investment round.

The difference is organic vs. contractual demand.

Organic demand: A customer evaluates options, picks the best fit, and pays with their own budget.

Contractual demand: A customer uses your service because you gave them credits that can only be spent with you.

How to Spot Circular Deals in Your Own Vendor Contracts

If you're negotiating a cloud deal, watch for these patterns:

Red flag #1: "We'll invest in your AI initiative, but you have to use our infrastructure."

Red flag #2: "We're offering $X million in cloud credits (non-transferable, expires in 12 months)."

Red flag #3: "This pricing is exclusive to our strategic partnership — you can't use other cloud providers for this workload."

Green flag: "Pay for what you use. No lock-in. Multi-cloud compatible."

The Bottom Line

The AI boom is real. Enterprise adoption is accelerating. But how much of Big Tech's AI revenue growth is genuine market demand, and how much is money they gave to startups being spent back on their own infrastructure?

For CFOs: Demand transparency on vendor revenue sources before signing multi-year cloud deals.

For CIOs: Build multi-cloud optionality so you're not locked into a vendor whose growth depends on funding rounds, not product quality.

For boards: Ask your finance team to validate that the cloud vendors you're betting on have diversified, organic revenue — not just circular flows from their own investments.

The real question isn't whether Big Tech is investing in AI. It's whether they're investing in customers — or in their own revenue reporting.