Anthropic puts 200 billion back into Google’s pocket: the most proper “left hand swaps with the right hand” in the AI era

Author: Ada, Deep Tide TechFlow

On May 5th, according to The Information, Anthropic committed to paying Google Cloud $200 billion over the next five years.

This multi-year agreement starts in 2027 and will account for over 40% of Google Cloud’s deferred revenue, reflecting contractual commitments from enterprise clients.

A company that didn’t exist five years ago, with just a contract, has taken nearly half of Google Cloud’s future revenue.

On the day the news broke, Alphabet’s stock rose 2% after hours.

But what’s more intriguing is another number. At the same time, Alphabet made a reverse investment in Anthropic, with a maximum of $40 billion.

Money leaves Google’s account, circles around, and returns to Google’s account. In the middle, there’s an accounting item called “Anthropic compute expenditure.”

So, is this the biggest cloud computing order in history, or the most elegant financial magic trick ever?

An “Exclusive Commitment” Not Just for Google

To understand the essence of this deal, let’s look at a set of not-isolated data points.

On April 20th, Anthropic announced an expanded partnership with Amazon, committing to spend over $100 billion on AWS technology over the next 10 years in exchange for up to 5 gigawatts of compute power. In return, Amazon will add up to $25 billion on top of its existing $8 billion investment.

Last November, Microsoft agreed to invest up to $5 billion in Anthropic, which in turn committed to purchasing $30 billion worth of Azure compute capacity.

In other words, Google: invest $40 billion, receive $200 billion. Amazon: invest $33 billion, receive over $100 billion. Microsoft: invest $5 billion, receive $30 billion.

Together, these three cloud giants are spending about $78 billion to secure a “contract commitment” worth $330 billion, with a net cash inflow of $250 billion on the books.

The core of this approach is turning capital expenditure into revenue. Money invested in Anthropic is recorded as investing activity cash flow, while the compute fees paid by Anthropic are counted as core revenue. The same money flows from one pocket to the other, creating a beautiful backlog on the financial statements.

Alphabet is both injecting capital into Anthropic and counting Anthropic’s compute procurement as future revenue, creating a self-reinforcing cycle of AI infrastructure prosperity.

Wall Street is the real winner in this game—if the backlog number is big enough, the P/E ratio can be sustained.

A High-Level Version of the Flywheel

The story of Strategy increasing positions at high levels isn’t finished, and the AI circle has amplified this same flywheel a thousand times.

Strategy’s logic is to issue stocks to raise funds, buy Bitcoin, let the Bitcoin price rise to boost market cap, issue more stocks, and buy more Bitcoin.

The cloud providers’ logic is to invest in AI companies, which pay for compute, leading to revenue growth, stock price increases, and more capital market investment, continuing the cycle.

The difference is that Bitcoin is a scarce asset, each unit representing real supply on the chain. Compute power isn’t. The “multi-gigawatt TPU capacity” coming online in 2027 doesn’t even have racks installed today.

In other words, a significant part of the $200 billion is Anthropic’s advance commitment to purchase a batch of chips that haven’t been manufactured yet, with Google using this promise to persuade the capital market.

Isn’t this just a forward contract? The difference is that commodity futures have delivery dates and margin requirements, but this contract doesn’t. What happens if Anthropic can’t pay in 2027? Who bears the breach costs?

It won’t be Google. It has already included the backlog in its earnings call PPT. On April 29th, Alphabet disclosed that Google Cloud revenue grew 63% year-over-year to over $20 billion, with cloud backlog reaching about $462 billion. This number supports Alphabet’s current market value.

It won’t be Anthropic either. As long as it keeps raising funds, the next valuation round will still be higher.

In the end, the retail investors who think they’re buying the “AI shovel seller” story might be the ones footing the bill.

$5 Billion Leverage for $330 Billion

Does Anthropic’s scale justify this number?

According to media reports, Anthropic’s annual revenue grew from $1 billion in 2023 to $5 billion in 2025.

A company with only $5 billion in annual revenue signing a 5-year, $200 billion, 10-year, $1 trillion, plus $30 billion contract—totaling $3.3 trillion.

Even if Anthropic’s revenue increases tenfold, it still wouldn’t reach $3.3 trillion over five years.

So, where does the money come from?

The only way is continued fundraising.

And the biggest potential investors are precisely these three cloud providers themselves.

That’s the entire secret of the cycle. Anthropic doesn’t need to actually make money; it just needs to stay “continually raising funds,” treating each new round of capital as next year’s compute bill. As valuations go up, it can raise even more.

Who does this sound like?

Strategy. It also doesn’t need Bitcoin to generate real cash flow, only to maintain the ability to issue stocks and bonds indefinitely. The only difference is that Strategy’s balance sheet includes Bitcoin, a globally priced asset.

The valuation logic of AI companies now resembles that of SaaS companies in 2021. Back then, everyone was chasing ARR; today, it’s about compute commitments. Essentially, it’s all about discounting the future to the present—only the question is whether the future will actually materialize.

What is OpenAI doing?

In the same 8-K filing where Amazon added to Anthropic, OpenAI also committed to consuming about 2 gigawatts of Trainium compute via AWS infrastructure starting in 2027.

Two months ago, Amazon invested $50 billion in OpenAI and signed a $100 billion cloud computing contract.

The script is exactly the same.

In other words, the three cloud giants and two model companies are playing the same game multiple times. Each time, headlines tout “the largest-ever,” “strategic partnership,” and “compute revolution.”

Behind each is the same money circulating.

So, who will be the first to stop?

It won’t be the cloud providers—they rely on this narrative for their current market value. Alphabet has raised its 2026 capital expenditure guidance to as much as $190 billion, a scale that must be “hedged” with Anthropic and OpenAI revenue, or Wall Street won’t approve.

It won’t be the model companies either—stopping means losing access to the next funding round, which could be fatal.

The first to be kicked out might be the second-tier players who haven’t aligned properly.

Will the concert stop?

The fragility of all this lies in the word “兑现” (realization/fulfillment).

In 2027, when TPU comes online, if Claude’s commercialization can’t keep pace with compute expansion, what will Anthropic use to digest this $200 billion?

If a contract is renegotiated, canceled, or split, Google Cloud’s backlog of $462 billion will immediately be exposed.

But today, no one wants to be the first to break the silence. CFOs are writing guidance, analysts are issuing buy ratings, CEOs are cautious during earnings calls. Everyone is betting that before the music stops, they’ve already secured the closest seat.

It’s not about whether it’s a bubble or not; it’s about how to deflate the bubble. Everyone knows this is a cycle of trading, but everyone also knows that as long as the AI story continues, no one dares to short the backlog.

Contracts are written on paper, money circulates among the three companies, valuations bounce between primary and secondary markets. Everyone holds a “future promise,” treating it as “current asset.”

Until one day in the future, a company’s earnings fall short. At that moment, the $200 billion could suddenly have a different name—or become a liability.

Until then, the celebration continues.