Anthropic commits to $200bn spending on Google Cloud over five years

Evidence of insatiable demand for AI or ill-disguised vendor financing?

Anthropic's cloud spending commitments and hyperscale multi-billion-dollar investments highlight the uncertainties inherent in AI infrastructure boom and raise concerns over future profitability, project delays, and the risks posed by private credit financing.

Anthropic has committed to spend $200 billion with Google Cloud over five years as part of a recent agreement, The Information reported on Tuesday, citing a source with knowledge of the matter.

This commitment indicates that Anthropic now accounts for more than 40% of the revenue backlog Google disclosed to investors last week - a backlog which reflects contractual commitments from cloud customers.

Last month, Anthropic signed another deal with Google and its chip partner Broadcom for multiple gigawatts of tensor processing unit (TPU) capacity, which it expects to come online starting in 2027.

In addition, Google Cloud parent Alphabet is investing another $40bn in Anthropic – adding to the $3bn its already sunk. Of the most recent $40bn, $10bn has been invested immediately with the remaining $30bn contingent on ‘performance milestones.’

It all sounds very circular. Google invests in Anthropic. Anthropic promises to buy compute from Google. Everyone promises to buy chips from Nvidia.

Anthropic also signed a multi-year deal with CoreWeave in April and is apparently due to secure more capacity from Amazon by the end of this year.

Promises, promises

Anthropic’s April deal spree looks impressive until you start looking at how much money Anthropic thinks it’s going to make, which is, according to The Wall Street Journal, $18 billion in 2026, $55 billion in 2027, $102 billion in 2028, and $148 billion in 2029.

But that’s revenue. Anthropic isn’t making a profit, and says it won’t do so until 2028, after losses of $11bn in 2026 and 2027.

It isn’t clear what will change between now and 2028 to enable Anthropic to start turning a profit, and it isn’t clear what ‘performance milestones’ Google has put in place before Anthropic gets its hands on the next $30bn.

Also unclear is when the multiple gigawatts of TPU capacity Anthropic has said it “expects to come online starting in 2027” will actually come online. Datacentres take a very long time to build, especially ones filled with AI infrastructure, because the chips require more power, and thus more cooling than normal cloud infrastructure. Public opposition to datacentres is delaying and reducing projects in size.

The scale of Anthropic’s commitment to Google is still a fraction of OpenAI’s $300bn Stargate deal with Oracle (for which the delivery has been shunted forward several times now) but the aggregate value of the many deals now accounts for a revenue backlog of $2 trillion across Amazon, Google, Microsoft and Oracle, according to The Information.

This $2 trillion will come from just two customers – Anthropic and OpenAI.

Today we’re going to party like its 2007

A revenue backlog is a promise to spend money in the future. What happens if Anthropic or OpenAI breaks those promises? Today, a report by the Financial Stability Board (FSB), a global watchdog has issued a warning about the extent to which the AI infrastructure boom is financed by private credit.

According to the report, the AI industry accounted for more than a third of private credit deals in 2025. The FSB warned that a “sharp correction in asset valuations, which have increased rapidly, could lead to sizeable credit losses to private credit investors”.

The FSB said: “This could be triggered by any significant shortfall in the supply of electricity, a critical factor in the construction and operation of datacentres, which could lead to delays or cancellations of projects.”

This is already happening. Projects already are being delayed and/or cancelled.

Private credit – should I care?

Yes. Private credit facilities mushroomed in the wake of tighter regulation of investment banking after the 2008 crash. It’s a sector that’s coming under increasing scrutiny, as investors get more jittery about what amounts to unregulated lending. Lenders like Blue Owl and Apollo Investment Management have both had to limit investor withdrawals after surges in requests in the last few months.

To some observers, this is all alarmingly reminiscent of the run up to the 2008 financial crash. Others seem completely unbothered, presumably on the proviso that it will be different this time. Shares in Alphabet rose around 2% in extended trading yesterday.

Alarm bells are ringing, but the promises of what LLMs can deliver in the future – always in the future – mean that few seem to be listening.