Anthropic fills hyperscaler backlogs

Cloud backlogs sound boring. But they’re becoming one of the clearest ways to see who really controls the AI buildout. The new twist is Anthropic. Its five-year, $200 billion Google Cloud commitment, reported on May 5, landed just days after Google said it would invest up to $40 billion in the company and expand capacity to 5 gigawatts. That turns Anthropic from “important customer” into something closer to a load-bearing tenant inside the cloud economy. ### What is a cloud backlog? A cloud backlog is basically contracted revenue that hasn’t been recognized yet — the money providers expect to collect over coming years from signed deals. Microsoft calls it commercial remaining performance obligation. Oracle calls it remaining performance obligations. Alphabet now reports backlog too, and said the majority is tied to normal Google Cloud Platform contracts, with just over half expected to turn into revenue within 24 months. (finance.yahoo.com) ### Why does Anthropic matter so much here? Because Anthropic is no longer just buying some extra compute. It is reserving industrial-scale capacity. Google and Anthropic said in April that Anthropic had secured 5 gigawatts of compute capacity that starts coming online next year, and Google’s investment can rise to $40 billion if milestones are met. Then came the May 5 report that Anthropic had committed to spend $200 billion with Google Cloud over five years. (microsoft.com) That single relationship is big enough to reshape how investors read Google’s backlog. ### How big is Google’s exposure? Very big. Alphabet’s reported backlog was about $467.6 billion in Q1 2026. Reuters’ pickup of the Anthropic report said the $200 billion commitment represents more than 40% of that figure. That does not mean Google books $200 billion tomorrow. But it does mean one AI lab may account for an unusually large share of future contracted cloud revenue. (cnbc.com) ### Is this just a Google story? No — Microsoft is even deeper into this pattern. Microsoft said its commercial remaining performance obligation reached $627 billion in the quarter ended March 31, 2026, up 99% year over year. Microsoft also explicitly noted that its OpenAI investment affected reported results, and the market already treats Azure’s AI growth and OpenAI demand as tightly linked. Oracle is in the same orbit after its own giant OpenAI infrastructure deals pushed RPO to $523 billion. (quartr.com) ### So why are people saying two labs fill half the pipeline? Because when you stack the disclosed and widely reported commitments together, OpenAI and Anthropic start to dominate the visible order book across Microsoft, Google, Oracle, and parts of Amazon’s AI capacity story. Some of the “half the backlog” charts making the rounds come from secondary analysis, so treat the exact percentages carefully. But the underlying direction is real — hyperscaler growth is becoming more concentrated around a tiny number of frontier-model buyers. (microsoft.com) ### Why is that a problem? Because concentration cuts both ways. These contracts make cloud revenue look huge and durable. But they also tie hyperscalers to a handful of customers that are expensive to serve, politically visible, and sometimes also competitors. Google sells cloud and TPUs to Anthropic while competing with Gemini. Microsoft powers OpenAI while also selling its own AI stack. If one lab slows spending, changes partners, or hits a financing wall, the “backlog” story suddenly looks less diversified than it did on earnings day. (finance.yahoo.com) ### Does this crowd out everyone else? In practice, yes — at least on the margin. Power, data-center shells, networking gear, and advanced chips are finite. A few multiyear giga-scale commitments can soak up the best capacity and force everyone else into longer waits or worse economics. That is why even Akamai landing a $1.8 billion Anthropic deal mattered last week — frontier labs are now spreading demand beyond the classic hyperscalers because the whole system is constrained. (cnbc.com) ### Bottom line? The story is not just that Anthropic signed a giant cloud deal. It’s that hyperscaler backlogs are starting to look less like broad enterprise demand and more like concentrated bets on a few AI labs. Anthropic didn’t create that shift by itself. But its Google deal made it impossible to ignore. (forbes.com)