Google spends 40 billion to secure Anthropic: The AI power struggle takes shape

Google will invest up to $40 billion in Anthropic, further deepening the partnership between the two companies. Both are partners and competitors in the artificial intelligence race.

Anthropic stated on Friday that Google committed to immediately invest $10 billion in cash at the latest valuation of $380 billion. Google will make an additional investment of $30 billion after Anthropic achieves performance milestones, while also significantly expanding Anthropic’s computing power capacity.

Anthropic is an important customer for Google’s chips and cloud services, and Google is actively expanding these businesses to compensate for the maturing growth of its core revenue source—search advertising. Google Cloud will provide 5 gigawatts of computing power to Anthropic over the next five years, starting gradually from 2027, with the possibility of additional gigawatts later. This agreement is a further expansion of the cooperation signed earlier this month between Anthropic, Google, and Broadcom.

Google’s TPU is one of the most competitive alternatives to NVIDIA chips. In this industry with extremely high demand for computing power, TPU is a scarce and valuable resource for Anthropic and other AI developers.

01 The End of the “Big Three” Pattern, Two Strong Rivals Forming

This investment marks a fundamental restructuring of the AI industry competition landscape.

Over the past two years, the top tier of AI has been defined as the “Big Three”—OpenAI, Google, and Anthropic—standing as a tripartite dominance. Now, this narrative has come to an end. Listing Anthropic’s funding rounds over the past half-year reveals a thought-provoking reality:

Amazon: $5 billion in cash, up to $25 billion, plus 5 gigawatts of Trainium capacity and a $1 trillion AWS procurement contract;

Google: $10 billion in cash, up to $40 billion, plus 5 gigawatts of TPU capacity;

NVIDIA: up to $10B, 1 gigawatt of GPU supply;

Microsoft: up to $5 billion, with Anthropic purchasing $30 billion worth of capacity on Azure.

All four top Silicon Valley players are now listed as shareholders of Anthropic. The total committed capacity exceeds 11 gigawatts—equivalent to the power output of 10 nuclear plants.

The landscape has shifted from “three companies dividing the world” to a duel between the Anthropic camp and OpenAI. The era of the “Big Three” is over.

02 Core Divide: Large Models + ASICs or Large Models + GPUs

Behind this restructuring, a clearer technical pathway division is emerging.

Anthropic is pursuing the “Large Models + ASIC” route. Google’s TPU and Amazon’s Trainium are both specialized chips (ASICs) designed specifically for AI workloads. Anthropic’s future computing foundation will mainly be supported by these two ASIC systems. One reason Google is willing to lock in Anthropic with $40 billion is this—Google’s capital expenditure plan for this year reaches $185 billion, heavily investing in data centers and TPU capacity. Without major clients to absorb the TPU inventory, it becomes the most expensive stockpile. Anthropic serves as both a conduit for enterprise clients and the best ballast for Google’s TPU capacity.

OpenAI, on the other hand, is following the “Large Models + GPU” route. Its core computing power comes from the Stargate project, tightly integrated with NVIDIA—a massive infrastructure plan valued at $500 billion, with the backbone based on NVIDIA GPUs. The advantage of this route is the mature NVIDIA GPU ecosystem and comprehensive software stack; however, the deployment cycle is long, with full production of Stargate expected around 2029, and the first Texas data center’s physical progress remains slow.

The underlying logic of these two routes is fundamentally different. ASICs offer high energy efficiency and lower unit cost but are highly customized, with less ecosystem flexibility; GPUs are versatile with a broader developer ecosystem but consume more power and cost more.

Currently, the competition is somewhat a proxy war between the ASIC camp and the GPU camp: Google and Amazon are validating and promoting their own ASIC chips through Anthropic, while NVIDIA consolidates its dominance in AI infrastructure via OpenAI and Stargate.

03 Google’s Strategy: Better to Leverage Than Fight

Google is willing to spend $40 billion on a nominal competitor, and three numbers explain why.

First, Anthropic’s annualized revenue as of March 2026 has surpassed $30 billion, from only about $1 billion at the start of 2025—a 30-fold increase in one year. Claude Code has gained popularity among programmers, enterprise channels are fully open, and B2B clients have expanded from startups to Fortune 500 companies.

Second, Anthropic’s implied valuation in the secondary market is close to $1 trillion, continuously surpassing Gemini’s products in developer market share within the AI enterprise market.

Third, Google, despite holding DeepMind, Gemini, and the world’s largest TPU cluster, has seen Gemini’s market share in enterprise AI continuously suppressed by Claude since its release over two years ago. Instead of fighting head-on, it’s better to leverage the strengths.

Google’s investment in Anthropic is a hedge operation of “if you can’t beat them, buy them”: if Anthropic wins in the AI enterprise market, Google can at least gain equity returns; if Gemini emerges, Google wins on both sides; if Gemini doesn’t succeed, TPU shipments are stable, and AI business can also be stabilized through Anthropic.

04 The Most Embarrassing Role: OpenAI

Anthropic’s computing power issues are resolved, but OpenAI’s own power capacity problems are becoming more apparent.

Within four days, Anthropic secured billion-dollar injections from Amazon and Google, locking in over 11 gigawatts of capacity commitments. OpenAI, after nearly a year of coordinating the Stargate project, has frequent rumors of funding gaps. SoftBank’s promised funds are still being disbursed in batches, and Microsoft is no longer the sole provider of OpenAI’s computing power.

Deeper issues are that OpenAI’s traditional advantages are being eroded. GPT series still lead in single-point performance, but Claude Code’s programming scene and Claude’s share in enterprise applications are expanding. Meanwhile, Microsoft, OpenAI’s core ally, has quietly appeared on Anthropic’s investor list—an extreme form of “hedging”: supporting OpenAI’s Azure while also providing capacity to Anthropic.

If Anthropic has bundled cloud giants (Amazon, Google), GPU giants (NVIDIA), and Microsoft as four strong backers, OpenAI still needs to quickly find a comparable strategic partner among sovereign capital and cloud giants—and do so fast.

Benchmark scores no longer determine the outcome of the AI industry. The real battleground of this reshuffle is a three-way game of cash + computing power + ecosystem. When Google invested $40 billion, the first round’s victory and defeat were already decided.