Anthropic Hits $30B Revenue, Inks $40B Google Deal, The Numbers Behind the AI Arms Race

Anthropic's numbers are staggering. The company behind Claude has seen its annualized revenue jump from roughly $9 billion to over $30 billion, 80 times internal projections, while executing a series of deals that signal the AI infrastructure race is only accelerating.

For those who haven't been following Anthropic closely, here is what you need to know. Anthropic was founded in 2021 by Dario and Daniela Amodei, two former OpenAI executives who left over disagreements about AI safety. Their thesis was simple: build AI systems that are genuinely safe and aligned with human values, not just powerful. Claude, their flagship model, has become the main credible alternative to OpenAI's GPT series. Claude is known for being more thoughtful, more careful in its reasoning, and generally more reluctant to generate harmful content. This "constitutional AI" approach has attracted customers who find OpenAI's breakneck pace unsettling.

Google's investment here is not just a financial bet. It is strategic. Google has its own AI models, Gemini among them, but the company has struggled to translate research dominance into the kind of developer mindshare and enterprise adoption that OpenAI and Anthropic have captured. By backing Anthropic with $40 billion, Google gets a seat at the table in the frontier model race without putting all its eggs in the Gemini basket. It is a hedge, and a massive one at that.

The OpenAI comparison is instructive. OpenAI has Microsoft. Anthropic has Google and Amazon (AWS invested $8 billion in 2024 and 2025). The AI industry is consolidating around two poles, each backed by cloud providers who want to sell the compute that powers these models. For users, this means Claude and ChatGPT are likely to remain the two dominant consumer-facing AI products for the foreseeable future. Competition between them is intense, and that intensity is good for anyone who uses these tools.

Three deals in eight days

In the span of just over a week, Anthropic closed three major agreements:

• A $40 billion additional investment from Google at a $350 billion valuation
• A lease on SpaceX's Colossus 1 facility, over 300 megawatts of power and 220,000 NVIDIA GPUs
• A $1.8 billion CDN deal with Akamai, the largest in the company's 28-year history

The SpaceX deal deserves a closer look. Colossus 1 is not a typical data center. It is a facility purpose-built for training the largest AI models ever constructed. 300 megawatts is roughly the power consumption of a small city. 220,000 GPUs is a number that would have sounded absurd even three years ago. Anthropic is effectively locking in the physical infrastructure needed to train Claude's next several generations. These deals cannot be done quickly. The lead times on power contracts, GPU procurement, and facility construction run into years. Anthropic is placing bets on 2027 and 2028 right now.

Anthropic also launched a $1.5 billion enterprise AI services joint venture with Blackstone, Goldman Sachs, and Hellman & Friedman, bringing Wall Street directly into the AI services business. This is the part I find most interesting. It signals that Anthropic sees enterprise services, helping large companies deploy and customize Claude, as a revenue stream that could rival the consumer subscription business. Goldman and Blackstone do not throw money at things without seeing a clear path to returns. Their involvement suggests enterprise AI spending is very real and very large.

OpenAI's response

OpenAI countered with its own moves: a $10 billion "DeployCo" joint venture with TPG and Bain, featuring a guaranteed 17.5% annual return to private equity backers. The company also closed a $122 billion funding round at an $852 billion valuation, the largest private financing round in history.

That guaranteed 17.5% return is worth paying attention to. Private equity firms do not typically get guaranteed returns in venture deals. The fact that OpenAI agreed to it suggests the company needed the capital badly enough to accept terms that would be unthinkable for a normal software company. Training frontier models is astronomically expensive, and even the market leader needs creative financing to keep the lights on.

OpenAI's compute spending tells the story in numbers: $30 million in 2017. A projected $500 billion in 2026.

What this means for Claude users

If you use Claude regularly, this wave of investment translates into a few concrete things. First, Claude is not going anywhere. A company with $40 billion in fresh Google investment and 220,000 GPUs under contract is not running out of money or compute power. Second, expect Claude to improve rapidly. The SpaceX compute deal means the next generation of models will be trained on infrastructure that dwarfs what produced Claude Opus 4.7. Third, enterprise features, longer context windows, better tool use, deeper integrations with business software, are likely coming faster now that Anthropic has Wall Street partners explicitly focused on enterprise deployment.

There is a tradeoff worth acknowledging. As Anthropic grows from a research-focused safety lab into a $350 billion company with private equity partners, the company's original safety-first identity will be tested. I think the Amodeis genuinely care about AI safety in a way that is not just PR. But when you have Goldman Sachs and Google expecting returns on tens of billions of dollars, the pressure to ship faster, to cut corners on safety testing, to prioritize capabilities over caution, becomes real. Claude users should watch this tension closely.

The big question

Big Tech's total AI infrastructure spending could hit $725 billion this year. That is real money, committed to physical assets, not R&D experiments. The railroad analogy is getting harder to avoid: build first, figure out revenue later. The question is not whether the technology is real. It is whether the economics work before the money runs out. For now, the checks are still clearing, and the GPUs are still spinning up.