Anthropic, the AI company behind the Claude chatbot, has hit a $30 billion annualized revenue run rate, according to Bloomberg. If accurate, that figure would place Anthropic ahead of OpenAI, which was last publicly reported to be targeting roughly $12.7 billion in annualized revenue in early 2025. Anthropic has also locked in a multi-year deal for access to Google’s custom AI chips, a partnership now confirmed in a federal securities filing by Broadcom. The company is reportedly exploring a valuation near $900 billion.
Two years ago, Anthropic was a well-funded but relatively niche safety-focused lab. Now it is generating more revenue than most Fortune 500 companies and locking up enough electrical power to run a small city. The speed of that transformation says as much about the AI market’s appetite as it does about Anthropic itself.
The chip deal, confirmed in an SEC filing
The hardest facts in this story come from Broadcom’s Form 8-K, filed with the Securities and Exchange Commission. The filing establishes three things. First, Broadcom has a long-term agreement with Google covering the production and supply of Tensor Processing Units (TPUs), the custom chips Google designs for AI workloads. Second, that agreement includes supply assurance extending through 2031. Third, the deal has expanded into a three-party collaboration among Broadcom, Google, and Anthropic, under which Anthropic will access approximately 3.5 gigawatts of computing capacity beginning in 2027.
To put 3.5 gigawatts in perspective: a single large-scale data center typically draws between 50 and 100 megawatts. Anthropic’s planned allocation could power 35 to 70 such facilities. That is an enormous commitment of electricity and hardware, reflecting the raw physical demands of training and running frontier AI models at scale.
The structure of the deal is unusual. Most AI startups rent Nvidia GPUs through cloud providers like Amazon Web Services or Microsoft Azure. Anthropic is instead building its compute stack around Google’s TPUs, chips purpose-built for the matrix math at the heart of modern AI. Google provides the chip architecture and data center capacity. Broadcom manufactures the TPU hardware. Anthropic deploys the compute to train and serve Claude.
Because the filing is a legally mandated disclosure, subject to securities-law penalties for material misstatement, the basic parameters of the deal carry high confidence. The duration, the supply assurances, and the 3.5-gigawatt commitment are not marketing claims. They are on the public record.
The revenue numbers and what they actually show
The financial picture is less certain. Bloomberg reported that Anthropic has reached a $30 billion annualized revenue run rate, up from approximately $9 billion at the end of 2025. Anthropic is a private company and does not publish audited financial statements, so neither figure can be independently verified through public filings.
If both numbers are accurate, the growth from late 2025 to mid-2026 represents roughly a 3.3x increase, extraordinary by any standard but not the 30x figure referenced in the headline. The larger multiple likely traces back to Anthropic’s run rate earlier in 2025, which was reported to be closer to $1 billion. From that baseline, a jump to $30 billion over 15 months would indeed approach 30x. The discrepancy reflects different starting points and the difference between annualized run rate, trailing twelve-month revenue, and bookings. Without Anthropic’s own disclosure, the precise growth multiple is impossible to pin down.
The claim that Anthropic has “passed OpenAI in revenue” adds another layer of complexity. OpenAI, also privately held, has not disclosed a current run rate through any regulatory filing. The most recent widely reported figure, from early 2025, put OpenAI at roughly $12.7 billion in annualized revenue. OpenAI has almost certainly grown since then, but by how much is unknown. The comparison is directional: Anthropic appears to be in the same revenue tier as OpenAI, and may have overtaken it, but a definitive ranking would require audited numbers from both companies.
There is also a question of durability. Annualized run rates extrapolate a recent period’s revenue across a full year. If Anthropic’s current numbers are driven by a handful of large enterprise contracts or a surge in API usage that may not persist, the annualized figure could overstate the company’s steady-state revenue. None of these details are visible in the limited public evidence.
The $900 billion valuation
Bloomberg’s reporting references a potential $900 billion valuation in the context of ongoing funding discussions. No term sheet, investor letter, or regulatory filing confirms that Anthropic has received or accepted a valuation at that level. Private-market valuations are often preliminary and can shift substantially before a round closes, especially in a sector as volatile as generative AI.
For context, Anthropic’s last confirmed valuation was $61.5 billion, set during a funding round in early 2025. A jump to $900 billion would represent roughly a 15x increase in just over a year. That kind of leap is not unprecedented in tech (OpenAI’s own valuation surged from $80 billion to $300 billion in a matter of months during 2024), but it would make Anthropic one of the most valuable private companies in history, surpassing SpaceX and rivaling the market capitalization of public giants like Walmart or JPMorgan Chase.
Until a completed round is disclosed, the $900 billion figure is best understood as a signal of investor appetite, not an established market price.
Who is backing Anthropic and who is using Claude
Anthropic’s rise has been fueled by some of the largest corporate investments in AI history. Amazon has committed up to $8 billion in the company. Google has invested at least $2 billion. Salesforce, Spark Capital, and other institutional investors have also participated in multiple rounds. That backing has given Anthropic the capital to compete for talent, secure compute, and build out its product line without the pressure of an IPO.
On the product side, Claude has gained traction across both consumer and enterprise markets. Anthropic offers a free tier, a $20-per-month Pro plan, and enterprise API access that has attracted companies in finance, legal, healthcare, and software development. The company has also expanded Claude’s capabilities with features like computer use, extended thinking, and a 200,000-token context window, moves designed to make the model useful for complex, multi-step tasks rather than just chat.
Revenue growth at this scale suggests that enterprise adoption, not just consumer subscriptions, is driving the numbers. Large API contracts with companies building on top of Claude could account for a significant share of the $30 billion run rate, though Anthropic has not broken out its revenue by segment.
What this means for the AI industry’s power structure
The strategic implications of the Broadcom deal may matter more than the revenue headline. By locking in TPU supply through 2031, Anthropic insulates itself from the GPU shortages and pricing swings that have constrained competitors. It also deepens Anthropic’s dependence on Google’s chip ecosystem, a relationship that carries its own risks if Google’s priorities shift or if TPU performance falls behind Nvidia’s roadmap.
For Google, the deal validates its years-long investment in custom silicon. TPUs have been central to Google’s internal AI work since 2015, but landing Anthropic as a major external customer signals that the chips can compete for workloads beyond Google’s own products.
For policymakers, the 3.5-gigawatt figure is a stark reminder of AI’s physical footprint. Training and serving increasingly capable models requires not just capital and talent but vast amounts of electricity and specialized hardware. A small number of companies are now securing long-term control over critical compute resources, a concentration that could shape who gets to build the next generation of models and on what terms.
The broader picture is one of rapid consolidation. Anthropic, OpenAI, Google DeepMind, xAI, and Meta are all racing to lock up chips, energy, and talent. The gap between these frontier labs and everyone else is widening. Anthropic’s chip deal and reported revenue suggest it has firmly established itself in that top tier, but the AI market is moving fast enough that today’s leader can become tomorrow’s cautionary tale. What is documented, the Broadcom filing, the multi-year supply commitment, the 3.5-gigawatt allocation, is real. What is projected, the $30 billion run rate, the $900 billion valuation, the claim of passing OpenAI, still depends on numbers that no one outside Anthropic’s boardroom can fully verify.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
