Detailed Analysis
Anthropic's annualized revenue run rate surpassed $30 billion in early April 2026, marking a more than threefold increase from the approximately $9 billion the company recorded at the close of 2025. The milestone was disclosed alongside a significant compute partnership involving Google and Broadcom, under which Anthropic will gain access to roughly 3.5 gigawatts of TPU-based computing capacity beginning in 2027. The deal, detailed in a Broadcom regulatory filing, encompasses a long-term agreement to develop and supply custom TPUs for future Google generations as well as networking components through 2031, with Anthropic positioned as a primary consumer of that capacity. Anthropic CFO Krishna Rao characterized the arrangement as the company's "most significant compute commitment to date," framing it as a direct response to the pace of commercial expansion.
The revenue trajectory underlying this announcement is striking in its velocity. Anthropic entered 2025 at roughly $1 billion in annualized run rate, reached $4.5 billion by mid-year, climbed to $9 billion by year-end, crossed $14 billion in February 2026 at the time of its Series G funding announcement, and has now more than doubled that figure within two months. A core driver of this growth is the company's deliberately enterprise-heavy commercial model: approximately 80% of revenue originates from business customers, and Anthropic now counts over 1,000 enterprises each spending more than $1 million annually — double the figure reported in February. That customer profile tends to produce higher retention rates and lower churn than consumer-oriented strategies, creating a compounding revenue base that has allowed Anthropic to sustain what appears to be near-exponential growth.
The compute infrastructure underpinning this growth reflects Anthropic's strategy of diversifying across multiple hardware ecosystems rather than concentrating dependence on a single supplier. The new Broadcom-Google arrangement builds upon the company's November 2025 commitment to $50 billion in US AI infrastructure, which already spans AWS Trainium chips, Google TPUs, and Nvidia GPUs. By securing 3.5 gigawatts of dedicated TPU capacity starting in 2027, Anthropic is effectively pre-positioning its compute supply chain to match projected demand curves — a critical consideration given that training and inference costs remain one of the primary constraints on frontier AI model development. Broadcom's regulatory filing notably acknowledged risks tied to Anthropic's ongoing commercial success, underscoring that this is a long-duration bet on the company's continued growth rather than a purely transactional arrangement.
The broader competitive context amplifies the significance of these figures. Anthropic's $30 billion run rate reportedly places it ahead of OpenAI's estimated $24–25 billion run rate, a reversal of the longstanding assumption that OpenAI held a commanding commercial lead. Broadcom CEO Hock Tan has separately predicted that his company will generate over $100 billion in AI chip revenue by 2027, driven by hyperscalers and frontier AI labs that lack the in-house silicon expertise to build custom accelerators independently — a category in which Anthropic firmly sits. The convergence of Anthropic's enterprise revenue momentum, its multi-supplier compute strategy, and Broadcom's emergence as a major custom AI silicon provider reflects a structural shift in how frontier AI companies are organizing supply chains, customer acquisition, and long-horizon infrastructure investment simultaneously, rather than sequentially.
Read original article →