Anthropic's rapid growth and massive new computing deal with Google and Broadcom provide powerful validation for the huge AI infrastructure spending by cloud providers.
This is significant because it confirms that the immense demand for AI is not just hype; it's a structural trend. AI companies like Anthropic are growing so fast that they need to secure enormous amounts of computing power years in advance. Anthropic's commitment to an additional 3.5 gigawatts of Google's Tensor Processing Units (TPUs) from 2027, on top of 1 gigawatt already secured for 2026, is a clear signal. This justifies the billions of dollars in capital expenditures (CAPEX) that companies like Google and Amazon are pouring into data centers and specialized chips.
Let's look at the causal chain. First, for Google, this deal is a direct win. The agreement with Anthropic translates into a long-term, predictable revenue stream for its cloud division. This comes right after Google Cloud's backlog—the amount of contracted future revenue—jumped by an incredible 55% in a single quarter to $240 billion, largely driven by big AI deals. The new 3.5GW commitment reinforces this trend, ensuring a robust revenue pipeline for years to come.
Second, this news also has positive implications for Amazon Web Services (AWS). Anthropic has repeatedly named AWS its primary cloud partner for training its AI models. With Anthropic's annualized recurring revenue (ARR) soaring from about $9 billion to $30 billion in just a few months, a substantial portion of that is spent on AWS infrastructure. This supports Amazon's aggressive $200 billion CAPEX guidance for 2026 and suggests that its revenue from AI workloads could exceed market expectations.
This event didn't happen in a vacuum. It's the culmination of a series of developments over the past 18 months, including major funding rounds for Anthropic and consistent reports of rising AI-driven demand from both Google and Amazon. What the 3.5GW deal changes is the timescale and magnitude of this demand. It transforms the narrative from a series of one-off capacity increases into a clear, multi-year expansion track, solidifying the long-term growth story for the entire AI infrastructure ecosystem.
- CAPEX (Capital Expenditures): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, plants, buildings, technology, or equipment.
- ARR (Annualized Recurring Revenue): A key metric used by subscription-based companies to represent the total revenue they expect to receive from customers over a year.
- TPU (Tensor Processing Unit): Google's custom-developed accelerator chip designed specifically for artificial intelligence and machine learning workloads.