Anthropic Scales AI Infrastructure with Massive Google and Broadcom TPU Deal
Anthropic is aggressively expanding its compute capabilities to keep pace with an unprecedented surge in demand for its Claude models. In a strategic move announced on April 6, 2026, the AI company has signed a new agreement with Google and Broadcom to secure multiple gigawatts of next-generation Tensor Processing Unit (TPU) capacity, expected to come online starting in 2027.
This expansion isn’t just about raw power; it’s a calculated effort to diversify hardware dependencies and ensure the resilience of the infrastructure powering the frontier of AI development.
The $21 Billion Hardware Bet: Broadcom and TPU v7p
A central piece of this infrastructure push is a massive $21 billion commitment to Broadcom. This deal focuses on securing nearly 1 million Google TPU v7p units, with deliveries scheduled to arrive by late 2026.
The financial commitment arrived in two waves: $10 billion during Q3 of fiscal 2025, followed by an additional $11 billion in Q4. Notably, Broadcom is changing how it delivers this technology. Instead of providing loose components, Broadcom is shipping fully assembled “Ironwood Racks”—complete rack-level AI systems that Anthropic can deploy directly into its data centers.
Diversifying the Compute Stack
Anthropic is avoiding vendor lock-in by employing a multi-cloud and multi-hardware strategy. The company currently trains and runs Claude on a diverse range of AI hardware, including:
- Google TPUs: High-efficiency custom chips for AI workloads.
- AWS Trainium: Amazon’s custom AI training chips.
- NVIDIA GPUs: The industry standard for general-purpose AI acceleration.
By matching specific workloads to the chips best suited for them, Anthropic improves performance and ensures greater resilience for customers who rely on Claude for critical business operations.
Explosive Business Growth Driving Demand
The scale of this investment is a direct response to a massive spike in revenue and customer adoption. Anthropic’s run-rate revenue has now surpassed $30 billion, a staggering leap from approximately $9 billion at the end of 2025.
This growth is driven by high-value enterprise adoption. In February 2026, Anthropic reported over 500 business customers spending more than $1 million annually. As of April 2026, that number has doubled to exceed 1,000 customers.
Strengthening U.S. Computing Infrastructure
The majority of this new compute capacity will be located within the United States. This move expands on a commitment made in November 2025 to invest $50 billion in strengthening American computing infrastructure, signaling a long-term bet on domestic AI scaling.
Beyond the chips, Anthropic continues to deepen its integration with Google Cloud services. The company uses a suite of tools including BigQuery, Cloud Run, and AlloyDB to power its data and AI development. This ecosystem already supports thousands of customers accessing Claude models, including major firms like Shopify, Replit, Palo Alto Networks, Cursor, and Coinbase.
Key Takeaways: Anthropic’s Compute Expansion
| Metric/Detail | Information |
|---|---|
| Broadcom Deal Value | $21 Billion |
| Hardware Secured | Nearly 1 million Google TPU v7p units |
| Delivery Method | “Ironwood Racks” (Fully assembled systems) |
| Revenue Growth | From ~$9B (End of 2025) to >$30B (April 2026) |
| High-Value Customers | 1,000+ spending >$1M annualized |
| US Investment | $50 Billion commitment (started Nov 2025) |
The Road Ahead
With multiple gigawatts of TPU capacity slated to come online in 2027, Anthropic is positioning itself to handle the exponential growth of generative AI. By securing a massive pipeline of custom silicon and diversifying its hardware providers, the company is building a foundation capable of supporting the next generation of foundation models and AI agents.