Anthropic Secures Major Compute Deal with Google and Broadcom
Anthropic partners with Google and Broadcom for multi-gigawatt TPU capacity to support AI model demand, valued at $380 billion.
Anthropic Secures Major Compute Deal with Google and Broadcom
Anthropic, the AI safety-focused startup behind the Claude model family, has announced a significant expansion of its partnerships with Google and Broadcom. This deal, revealed on April 6, 2026, involves securing multiple gigawatts of next-generation TPU (Tensor Processing Unit) capacity, set to become operational starting in 2027. The agreement is described as the company's "most significant compute commitment to date," aiming to support the growing demand for its frontier AI models amid a $30 billion annual run-rate revenue and a recent $30 billion Series G funding round valuing Anthropic at $380 billion (Anthropic).
Strategic Infrastructure and Partnerships
Anthropic's CFO, Krishna Rao, emphasized the partnership's role in scaling infrastructure to match exponential customer growth. The compute will be delivered via Google Cloud services and Google-built TPUs supplied through Broadcom, diversifying Anthropic's hardware stack, which includes AWS Trainium, Google TPUs, and NVIDIA GPUs. This multi-platform strategy optimizes workloads for performance and resilience, with Amazon remaining Anthropic's primary cloud and training partner via Project Rainier (Google Cloud Press Corner).
Google Cloud highlighted thousands of enterprise customers accessing Claude through its platform, including Coinbase, Cursor, Palo Alto Networks, Replit, and Shopify. Beyond TPUs, Anthropic leverages Google tools like BigQuery, Cloud Run, and AlloyDB for data and AI development.
Past Performance and Track Record
Anthropic's compute strategy has evolved rapidly. In October 2025, it expanded Google TPU access to over one gigawatt, following initial deals that powered Claude 3.5 Sonnet and Opus models. Revenue has surged from enterprise adoption, with Claude available across AWS Bedrock, Google Vertex AI, and Microsoft Azure Foundry, a unique multi-cloud footprint (TechCrunch).
Competitor Comparison
| Company | Key Compute Partners | Recent Capacity Deals | Valuation/Revenue | Multi-Cloud Availability |
|---|---|---|---|---|
| Anthropic | AWS (primary), Google TPUs, NVIDIA, Broadcom | 3.5GW+ TPUs (2027), $50B U.S. invest | $380B / $30B run-rate | Yes (AWS, Google, Azure) |
| OpenAI | Microsoft Azure (exclusive), NVIDIA | Multi-$100B over years; Oracle supp. | ~$150B (est.) / $10B+ | No (Azure-primary) |
| xAI | NVIDIA, custom Dojo-like | $6B funding for Memphis supercluster | $24B / N/A | Limited |
| Google DeepMind | Internal TPUs | In-house scaling to 10%+ global power | N/A (Google: $2T) | Internal |
Anthropic's diversified stack contrasts with OpenAI's Microsoft reliance, offering resilience but higher integration costs.
Market Timing and Strategic Context
The timing aligns with AI compute shortages intensified by U.S. export controls on advanced chips to China, pushing firms toward domestic capacity. Google's 2024 global power use hit 3.7GW, framing this deal as a 10-15% equivalent expansion (News YCombinator).
Broader Implications
This positions Anthropic to challenge leaders in the $1 trillion AI race, enhancing Claude's edge in safe, reliable AI for critical applications. For Google, it validates TPUs against NVIDIA; for Broadcom, it cements AI silicon revenue. Risks include power grid strains and regulatory scrutiny, but the deal reinforces U.S. AI leadership (TechCrunch).
