In one of the biggest infrastructure moves of the AI era, Broadcom has locked in a long-term agreement to build Google’s next-generation TPU chips — and expanded access to those chips for Anthropic, whose revenue has tripled in just four months.
Broadcom · Google TPU · Anthropic AI Infrastructure
Broadcom is the primary manufacturer of Google’s custom Tensor Processing Units (TPUs). | Illustration: WebPandits
In a move that underscores the staggering scale of today’s AI infrastructure race, semiconductor giant Broadcom confirmed on Monday that it has agreed to a long-term partnership with Google to develop and supply the next generation of custom Tensor Processing Units (TPUs) — and simultaneously announced an expanded compute access agreement with AI startup Anthropic that will give it access to 3.5 gigawatts of processing power starting in 2027.
The dual announcement, disclosed in a regulatory filing with the U.S. Securities and Exchange Commission (SEC), sent Broadcom’s stock up roughly 3% in after-hours trading and cemented the company’s position as the invisible backbone of the AI chip economy.
What Was Announced?
Broadcom revealed two distinct but interlinked agreements. The first is a Long Term Agreement with Google for Broadcom to design and supply custom TPUs for Google’s future chip generations — extending a manufacturing partnership in place since 2016. The deal also includes a supply assurance agreement for networking and other components to power Google’s next-generation AI data center racks through to 2031.
The second agreement involves all three companies — Broadcom, Google, and Anthropic — in an expanded strategic collaboration. Under its terms, Anthropic will access approximately 3.5 gigawatts of next-generation TPU-based AI compute capacity via Broadcom, beginning in 2027.
⚡ Key Facts at a Glance
▸ Broadcom signs long-term deal with Google to build next-gen TPU chips through 2031
▸ Anthropic gets access to ~3.5 gigawatts of TPU compute capacity starting 2027
▸ Broadcom also supplies networking components for Google’s next-gen AI racks
▸ Anthropic’s revenue run rate surged from $9B (end-2025) to $30B+ (April 2026)
▸ Enterprise customers spending $1M+/year on Anthropic: now over 1,000 — up from 500 in February
▸ AVGO shares rose ~3% in after-hours trading following the announcement
Anthropic’s Explosive Growth: Revenue Triples in Four Months
Perhaps the most striking number in Monday’s announcement wasn’t about chips at all — it was about revenue. Anthropic revealed that its annualized revenue run rate has now surpassed $30 billion, up from approximately $9 billion at the end of 2025 — a more than three-fold increase in just four months.
The number of enterprise customers spending more than $1 million annually on its AI tools has also more than doubled — from around 500 at the end of February to over 1,000 today — signalling that Claude AI has moved well beyond experimental use into core business workflows.
“Our run-rate revenue has now surpassed $30 billion — up from approximately $9 billion at the end of 2025. We are making our most significant compute commitment to date to keep pace with our unprecedented growth.”
— Anthropic spokesperson, April 7, 2026Why This Deal Matters for Big Tech
Google’s TPUs — Tensor Processing Units — are the primary alternative to Nvidia’s market-leading GPUs for AI workloads. While Nvidia dominates with its H100 and B200 series, Google’s TPUs have grown increasingly attractive as companies seek more cost-efficient alternatives. Broadcom has been the behind-the-scenes design partner for Google’s TPUs since 2016. By locking in a deal to develop future generations, Broadcom is positioning itself at the centre of Google’s AI roadmap for years to come.
Broadcom’s Growing AI Empire
The deals don’t exist in isolation. Broadcom is rapidly becoming the indispensable chip partner for multiple AI giants simultaneously. In 2025, it also signed a partnership with OpenAI to develop custom accelerators. CEO Hock Tan has predicted his company could win over $100 billion in AI chip revenue in 2027 alone. Mizuho analysts estimate Broadcom could collect as much as $21 billion in AI revenue from Anthropic in 2026, rising to $42 billion in 2027.
📈 Related Stocks · After Hours · April 7, 2026
Anthropic’s Multi-Cloud Strategy
Despite the scale of this deal, Anthropic has been careful to stress it is not becoming exclusively dependent on any single hardware supplier. The company continues to use a mix of Google TPUs, Amazon Web Services’ Trainium chips, and Nvidia GPUs — deliberately distributing workloads to match tasks to the most efficient available hardware.
The Geopolitical Dimension
The agreement also lands against a backdrop of intense geopolitical significance. AI infrastructure spending is no longer just corporate strategy — it has become national industrial policy. Anthropic noted that its expanded compute commitment aligns with a broader pledge to invest $50 billion in U.S. computing infrastructure. For India’s growing tech and startup ecosystem, the message is clear: the foundation layer of global AI — the chips, the racks, the power — is being locked up by a small number of vertically integrated Western players.
Bottom Line
Monday’s announcements mark a new chapter in the industrialisation of AI. Custom silicon is no longer a curiosity — it is the beating heart of frontier AI infrastructure. Broadcom is now the indispensable manufacturer for at least three of the world’s most powerful AI companies simultaneously. Anthropic, the company behind Claude, has quietly become one of the fastest-growing technology businesses in history.
The AI infrastructure race is no longer about who has the best model. It is increasingly about who has locked in the compute to run it.
“This groundbreaking partnership with Google and Broadcom is a continuation of our disciplined approach to scaling infrastructure — we are building the capacity necessary to serve the exponential growth we have seen in our customer base.”
— Anthropic, official blog post, April 7, 2026