Broadcom shares jumped roughly 3% after securing multi-year AI chip contracts expected to significantly boost its AI revenue stream. The company signed a deal to supply Anthropic with 3.5 gigawatts of Google-powered AI computing capacity starting in 2027, guaranteeing the AI firm close to 5GW of computing power. Simultaneously, Broadcom locked in a long-term agreement with Google to build and deliver custom AI chips and hardware for future AI racks until 2031, extending a strategic partnership critical as hyperscalers ramp AI infrastructure investments.
Anthropic’s deal with Broadcom could cost it hundreds of billions of dollars. For context, setting up a single gigawatt of power costs about $35 billion-$50 billion, with chips making up the lion’s share of the expense.
Broadcom nonetheless believes that acknowledges that Anthropic’s massive appetite for AI compute is tied directly to the startup’s ongoing commercial wins. That said, the San Francisco startup’s revenue has more than tripled, climbing from $9 billion in December 2025 to $30 billion by the close of March.
“We are building the capacity necessary to serve the exponential growth we have seen in our customer base while also enabling Claude to define the frontier of AI development,” noted Krishna Rao, Anthropic’s chief financial officer.
Meanwhile, Broadcom’s deal with Google underscores Google’s ambition to make its internal chips a legitimate alternative to Nvidia’s GPUs. In recent years, TPUs have become a key part of Google’s cloud revenue.
Speaking on the new AI contracts, D.A. Davidson analyst Gil Luria noted that they “put the spotlight back on Broadcom as a major winner.” One Reddit user even suggested Broadcom’s share price could double in just three years.
Overall, AI leaders are still casting a wide net to maximize compute access. For instance, Anthropic claimed it is running its Claude network on several AI hardware platforms, including AWS Trainium, Google TPUs, and Nvidia GPUs. It contended that using different chips enhances performance and resilience.
OpenAI has also been seeking alternatives to Nvidia’s GPU since last year, according to sources. For some time, Nvidia held a monopoly on providing chips for the company. That changed in September, when they announced a $100 billion deal with OpenAI, in which they would trade an equity stake for the capital OpenAI needs to purchase advanced hardware. The deal was originally expected to close in weeks, but talks stretched into months, and OpenAI reached separate agreements with AMD.
Sources had claimed that OpenAI wasn’t happy with Nvidia’s speed on specific tasks, such as coding. They plan to fill the gap with new hardware that would meet about 10% of their total inference requirements. However, Nvidia CEO Jensen Huang brushed aside talk of tension, labeling such reports “nonsense” while emphasizing that Nvidia is moving forward with a huge financial commitment to OpenAI.
“Customers continue to choose NVIDIA for inference because we deliver the best performance and total cost of ownership at scale,” Nvidia stated. In a post on X, OpenAI Chief Executive Sam Altman also described Nvidia as the top AI chipmaker and said the firm wants to continue working closely with it for the foreseeable future.
Still, as the battle for AI inference heats up, firms like Broadcom are well-positioned to cash in on the growing demand.
There’s a middle ground between leaving money in the bank and rolling the dice in crypto. Start with this free video on decentralized finance.
Log in to Reply
Log in to comment your thoughtsComments