Hock Tan's forecast of over $100 billion in AI chip revenue would represent a fivefold jump from 2025, powered by partnerships with OpenAI, Anthropic, and four other major customers.
Broadcom just put a number on the AI gold rush, and it’s measured in gigawatts. CEO Hock Tan confirmed that the chipmaker plans to ship roughly 10 gigawatts of AI compute capacity to six major customers by 2027, with even larger shipments expected in 2028 and 2029.
To translate that into dollars: Broadcom is projecting AI chip revenue exceeding $100 billion in 2027. That’s up from $20 billion in 2025.
The gigawatt economy takes shape #
The 10-gigawatt figure comes from Broadcom’s Q1 FY2026 earnings call, where Tan laid out a roadmap that reads less like a semiconductor forecast and more like a power grid expansion plan. The company reported $8.4 billion in AI semiconductor revenue for the quarter, representing 106% year-over-year growth.
The six major customers haven’t all been named publicly, but two of the most prominent partnerships tell a clear story. Broadcom’s collaboration with OpenAI, first announced on October 13, 2025, centers on developing and deploying custom-designed AI accelerators totaling 10 GW of capacity. The first shipment of those custom chips is targeted for 2027, with initial capacity exceeding 1 GW.
Anthropic, meanwhile, has its own aggressive ramp planned. The Claude-maker intends to deploy 1 GW of TPUs in 2026, scaling to 3 GW the following year. Between just these two customers, Broadcom already has a significant chunk of its 2027 target accounted for, with Google and Meta likely filling the rest of the pipeline.
The revenue math gets interesting fast #
Analysts estimate that each gigawatt of AI compute infrastructure generates between $12 billion and $20 billion in revenue. At 10 GW, that math lands somewhere between $120 billion and $200 billion in total addressable revenue, which makes Tan’s $100 billion-plus forecast look almost conservative depending on where Broadcom falls in the value chain.
But not everything is smooth sailing. Reports surfaced in May 2026 highlighting potential financing challenges for the OpenAI-Broadcom partnership, with obstacles estimated at around $18 billion. Building out 10 GW of custom AI accelerator capacity requires not just engineering talent but enormous capital commitments, and even the best-funded AI companies face questions about whether their spending plans pencil out.
What this means for investors #
Broadcom’s positioning in the AI chip market is fundamentally different from Nvidia’s. While Nvidia sells general-purpose GPUs that any customer can buy off the shelf, Broadcom is building custom silicon for specific hyperscalers.
That concentration risk cuts both ways. Having six major customers driving a $100 billion revenue forecast means each relationship is worth tens of billions of dollars. OpenAI alone represents a massive share of that 10 GW target.
The $18 billion financing challenge reported around the OpenAI partnership deserves scrutiny. AI companies are collectively spending hundreds of billions on infrastructure, but much of that spending depends on continued access to capital markets and sustained investor enthusiasm for AI.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our