
What happened
Broadcom CEO Hock Tan told Wall Street to expect its custom AI XPU, networking, and other AI chips to grow from $2.6 billion in F2022 to $230 billion by F2028. In Q3 F2026, AI-related revenue was $16.74 billion, up 3.22X year on year.
Why it matters
The forecast implies Broadcom is becoming Nvidia's biggest counterbalance. It has six XPU customers, around 30 gigawatts of compute capacity, and cost claims of $12 billion per gigawatt versus $18 billion for Nvidia Grace-Hopper and $40 billion for Vera-Rubin.
What to watch
The $230 billion target hinges on HBM memory, wafers, and packaging, which Broadcom says it has secured; if demand rises, Tan says it can add capacity. Watch Anthropic's planned deployments: 1 gigawatt of Ironwood TPU 7 this year, 5 gigawatts of TPU 8i in 2027, and 10 gigawatts of TPU 9 in 2028.
WHO IT HITSThis lands hardest on Broadcom's XPU customers such as Google, Anthropic, OpenAI, and Meta Platforms, who are deciding how much data-center capacity to commit per gigawatt. It also affects Nvidia and cloud builders planning AI clusters, because Broadcom's cost figure for custom accelerators is below Nvidia's systems if the body's $12 billion-per-gigawatt comparison holds.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
Broadcom's relationship with Google on custom TPUs stretches back seven years, including design help, TSMC foundry, and packaging. That base gave Broadcom a head start when the GenAI boom began, and the company now says the business has secured HBM memory, chip wafers, and packaging to fulfill its forecast. Tan told Wall Street not to expect quarterly updates to the forecast, a signal that he wants investors to focus on the multiyear trajectory rather than near-term swings.
The customer list behind that trajectory is unusually concentrated. Google, Anthropic, OpenAI, and Meta Platforms are named as customers, with the other two XPU customers possibly ByteDance or Apple. Anthropic is set to deploy 1 gigawatt of Ironwood TPU 7 this year and 5 gigawatts of TPU 8i in 2027, making it Broadcom's largest XPU customer in 2027 by Tan's account. OpenAI is on track for 1.3 gigawatts of Jalapeno clusters in 2027 and another 5 gigawatts in 2028, which would make OpenAI the second biggest and push Google to number three.
The outcome hinges on whether Broadcom can convert these gigawatt commitments into shipped systems, and whether Nvidia's grip on cloud customers keeps Google from buying more TPUs for itself. Broadcom's Q3 F2026 results already show the financial weight of this shift: semiconductor solutions sales were $20.84 billion, up 127.4 percent, and AI-related revenue hit $16.74 billion. If the $230 billion forecast holds, Broadcom's role as a chip shepherd for tech titans could look less like a supplier story and more like a structural counterweight to Nvidia, though Tan's numbers assume capacity additions that are not yet built.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Much of the attention on AI infrastructure buildouts is now tied to sheer compute power, with dominance define…

Barron's reported September 10 that Kepler Computing emerged from stealth with a memory architecture using fer…

Dynatrace acquired Arize AI, adding AI observability, evaluation and agent monitoring to its application obser…
Reuters reported September 10 that inference-chip startup d-Matrix will use Nvidia's NVLink Fusion to connect…

Amazon announced Shop the Scene, which lets U.S

Mecka AI, which collects human motion data to train robots, is nearing a round led by Sequoia Capital at a val…
