Broadcom has told Wall Street that it expects its AI semiconductor revenue to reach approximately $230 billion by fiscal 2028 — a fourfold increase from the roughly $58 billion the company anticipates for the current fiscal year.

CEO Hock Tan delivered the forecast during Broadcom's fiscal Q3 2026 earnings call on September 2, alongside quarterly results that underscored how central AI has become to the company's business. The projection also highlights a broader shift in the AI chip industry: as hyperscalers pour capital into data center buildouts, an increasing share of that spending is flowing to custom silicon designed for their specific workloads rather than to merchant GPU suppliers alone.

The Numbers Behind the Ambition

Broadcom's Q3 FY2026 results set records on most measures. Total revenue came in at $29.6 billion, up 86% year-over-year. The AI semiconductor division, which includes custom accelerators and AI networking chips, generated $16.7 billion in the quarter alone — a 221% surge compared with the same period a year earlier.

For Q4, the company guided AI semiconductor revenue to $21.7 billion. If that figure is achieved, it would cap a fiscal year with roughly $58 billion in AI chip sales.

Tan projected that AI semiconductor revenue would roughly double to $115 billion in fiscal 2027, then double again to approximately $230 billion in fiscal 2028. He also estimated that adjusted earnings per share for fiscal 2028 would surpass $30.

Custom chips accounted for about 73% of Q3 AI revenue. Six key customers are driving this demand, including Google, OpenAI, Anthropic, and Meta, all of which Tan described as making multi-gigawatt-scale deployment commitments. Broadcom said it has secured supplies of cutting-edge wafers, high bandwidth memory, and substrates through 2028 — a signal that the company sees component availability, not just customer demand, as a gating factor for its growth plans and has moved to lock in capacity years in advance.

Wall Street's Mixed Reaction

Despite the blockbuster AI figures, Broadcom's overall Q3 revenue slightly missed analyst expectations, and the stock dipped in the aftermath. The divergence reflects a broader debate among investors about whether AI infrastructure spending can sustain its current pace, a question that directly shapes how analysts value semiconductor suppliers tied to hyperscaler capex.

Analyst projections for Broadcom's 2028 AI revenue show some dispersion. While Tan pegged the number at $230 billion, some analyst estimates land closer to $200 billion. RBC has projected that the total AI semiconductor market could exceed $550 billion by 2028. If that figure holds, Broadcom's $230 billion target would represent roughly 40% of the entire AI chip market.

The Nvidia Alternative Play

Broadcom's positioning in AI semiconductors is fundamentally different from Nvidia's. Whereas Nvidia sells standardized GPUs, Broadcom designs custom chips tailored to each hyperscaler's specific architecture. Google has its TPUs, Amazon has its Trainium chips, and Meta and OpenAI are investing in custom silicon. Broadcom sits at the center of many of these efforts, supplying design expertise and networking technology.

The company's AI networking business provides a dual revenue stream: as AI clusters scale, the networking fabric connecting the chips becomes as critical as the compute itself, and Broadcom's networking chips handle that interconnect layer. Whether the 2028 target is met will hinge largely on whether the named customers follow through on their multi-gigawatt deployment commitments and whether custom silicon continues to win share of AI data center budgets.