Broadcom reported fiscal third-quarter 2026 revenue of $29.59 billion on Thursday, an 86% increase from $15.95 billion a year earlier, with AI semiconductor revenue of $16.7 billion growing 221% year over year and 54% sequentially. The results, detailed in the company's earnings release and summarized by Pulse 2.0, cap a quarter in which demand for Broadcom's custom AI accelerators and networking technology outpaced even the most aggressive Wall Street forecasts.
The custom-silicon boom is reshaping the economics of AI infrastructure, and this quarter's numbers show how quickly the center of gravity is shifting. For readers tracking the latest AI developments across chips, models and cloud economics, Broadcom's results are the clearest signal yet that hyperscaler spending on bespoke silicon has become a market of its own.
The Numbers Behind the Surge
According to the earnings release cited by Pulse 2.0, Broadcom's semiconductor solutions segment generated $20.84 billion in revenue, up 127% year over year. GAAP operating income rose 171% to $15.96 billion, while non-GAAP operating income climbed 92% to $20.10 billion. GAAP net income increased 216% to $13.09 billion, and non-GAAP net income rose 95% to $16.37 billion.
Earnings per share followed the same trajectory: GAAP diluted EPS of $2.68, up 215%, and non-GAAP diluted EPS of $3.32, up 96%. Cash flow from operations reached $14.2 billion, and free cash flow grew 95% to a record $13.7 billion — 46% of total revenue.
Hock Tan Guides Even Higher
The guidance is where the story gets more striking. Broadcom expects fiscal fourth-quarter revenue of approximately $34.8 billion, representing 93% year-over-year growth, with non-GAAP operating income holding at roughly 66% of projected revenue. AI semiconductor revenue is forecast to accelerate further to $21.7 billion, up 236% year over year.
"Demand for our custom AI accelerators and networking continues to be very strong," CEO Hock Tan said in the release. "Q3 AI semiconductor revenue of $16.7 billion grew 221% year-over-year, and 54% quarter-over-quarter. In Q4 the momentum continues, and we expect AI semiconductor revenue to accelerate to $21.7 billion, up 236% year-over-year."
CFO Amie Thuener emphasized the profitability of the growth: "Broadcom achieved record revenue, operating profit and free cash flow in Q3. We delivered non-GAAP operating income growth of 92% year-over-year, as consolidated revenue grew 86% year-over-year to $29.6 billion."
The company also declared a quarterly dividend of $0.65 per share, which several outlets, including ad-hoc-news, flagged alongside the record results.
Why Custom Accelerators Matter Now
Broadcom's AI revenue is built largely on custom accelerator programs for hyperscale customers — chips designed for specific companies' workloads rather than sold as general-purpose GPUs. Multiple reports on the results, including coverage from Martin Cid Magazine and Tech Times, framed Broadcom's custom silicon as an increasingly credible alternative to Nvidia's merchant GPU business, which has dominated AI training and inference hardware for two years.
The strategic logic for hyperscalers is straightforward. Custom accelerators can be tuned for specific model architectures and inference patterns, potentially improving cost-per-token economics at data center scale. As AI inference workloads grow — particularly for the agentic systems now common across the industry — the economics of specialized silicon become more attractive relative to general-purpose GPUs.
Broadcom's networking portfolio compounds the advantage. AI clusters depend on high-speed interconnects to move data between thousands of accelerators, and Broadcom supplies much of that plumbing. When a hyperscaler buys a custom accelerator program from Broadcom, the networking gear often follows.
What It Means for the AI Hardware Market
The quarter reinforces a bifurcation in AI hardware. Nvidia remains the default choice for frontier training runs, but the inference market — where most deployed AI workload ultimately lives — is where custom silicon is gaining fastest. Broadcom's 221% AI growth rate against a much larger base than a year ago suggests hyperscalers are moving real production workloads onto custom chips, not merely running pilots.
For competitors, the bar keeps rising. AMD has been pushing its Instinct line, Intel is reorganizing around foundry and AI, and a wave of startups is targeting specific niches of the accelerator market. Broadcom's results show that the company's model of co-developing chips with a handful of very large customers can scale to tens of billions of dollars per quarter.
There are risks worth noting. Custom accelerator programs concentrate revenue among a small number of hyperscale customers, and any change in a major customer's capital plans would ripple directly through Broadcom's outlook. The company's Q4 guidance assumes the current spending pace continues into the end of the fiscal year.
Wall Street's Reaction
Coverage of the results was largely positive, with outlets including 24/7 Wall St. and The Motley Fool highlighting the 221% AI growth figure. Some coverage, such as a 36Kr analysis published Thursday, noted that the results — while described internally as exceeding expectations — landed against market projections that had grown extraordinarily ambitious, and shares moved modestly rather than soaring.
That dynamic reflects how completely AI expectations now permeate the semiconductor sector: a quarter with 86% revenue growth and a 221% surge in AI revenue is treated as strong-but-expected rather than extraordinary. Broadcom's ability to guide Q4 AI revenue up another 30% sequentially suggests the demand curve has not yet bent.
For now, Broadcom's results put hard numbers on a trend the industry has been discussing for two years. The custom AI silicon market is no longer a hedge against Nvidia — it is a multi-billion-dollar-quarter business of its own, and Broadcom is its primary beneficiary.
Stay Ahead of AI
Follow every earnings report, model launch and policy shift shaping the AI industry — read more AI news →
