Broadcom published results for the third quarter of fiscal 2026, which ended on 2 August. Revenue rose 86%, GAAP net income 216%, and the company guides to 93% growth next quarter.
Those are not numbers one sees at this scale. But they are a result rather than a cause, and the cause sits in a single line the chief executive chose to open with.
Errors or inaccuracies are possible. Spotted something that looks wrong? Write to me and I will correct it.
The Overall Picture
| Third quarter | 2025 | 2026 | |
|---|---|---|---|
| Revenue | 15,952 | 29,591 | +86% |
| GAAP operating income | 5,887 | 15,955 | +171% |
| GAAP net income | 4,140 | 13,088 | +216% |
| Diluted earnings per share | 0.85 | 2.68 | +215% |
| Cash flow from operations | 7,166 | 14,197 | +98% |
| Free cash flow | 7,024 | 13,665 | +95% |
(in millions of dollars, except per-share figures)
Free cash flow is 46% of revenue. That is a level normally associated with software companies rather than chip makers, and it follows from Broadcom designing and selling but not manufacturing.
And the Number the Report Rests On
Hock Tan, the company's president and chief executive, opened the release like this:
"Demand for our custom AI accelerators and networking continues to be very strong. Q3 AI semiconductor revenue of $16.7 billion grew 221% year-over-year, and 54% quarter-over-quarter."
And he added a forecast: "In Q4 the momentum continues, and we expect AI semiconductor revenue to accelerate to $21.7 billion, up 236% year-over-year."
Note the quarterly figure, not the annual one. Growth of 221% over a year is impressive but measures a low starting point. Growth of 54% in a single quarter says something else entirely: the pace is not moderating.
How Much of the Company That Actually Is
This is the question that turns the number from impressive into meaningful.
| Third quarter | |
|---|---|
| Total revenue | 29,591 |
| Of which, the semiconductor segment | 20,839 |
| Of which, AI chips | 16,700 |
AI chips are about 56% of the entire company's revenue, and about 80% of the semiconductor segment.
And in next quarter's guidance the gap widens further: $21.7 billion out of $34.8 billion, or about 62% of revenue.
And What Happened to the Other Half of the Company
Broadcom is built from two segments, and it is worth seeing what happened to each.
| Segment | 2025 | Share | 2026 | Share | |
|---|---|---|---|---|---|
| Semiconductors | 9,166 | 57% | 20,839 | 70% | +127% |
| Infrastructure software | 6,786 | 43% | 8,752 | 30% | +29% |
Infrastructure software grew 29%. That is respectable growth in its own right - and this is a company that bought VMware in an enormous deal precisely to build that segment.
But it fell from 43% of revenue to 30%, not because it shrank but because the other half more than doubled.
Broadcom was a company that was half software. Within a year it became a company that is three-quarters chips, most of them AI chips.
The Guidance
| Fourth quarter, ending 1 November | |
|---|---|
| Revenue | about $34.8 billion, growth of 93% |
| Of which AI chips | $21.7 billion, growth of 236% |
| Adjusted operating margin | about 66% of revenue |
Chief financial officer Amie Thuener noted that the adjusted operating margin is expected to hold at 66% - unchanged from a year ago.
And that is worth pausing on. A company that nearly doubles revenue in a year usually pays for it in margin, because the growth comes from newer products or less profitable customers. Here revenue nearly doubled and the margin stayed put.
What I Suggest Watching
The quarterly figure, not the annual one. Guided annual growth of 236% sounds more dramatic than 221%, but both are measured against a low base. The move from $16.7 to $21.7 billion is growth of about 30% in a single quarter - and that is the number that says whether the pace is holding.
And customer concentration. Broadcom sells custom accelerators, that is, chips designed for a specific customer. Such a model creates deep and long relationships, and it also means a small number of customers account for a large share of revenue. The release does not break out customers, so I cannot quantify it - which is precisely why I note this as an open question rather than a figure.
הזווית שלי
דעה אישית של אילן אברמוב - לא ייעוץ ולא המלצה
What I take from this report is not the size of the numbers but the direction of the second derivative.
In every technology cycle a moment arrives when annual growth still looks excellent while quarterly growth has already begun to moderate. That is the early sign, and it appears long before the headlines change. Broadcom reported the opposite: 54% in the quarter, with guidance implying about another 30% in the next one.
And what that says about the wider debate: the thesis that AI spending is cooling is not supported by this report. Nor is it refuted by it - Broadcom sells mainly to a handful of large cloud operators, and one supplier's results do not measure the whole market. But anyone arguing for a slowdown has to explain this number rather than route around it.
And the second thing, which interests me more over the long run: Broadcom changed materially within a single year. A company that was half infrastructure software - a stable, contracted business with high margins and slow growth - is today mostly a business of custom chips for a small number of customers. Those two businesses behave very differently in a downturn, and anyone who bought the first has gradually received the second.
This is not a criticism of management - it went where the demand went, which is exactly what managers are supposed to do. It is a statement about what is actually held, and that is worth re-examining when a company changes its revenue mix at this pace.
(It is important to stress: this is my personal opinion only, and nothing herein constitutes a recommendation to take any action.)
