An inline quarter, and a roadmap to $230 billion
A record quarter that the market read as merely inline, followed across two sessions by a stock down over two per cent despite AI semiconductor revenue guided to double in FY27 and double again in FY28 — $58bn to $115bn to $230bn — with demand already ahead of that number. Buying the pullback with defined risk, on a name still down more than a quarter from its June high.
3 Sep 2026 · Q3 FY26 earnings and FY27/FY28 AI guidance · 3 min read
Sources
Summary
Q3 revenue was $29.6bn, up 86%, a record. AI semiconductor revenue was $16.7bn, up 221% year on year and above the company's own $16.0bn guide. None of that moved the stock — the quarter read as inline, not a beat, and the print itself is what sold off. What changed the picture was the call: management now guides AI semiconductor revenue to $115bn in FY27 and $230bn in FY28, against roughly $58bn this year, and said actual customer demand already exceeds the FY27 number. The stock is down across the two sessions since regardless, and sits more than a quarter below its June high.
| Metric | Reported | YoY |
|---|---|---|
| Revenue | $29.6bn | +86% |
| AI semiconductor revenue | $16.7bn | +221% |
| Non-GAAP operating margin | 68% | record |
| Non-GAAP EPS | $3.32 | record |
| Free cash flow | $13.7bn | 46% of revenue |
| FY27 AI semiconductor guide | $115bn | vs. ~$58bn FY26 |
| FY28 AI semiconductor guide | $230bn | vs. $115bn FY27 |
Positives and negatives
Positives
- The guide is a floor, not a forecast. Tan said demand already exceeds the $115bn FY27 number — the binding constraint is fab and advanced-packaging capacity, not orders. A guide the company is already behind on is a different kind of number than the usual best estimate.
- The customer base is broadening beyond one hyperscaler. Anthropic is on track to overtake Google as the largest XPU customer in 2027 and hold that through 2028, deploying 5GW of next-generation TPU capacity in 2027 with 10GW more in 2028. OpenAI's own deal is 10GW between the second half of 2026 and the end of 2029. Meta's roadmap is live from 2027. That is diversification a single-customer AI supplier does not have.
- Valuation has not caught up to the growth. Around 18x consensus FY27 earnings for a business guiding to double AI revenue two years running is a growth multiple well behind the growth rate.
Negatives
- The market has now read three different stories off the same set of facts in three sessions — sold the print, bought the call, sold again since. That is not a confident tape, and it means the near-term price is not obviously done digesting this.
- The guide is management's own number, unaudited and multiple years out. A $58bn to $230bn path assumes capacity gets built and every named customer's compute ramp proceeds on schedule — a lot has to go right in sequence for four years running.
- Infrastructure Software grew 29%, a third of Semiconductor Solutions' 127%. The AI story is carrying the entire re-rating; the other 30% of the business is merely fine.
Conclusion
With the market near highs, Broadcom is one of the few names that isn't — down roughly 27% from the June high despite a quarter that beat its own AI guide and a call that raised the multi-year outlook twice over. The weakness stemmed from a largely inline print, not a bad one, and I think the growth and the guidance underneath it get underweighted by a market that is currently trading the reaction rather than the roadmap.
I don't generally like to chase, but this is close to the AAPL trade from last month, which has bounced almost 10% since I put it on: buying a name that has already done its correcting, on a catalyst the market has not fully priced. The AI capex buildout behind these numbers is relentless, and I think it gets Broadcom back toward $400 from here, with downside limited by where the correction has already taken it.
So: sold the November 260–290 put spread, financing the 450–480 call spread with the credit. The 260 strike sits well below the current print and is the level I'd expect to hold outside a genuine market-wide pullback — that is the scenario this trade is wrong in, not a company-specific one. Call it a buy on the roadmap, not the quarter.
Rating and trade idea
Buy- Structure
- Custom — short put spread financing a call spread
- Strikes
- 260–290P / 450–480C
- Expiry
- Nov
- Credit / debit
- $180 credit
- Max risk
- $2,820
What proves this wrong
A meaningful pullback in the broader market that breaches the 260 downside protection.
In the portfolio
- AVGO Custom · 295-310P 450-465C · opened 19 Aug '26+$64
- AVGO Custom · 260-290P 450-480C · opened 3 Sep '26Open