Any honest Broadcom (AVGO) stock price prediction has to start by killing the laziest explanation on the tape: that AVGO is down because the AI trade is cooling. It is not. On 2 September 2026 Broadcom reported fiscal Q3 AI semiconductor revenue of $16.7 billion, up 221% year over year, and guided Q4 AI revenue to $21.7 billion. The stock closed down 2.74% the next session, and three weeks later it finished Friday 18 September at $357.61, roughly 28% below its 52-week high of $495. Growth was never the problem. What the market is actually repricing is the quality and the ownership of that growth, and the answer to that question sits in the 10-Q rather than the press release.
Here is the number almost nobody is quoting. In the same quarter that AI revenue tripled, Broadcom disclosed that direct sales to a single semiconductor customer, a distributor, accounted for 50% of total company net revenue, against 32% a year earlier. Concentration did not merely persist through the AI build-out; it deepened by 18 percentage points in twelve months. Meanwhile the sell side’s consensus 12-month target of $525 sits 6.1% above the 52-week high, while the tape sits 27.8% below it. To be right, consensus needs a new record plus another 6% within a year. That gap between where analysts are and where the tape is, is the whole argument.
Key facts
- AVGO closed at $357.61 on Friday 18 September 2026 (regular session, 4:00 PM EDT), up 2.97% on the day, with a 52-week range of $289.96 to $495.00 — stockanalysis.com quote data, pulled 19 September 2026.
- Fiscal Q3 2026 revenue was $29.59 billion, up 86%, with non-GAAP operating income of $20.10 billion and free cash flow of $13.67 billion, or 46% of revenue — Broadcom Q3 FY2026 press release, Exhibit 99.1 to the 8-K, 2 September 2026.
- Q4 FY2026 guidance is revenue of approximately $34.8 billion (+93%) at a non-GAAP operating margin of roughly 66% — same 8-K, Item 2.02.
- One distributor customer accounted for 50% of net revenue in the quarter ended 2 August 2026, versus 32% in the year-ago quarter; the top five end customers were about 50% of nine-month revenue — Broadcom Form 10-Q for the quarter ended 2 August 2026, filed 10 September 2026.
- Infrastructure software delivered an 83.7% segment operating margin ($7.33bn on $8.75bn) against 61.3% for semiconductor solutions — 10-Q segment tables.
- Market capitalisation was $1.707 trillion with a 1-year consensus target of $525.00 and an annualised dividend of $2.60 — Nasdaq AVGO summary data, 19 September 2026.
Why a 221% growth print produced a red candle
The mechanics are worth walking through slowly, because the headline and the reaction genuinely do point in opposite directions. Broadcom’s fiscal third quarter ended 2 August 2026. Revenue of $29.59 billion beat the prior year by 86%. GAAP operating income more than doubled to $15.96 billion. Free cash flow reached $13.67 billion. Then AVGO closed at $357.16 on 3 September, down 2.74% from $367.24, and has spent the fortnight since trading either side of $360.
Markets do not punish growth. They punish growth they have already paid for.
By the time the print landed, the stock had already run from a $293.41 low on 30 March 2026 to a $481.57 closing high on 2 June, a 64% move in ten weeks. The guidance, remarkable as it is in absolute terms, arrived into a price that had assumed something close to it. Having tracked the custom-ASIC order cycle since the first hyperscaler XPU programmes moved from evaluation into volume, the pattern is familiar: the re-rating happens on the announcement of a design win, not on the revenue that eventually flows from it, and the revenue phase tends to be the boring part.
There is a second, less comfortable reading. The AI revenue is now large enough that its composition matters more than its growth rate. Of Q4’s guided $34.8 billion, $21.7 billion is AI semiconductors, leaving roughly $13.1 billion of everything else. AI has gone from 56.4% of total revenue in Q3 to a guided 62.4% in Q4. Within the semiconductor segment alone it is about 80%. A business that concentrated is no longer diversified by any reasonable definition, however many product lines sit on the price list.
Broadcom’s own framing has not wavered. “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,” said Hock Tan, President and Chief Executive Officer of Broadcom Inc., in the results release. “In Q4 the momentum continues, and we expect AI semiconductor revenue to accelerate to $21.7 billion, up 236% year-over-year.” Note what that sentence does and does not claim. It is a statement about demand and revenue. It is not a statement about how many customers that demand comes from, and the filing that answers that question was published eight days later.
What the company actually disclosed, and what it did not
The 10-Q filed on 10 September is the document that changed the shape of the argument. Its language is blunt: “A relatively small number of customers account for a significant portion of our net revenue. Direct sales to one semiconductor solutions customer, which is a distributor, accounted for 50% and 46% of our net revenue for the fiscal quarter and three fiscal quarters ended August 2, 2026, respectively, and 32% and 30% of our net revenue for the fiscal quarter and three fiscal quarters ended August 3, 2025, respectively.”
Read that twice. Half the revenue of a $1.7 trillion company now transacts through one counterparty.
Two caveats matter and both cut against over-reading it. First, the counterparty is a distributor, not an end customer, so it aggregates demand from multiple underlying buyers; Broadcom separately states that its top five end customers were approximately 50% of nine-month revenue, and that distributors in aggregate were 56%. Second, routing volume through a distributor of record is ordinary practice in the contract-manufacturing chain and says nothing on its own about credit quality. What it does establish is that the disclosure moved from 32% to 50% in a single year, which is a change in kind rather than degree, and that Broadcom itself flags it as a risk that “increases the risk of quarterly fluctuations.”
The market has been pricing this concentration question for a while, mostly by proxy. When Marvell took custom AI chip work at Google, AVGO fell 4.7% in a single session, a move far larger than the revenue at stake, because a second source at a top XPU customer changes the terminal value of an ASIC franchise rather than next quarter’s shipments. The same logic explains why the stock reacted so strongly when OpenAI’s first in-house chip taped out on Broadcom silicon. Custom silicon revenue is contractual, multi-year and high-visibility right up until a customer qualifies somebody else.
The CFO’s commentary, by contrast, is a study in deliberate narrowness. “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,” said Amie Thuener, Chief Financial Officer of Broadcom Inc. She then guided margin, not mix: “we expect to maintain our non-GAAP operating margin at 66%, flat from a year ago.” Holding a 66% operating margin while revenue grows 93% is the single most impressive line in the release, and it is the one that most directly supports a bull case.
The annuity hiding inside the AI story
This is where the merchant-GPU comparison that dominates coverage stops being useful. Broadcom is two businesses with completely different economics stapled together, and the 10-Q segment tables make the gap explicit.
| Fiscal Q3 2026 (quarter ended 2 Aug 2026) | Semiconductor solutions | Infrastructure software |
|---|---|---|
| Net revenue | $20,839m | $8,752m |
| Segment operating income | $12,770m | $7,325m |
| Segment operating margin | 61.3% | 83.7% |
| Year-on-year revenue growth | +127% | +29% |
| Year-on-year segment profit growth | +145% | +40% |
| Annualised revenue run-rate | $83.4bn | $35.0bn |
Combine two sources and an insight falls out that neither states on its own. The 10-Q gives an infrastructure software segment running an 83.7% operating margin, which annualises to roughly $29.3 billion of segment operating profit on $35.0 billion of revenue. The Nasdaq market-cap figure gives $1.707 trillion of equity value. Even on a punitive 20x multiple, that software annuity alone accounts for something in the region of $585 billion, or about a third of the market capitalisation, for a business growing 29% with essentially no cyclicality and no customer-qualification risk. The AI franchise is being asked to justify the other two-thirds.
Valuation on the whole company is less stretched than the narrative implies. Non-GAAP diluted EPS of $3.32 annualises to $13.28, putting AVGO on roughly 26.9x annualised Q3 non-GAAP earnings at $357.61. Against Q4’s guided revenue annualised to $139.2 billion, the market cap is 12.3x sales. Those are not bubble multiples for 86% growth, which is precisely why the 28% drawdown is interesting rather than obvious. The broader capex question behind all of this has already been examined in FinanceFeeds’ work on what a deliberate slowdown does to $700 billion of AI capex, and it does not need rebuilding here.
Where the structural pressure actually sits
Customer concentration is the tension that matters for AVGO, and it is not primarily a regulatory one. It is contractual. A merchant GPU vendor sells a catalogue part to anyone with a purchase order; an ASIC vendor co-designs a part with one customer, amortises non-recurring engineering against one programme, and lives or dies on that programme’s renewal. The switching cost runs in both directions. That is why a second-source announcement at a single hyperscaler moves the stock more than a quarter’s results do.
Export control is the second axis, and it is genuinely bidirectional for Broadcom. The US Bureau of Industry and Security controls advanced accelerators and the networking silicon that lashes them together, so custom XPUs designed for US hyperscalers are largely insulated, while any China-destined derivative is not. Broadcom’s disclosed revenue geography has historically carried meaningful China exposure through the contract-manufacturing channel, and because roughly 56% of revenue moves through distributors, the end-market destination of a given wafer is a layer removed from the invoice.
Then there is the balance sheet, which is where the VMware acquisition still shows up. Broadcom filed a Form S-4 on 10 September 2026 registering an exchange offer for $654,004,000 of 4.926% Senior Notes due 2037, originally issued in April 2022. It is a housekeeping registration rather than a transaction, and it will not change indebtedness by a dollar; the exchange notes carry identical terms. Worth flagging only because a registration statement from a serial acquirer invites the wrong assumption, and there is no new deal inside this one.
Memory cost inflation is a live input on the bill of materials for AI systems, though it lands on Broadcom’s customers more than on Broadcom, whose products are logic rather than DRAM or NAND.
Cash gives management room that most of the peer group lacks. The company held $24.0 billion at quarter end and generated $13.67 billion of free cash flow in the quarter; the $0.65 quarterly dividend costs roughly $3.1 billion a payment, which is under a quarter of free cash flow. Annualised at $2.60 against Friday’s close, the yield is 0.73%. Small, but real, and it implies a holder base that includes income and quality mandates rather than pure momentum, which is part of why the drawdown has been orderly rather than violent.
The call: $465 bull, $395 base, $280 bear
All three levels are derived from Q4 guidance and the Q3 segment economics rather than from a target list, and the arithmetic is shown so it can be disagreed with.
| Case | Level | Move from $357.61 | Implied market cap | Multiple of annualised Q3 non-GAAP EPS ($13.28) | Probability |
|---|---|---|---|---|---|
| Bull | $465 | +30.0% | $2.22tn | 35.0x | 30% |
| Base | $395 | +10.5% | $1.89tn | 29.7x | 50% |
| Bear | $280 | −21.7% | $1.34tn | 21.1x | 20% |
The bull case at $465 assumes Q4 lands at or above the guided $34.8 billion with the 66% non-GAAP operating margin intact, and that the software segment holds its 83.7% margin. That puts the company on roughly 15.9x the $139.2 billion annualised revenue run-rate. It deliberately stops 6% below the $495 52-week high: a bull case that does not require a new record is a more honest one, and it is why this number sits well under the $525 consensus.
The base case at $395 is the guidance delivered and the multiple unchanged, roughly 29.7x annualised non-GAAP earnings, which is where the stock traded for much of the summer.
The bear case at $280 assumes AI semiconductor revenue stalls near the Q3 level of $16.7 billion instead of reaching the guided $21.7 billion, taking Q4 revenue to about $29.8 billion and the annualised run-rate to $119.2 billion. At 11.2x that reduced run-rate, the equity is worth about $1.34 trillion. The trigger would most plausibly be a second-source qualification at a major XPU customer rather than a demand collapse.
Invalidation. A weekly close below $289.96, the 52-week low, would break the base case and hand control to the bear path. On the other side, a weekly close above $420 would put the June highs back in play and argue the base case is too conservative. What would change my mind fastest is the concentration line in the next 10-Q: if the single-distributor share keeps climbing past 50%, the bear probability deserves to rise regardless of the growth rate. For context on how these levels compare across the custom-silicon complex, FinanceFeeds has run the same exercise on Marvell and on Arm.
FAQ
What is the Broadcom (AVGO) stock price prediction for the next 12 months?
This analysis sets a base case of $395, a bull case of $465 and a bear case of $280 against Friday’s close of $357.61, with probabilities of 50%, 30% and 20% respectively. Each level is derived from Broadcom’s own Q4 FY2026 revenue guidance of $34.8 billion and the 66% non-GAAP operating margin the company guided to, rather than from an aggregated analyst target.
Why did AVGO stock fall after such strong Q3 FY2026 earnings?
AVGO closed down 2.74% on 3 September 2026 despite AI semiconductor revenue growing 221%. The stock had already climbed 64% from its March low to its June peak, so the guidance arrived into a price that largely assumed it. The 10-Q filed a week later also disclosed that single-customer concentration had risen from 32% to 50% of net revenue.
How concentrated is Broadcom’s customer base?
Per the Form 10-Q for the quarter ended 2 August 2026, direct sales to one semiconductor customer, a distributor, were 50% of net revenue, up from 32% a year earlier. Distributors in aggregate accounted for 56% of nine-month revenue, and Broadcom estimates its top five end customers at approximately 50%. The company flags this as raising the risk of quarterly fluctuations.
Is Broadcom an AI stock or an infrastructure software business?
Both, in measurable proportions. Semiconductor solutions produced $20.84 billion of Q3 revenue at a 61.3% segment operating margin; infrastructure software produced $8.75 billion at 83.7%. Software annualises to roughly $35 billion of revenue and $29.3 billion of segment operating profit, growing 29% with no customer-qualification risk, which is a genuine annuity sitting underneath the AI cycle.
What dividend does Broadcom pay and what is the yield?
The board declared a quarterly cash dividend of $0.65 per share on 2 September 2026, payable 30 September to holders of record on 21 September. Annualised at $2.60 against the 18 September close of $357.61, that is a yield of 0.73%. The payment costs roughly $3.1 billion per quarter, under a quarter of the $13.67 billion of free cash flow generated in Q3.
What would invalidate this Broadcom (AVGO) stock price prediction?
A weekly close below the 52-week low of $289.96 would break the base case and favour the $280 bear level. A weekly close above $420 would argue the $395 base is too conservative. The specific disclosure to watch is the single-distributor revenue share in the next quarterly filing; a reading above 50% would justify raising the bear probability even if growth stays intact.
Disclaimer
This article is analysis and information, not investment advice, and no part of it is a recommendation to buy or sell any security. Figures are sourced to Broadcom’s SEC filings and to market data pulled on 19 September 2026 and were accurate as of the 18 September 2026 close. Equity prices move and capital is at risk. Readers should conduct their own research and consider their own circumstances before making any financial decision.
