
- Broadcom Q3 FY2026 Earnings Results: Revenue, EPS, AI Sales and Margins
- Did Broadcom Pass Our Two Engines and One Gearbox Test?
- Why Did AVGO Stock Barely Move After Broadcom Earnings?
- Broadcom AI Revenue Drove 84% of Q3 Growth: Why It Matters
- How VMware Is Protecting Broadcom’s Margins During the AI Boom
- Broadcom Q4 Guidance: Lower Margin, More Profit
- What Did Broadcom CEO Hock Tan Say?
- Broadcom’s $115B and $230B AI Revenue Forecast: The Math Explained
- The Demand-Delivery-Economics Test for Broadcom
- Broadcom’s Cash Flow May Be Better Than the Headline Beat
- Broadcom Q3 Earnings: What Went Right and What Investors Should Watch
- The Three Biggest Risks for AVGO Investors
- Is AVGO Stock Attractive After Q3 Earnings?
- What Should AVGO Investors Watch Next?
- Broadcom Q3 Earnings Analysis: The Bottom Line
Broadcom delivered record revenue, record operating profit and record free cash flow. AI semiconductor sales more than tripled, Q3 margins beat guidance, and management raised its FY2027 AI revenue outlook to $115 billion. Yet AVGO stock finished extended trading down just 0.82%. The market did not reject Broadcom’s AI story. It simply did not receive a surprise large enough to settle every debate around margins, execution and valuation.
Let's break down Broadcom’s Q3 FY2026 earnings, its Q4 guidance, the after-hours AVGO stock reaction and CEO Hock Tan’s two-year AI roadmap. Then we will test the result against our earnings-preview framework and identify what investors should track from here.
Broadcom Q3 FY2026 Earnings Results: Revenue, EPS, AI Sales and Margins
Broadcom reported its fiscal third-quarter results on September 2, 2026, for the quarter ended August 2.
| Metric | Q3 FY2026 | Q3 FY2025 | YoY change | Expectation or guide |
| Total revenue | $29.591B | $15.952B | +86% | $29.49B estimate |
| Semiconductor Solutions revenue | $20.839B | $9.166B | +127% | - |
| AI semiconductor revenue | $16.7B | About $5.2B | +221% | About $16.0B guide |
| Infrastructure Software revenue | $8.752B | $6.786B | +29% | About $8.9B guide |
| GAAP net income | $13.088B | $4.140B | +216% | - |
| GAAP diluted EPS | $2.68 | $0.85 | +215% | - |
| Adjusted diluted EPS | $3.32 | $1.69 | +96% | $3.24 estimate |
| Non-GAAP operating margin | 67.9% | About 65.5% | +240 bps | About 67% guide |
| Free cash flow | $13.665B | $7.024B | +95% | 46% of revenue |
Sources: Broadcom Q3 FY2026 earnings release filed with the SEC
Against the expectations used in our preview, total revenue beat by only about 0.3% and adjusted EPS by roughly 2.5%. Those are respectable, but not spectacular, surprises.
The quality of the beat was more interesting. AI semiconductor revenue exceeded Broadcom’s own $16 billion target by $700 million, or 4.4%. Non-GAAP gross margin came in at approximately 75%, one percentage point above guidance, while operating margin reached a record 67.9%.
In other words, Broadcom did not merely sell more AI chips. It sold more than expected without allowing the cost structure to run away.
Did Broadcom Pass Our Two Engines and One Gearbox Test?
Before the earnings release, we described Broadcom as two growth engines connected to one profitability gearbox. The first engine was AI semiconductors, the second was infrastructure software, and the gearbox was Broadcom’s ability to convert growth into operating profit and cash.
Here is how the actual results scored.
| Preview test | What we wanted to see | Q3 result | Verdict |
| AI engine | More than $16B of AI revenue with stronger forward demand | $16.7B, up 54% QoQ; Q4 guide of $21.7B | Strong pass |
| Software engine | Revenue near $8.9B with healthy ARR growth | $8.752B; ARR grew 15% | Mixed |
| Profitability gearbox | Gross margin near 74%, operating margin near 67% | About 75% and 67.9% | Strong pass |
| Cash conversion | Free cash flow moving with earnings | $13.665B, or 46% of revenue | Strong pass |
| Long-term visibility | A more credible FY2027 AI path | FY2027 raised to $115B; FY2028 set at $230B | Strong pass |
Operationally, this landed close to the “high-quality beat” scenario from the preview. AI revenue, margins, cash flow and long-term visibility improved together.
The one yellow flag was software. Infrastructure Software revenue was around $148 million, or 1.7%, below the approximately $8.9 billion anticipated in the preview. Annual recurring revenue, or ARR, grew 15%, compared with 17% growth discussed in the previous quarter.
That is not a broken engine. It is an engine that deserves another look when Broadcom reports Q4.
Why Did AVGO Stock Barely Move After Broadcom Earnings?
AVGO closed the regular September 2 session at $367.24, down 0.66%. The first earnings reaction sent the shares roughly 3% lower. The stock then briefly turned about 2% higher when Broadcom disclosed its FY2027 and FY2028 AI outlook on the conference call.
By 7:59 p.m. ET, AVGO was trading at approximately $364.23, down 0.82% from the regular close. The stock therefore travelled in both directions before ending almost where it started.
That muted finish becomes easier to understand when the guidance is compared with different Wall Street datasets.
| Q4 item | Broadcom guidance | Comparison | Result |
| Total revenue vs LSEG | $34.8B | $35.03B | $230M below |
| Total revenue vs FactSet | $34.8B | $34.68B | $120M above |
| Non-GAAP operating margin | 66% | About 66.5% expected | 50 bps below |
| AI semiconductor revenue | $21.7B | +236% YoY | Strong growth |
| FY2027 AI revenue | $115B | Previous target above $100B | Raised |
| FY2028 AI revenue | $230B | New outlook | Major long-term signal |
Sources: Reuters earnings coverage, Investor’s Business Daily and Barron’s.
The same $34.8 billion outlook was a slight miss against LSEG and a slight beat against FactSet. That left investors without a clean headline surprise. The 66% operating-margin guide also came in slightly below expectations.
Before the results, options were implying an approximately 7% move by the end of earnings week. The final 0.82% after-hours decline was less than one-eighth of that magnitude, although the two figures cover different time windows.
The stock’s response can therefore be summarised simply: Q3 was better than expected, Q4 was close to expected, and the long-term AI forecast was excellent but still carried execution risk.
Broadcom AI Revenue Drove 84% of Q3 Growth: Why It Matters
Broadcom’s 86% headline revenue growth looks diversified until the increase is separated into its three main sources.
| Revenue stream | Q3 FY2025 | Q3 FY2026 | YoY increase | Share of total growth |
| AI semiconductors | About $5.20B | $16.70B | About $11.50B | About 84% |
| Infrastructure software | $6.786B | $8.752B | $1.966B | About 14% |
| Non-AI semiconductors | About $3.97B | About $4.14B | About $0.17B | About 1% |
| Total | $15.952B | $29.591B | $13.639B | 100% |
Source: Author calculations using Broadcom’s reported segment and AI revenue.
Our preview estimated that AI would generate roughly 80% of Broadcom’s year-on-year revenue increase. The actual contribution was closer to 84%.
This is both the most bullish and the most uncomfortable number in the report.
It is bullish because Broadcom is no longer a company waiting for AI revenue to become meaningful. AI already represents approximately 56% of total quarterly revenue.
It is uncomfortable because Broadcom’s growth rate is becoming more dependent on a small group of very large customers continuing to build data centres at extraordinary speed.
How VMware Is Protecting Broadcom’s Margins During the AI Boom
AI receives the headlines, but infrastructure software continues to protect Broadcom’s profitability.
The software division generated only about 30% of Q3 revenue, yet its operating margin reached approximately 84%. Applying that rounded margin to reported software revenue produces estimated segment operating profit of around $7.35 billion.
That would represent roughly 37% of Broadcom’s total non-GAAP operating income, despite software contributing less than one-third of revenue.
| Software profitability measure | Q3 FY2026 |
| Infrastructure Software revenue | $8.752B |
| Share of Broadcom revenue | 29.6% |
| Software operating margin | About 84% |
| Estimated software operating profit | About $7.35B |
| Estimated share of total operating profit | About 37% |
Source: Author calculations using management’s rounded segment margins from the Q3 earnings call.
This is the balancing act inside Broadcom. Custom AI accelerators bring explosive revenue growth but have lower gross margins because they contain expensive components such as high-bandwidth memory. VMware contributes slower growth but exceptionally high profitability.
Think of VMware as the suspension system on a fast car. The AI engine creates the speed; software stops every change in the road from reaching the passengers.
The concern is that Q4 software revenue is expected to stabilise around $8.7 billion, slightly below Q3. If AI keeps taking a larger share while software becomes flat, the cushioning effect weakens.
Broadcom Q4 Guidance: Lower Margin, More Profit
Broadcom expects Q4 revenue of approximately $34.8 billion, up 93% year over year. AI semiconductor revenue is projected to reach $21.7 billion, representing 236% growth.
Here is the more revealing sequential comparison.
| Metric | Q3 FY2026 | Q4 FY2026 guide | Sequential change |
| Total revenue | $29.591B | $34.8B | +17.6% |
| AI semiconductor revenue | $16.7B | $21.7B | +29.9% |
| AI share of total revenue | 56.4% | 62.4% | +6.0 percentage points |
| Non-GAAP gross margin | About 75% | About 73% | -200 bps |
| Implied non-GAAP gross profit | $22.191B | About $25.40B | +14.5% |
| Non-GAAP operating margin | 67.9% | About 66% | -190 bps |
| Implied non-GAAP operating income | $20.095B | About $22.97B | +14.3% |
| Capital expenditure | $532M | About $1.4B | +163% |
Sources: Broadcom’s Q3 earnings release, earnings-call guidance and author calculations. Implied profit figures are estimates, not company guidance.
Approximately $5 billion of Broadcom’s projected $5.21 billion sequential revenue increase would come from AI. That means AI is responsible for roughly 96% of the expected Q4 growth.
Gross margin measures what remains after the direct cost of producing a product. Operating margin goes further by also accounting for expenses such as research, sales and administration.
Think of a restaurant. Gross margin is what remains from each thali after paying for its ingredients. Operating margin is what remains after paying the staff and rent as well. Broadcom’s AI “ingredients”, particularly memory, are becoming more expensive. But the company does not need to increase its corporate expenses at the same speed as revenue.
That is why Q4 gross margin can fall by two percentage points while estimated gross-profit dollars still rise by around 14.5%. The margin slice becomes thinner, but the overall pie gets much larger.
Management’s argument is therefore reasonable: investors should not examine gross margin in isolation. However, margins still matter. If gross margin keeps falling and operating expenses eventually begin rising faster, the current operating-leverage benefit could fade.
What Did Broadcom CEO Hock Tan Say?
Hock Tan described Q3 demand as “simply hot” and said Broadcom was still early in the custom-AI cycle.
Broadcom now has six major XPU customers. An XPU is a custom accelerator designed around a specific company’s AI workloads. A general-purpose GPU is like a Swiss Army knife: powerful and flexible. An XPU is closer to a chef’s knife designed for one kitchen. It may do fewer jobs, but it can perform its chosen job faster and more efficiently.
Custom accelerators accounted for 73% of Broadcom’s Q3 AI revenue. That implies approximately $12.2 billion of XPU revenue, with AI networking contributing an estimated $4.5 billion.
| Management signal | What Broadcom disclosed | Why it matters |
| XPU shipments | Increased more than 3.5 times YoY | Custom accelerators are now the main AI driver |
| AI networking | Increased more than 2.5 times YoY | Broadcom earns content beyond the accelerator |
| New TPU generations are shipping under a long-term agreement | Reduces concerns of an immediate displacement | |
| Anthropic | Expected to become Broadcom’s largest XPU customer in FY2027 | Customer mix is shifting beyond Google |
| OpenAI | Initial custom accelerator is shipping; later generations are in development | Creates a multi-generation opportunity |
| Meta | Production shipments of its custom accelerator are expected in Q4 | Adds another large deployment programme |
| Supply | Broadcom says supply is secured for its published outlook | Supports the credibility of the revenue targets |
Sources: Broadcom’s earnings call transcript, Reuters and The Wall Street Journal.
Broadcom also expects its AI networking business to grow broadly alongside custom accelerators. This is strategically important. A customer may use Broadcom networking even when some of its accelerators come from Nvidia, Marvell or another supplier.
Broadcom is therefore not making only the engine of the AI data centre. It also sells parts of the road system that allow thousands of engines to communicate.
Broadcom’s $115B and $230B AI Revenue Forecast: The Math Explained
The biggest announcement was not contained in the press release. It came during the earnings call. Management expects AI semiconductor revenue of approximately $115 billion in FY2027 and $230 billion in FY2028.
| Fiscal year | AI semiconductor revenue | Implied growth |
| FY2026 estimate | About $58B | +186% YoY |
| FY2027 outlook | About $115B | +98% |
| FY2028 outlook | About $230B | +100% |
Source: Broadcom management guidance from the Q3 FY2026 earnings call.
Moving from $58 billion to $230 billion in two years implies an annualised growth rate of approximately 99%. Broadcom is effectively forecasting that this already-large AI business will almost quadruple.
The FY2027 and FY2028 forecasts also add up to $345 billion of AI semiconductor revenue. Tan rounded this to approximately $350 billion of potential shipments over the two years.
That number needs careful handling. It is management guidance, not recognised backlog and certainly not cash already collected. Customers still need data-centre buildings, power, racks, memory, advanced packaging and financing before every planned gigawatt can start producing revenue.
Broadcom believes its six customers could deploy around 30 gigawatts of capacity across this period. Management estimates that its semiconductor content could remain around $20 billion to $30 billion per gigawatt, even as each new generation of chip becomes more powerful.
The size of the opportunity is no longer the hardest question. The harder question is how quickly the physical world can catch up with the demand.
The Demand-Delivery-Economics Test for Broadcom
A useful way to track Broadcom from here is to separate the AI story into three forms of proof.
| Test | What investors should examine | Current assessment |
| Demand proof | Orders, customer programmes and multiyear revenue outlook | Strong |
| Delivery proof | Wafers, HBM, substrates, packaging, power and completed data centres | Strong but constrained |
| Economics proof | Operating margin, cash flow, capital needs and financing exposure | Strong today; Q4 needs watching |
Demand proof is the easiest part. AI revenue grew 221%, customer demand exceeds Broadcom’s current supply plan, and management provided forecasts through FY2028.
Delivery proof is more complicated. Tan said Broadcom considers land, power and the readiness of data-centre buildings before including deployments in its outlook. Leading-edge wafers, substrates, high-bandwidth memory and other system memory can each become bottlenecks.
Broadcom is responding by expanding substrate and optical-component capacity in Singapore and the US. That helps explain why Q4 capital expenditure is expected to jump from $532 million to $1.4 billion.
Economics proof remains encouraging. Broadcom generated $13.665 billion of free cash flow in Q3, equal to 46% of revenue, even as AI revenue rose 54% sequentially. The important test is whether similar cash conversion survives a heavier mix of custom accelerators and higher capital spending.
Broadcom’s Cash Flow May Be Better Than the Headline Beat
Broadcom’s cash generation gives it room to invest while reducing debt.
| Balance-sheet measure | FY2025 year-end | Q3 FY2026 | Change |
| Cash and equivalents | $16.178B | $23.975B | +$7.797B |
| Short and long-term debt | $65.136B | $59.419B | -$5.717B |
| Net debt | About $48.958B | About $35.444B | Improved by $13.514B |
Source: Author calculations using Broadcom’s Q3 balance sheet.
During Q3, Broadcom produced $13.665 billion of free cash flow, paid approximately $3.1 billion in dividends and repaid $5.6 billion of long-term debt. Subtracting the dividends and debt repayment from free cash flow leaves roughly $5 billion before other cash movements. Cash on the balance sheet increased by approximately $4.35 billion sequentially.
This is what high-quality growth should look like: accounting profit, operating profit and cash all moving in the same direction.
Stock-based compensation also declined 13% year over year to $2.019 billion. It still equalled approximately 6.8% of revenue and should not be ignored, but the decline strengthens the quality of the adjusted earnings comparison.
Broadcom did not disclose a Q3 repurchase amount in the earnings release. Investors following the sharp Q2 slowdown in buybacks will need the next Form 10-Q for a complete answer.
Broadcom Q3 Earnings: What Went Right and What Investors Should Watch
| What went right | What needs watching |
| AI revenue beat Broadcom’s guide by $700M | Software revenue was below the preview expectation |
| Gross and operating margins beat guidance | Software ARR growth slowed from 17% to 15% |
| Free cash flow rose 95% to $13.665B | Q4 gross margin is expected to fall to 73% |
| FY2027 AI outlook increased to $115B | Q4 operating-margin guidance was slightly light |
| FY2028 AI outlook reached $230B | Nearly all near-term growth is now coming from AI |
| Net debt improved by about $13.5B in nine months | Customer and supply-chain concentration remain high |
| Google relationship received a long-term endorsement | AI financing structures add contingent risk |
Our view is that Q3 contained no major operating failure. The weaknesses were relative to expectations, not absolute performance.
Software grew 29%, but it missed the internal bar. Q4 revenue is expected to grow 93%, but it did not clearly beat every consensus dataset. Operating profit should rise, but the margin percentage is expected to decline.
These are reasons for disciplined monitoring, not evidence that Broadcom’s business has suddenly deteriorated.
The Three Biggest Risks for AVGO Investors
1. Customer concentration: Broadcom’s most recent Form 10-Q said its five largest end customers represented approximately 45% of revenue during the first half of FY2026, up from 40% a year earlier.
Concentration works wonderfully while customers are increasing orders. It becomes painful if even one customer delays a data centre, changes suppliers or negotiates lower pricing.
2. Google’s multi-supplier strategy: Google recently expanded its custom-silicon relationship with Marvell. The deal raised understandable concerns because Broadcom has historically been Google’s main TPU implementation partner.
However, the evidence does not yet show a simple replacement. Broadcom said its Google engagement had “never been stronger” and discussed a long-term agreement covering future TPUs and networking. Reuters also cited an analyst who viewed the Marvell agreement as an expansion of Google’s silicon spending rather than a direct displacement of Broadcom.
The sensible conclusion is that Google wants supplier diversity. Broadcom can continue growing while still losing some share of an expanding opportunity.
3. AI financing exposure: Broadcom’s previous Form 10-Q disclosed a backstop arrangement connected with AI racks and customer leases. Maximum exposure under the first arrangement could reach $29 billion as racks are deployed, although customer payments and the resale or assumption of the underlying assets could reduce that figure.
The $29 billion maximum is about 1.2 times Broadcom’s Q3 cash balance. It is not current debt or an immediate loss, but it is too large to treat as a footnote.
Management said outside financial partners independently underwrite and capitalise the assets, while any Broadcom residual-value guarantees should be modest. However, management did not provide a maximum exposure for possible future arrangements on the Q3 call.
Investors should therefore distinguish between semiconductor demand and the financing used to turn that demand into operational data centres.
Is AVGO Stock Attractive After Q3 Earnings?
The earnings strengthened Broadcom’s operating case. They did not automatically settle the valuation question.
At the final after-hours price of approximately $364.23, AVGO was around 25.8% below its June 2026 high of $495. Broadcom shares had gained only about 6% during 2026 through September 2, significantly trailing the wider semiconductor index, according to Reuters.
Tan also said Broadcom remained on track to exceed $30 in EPS in FY2028. Dividing the after-hours share price by $30 produces a ratio of approximately 12.1 times that future EPS objective.
That is not Broadcom’s current P/E and it is not a price target. It assumes that management delivers the objective and ignores the time, financing, competition and execution risk between now and FY2028. It is simply a useful way to see how much of the long-term argument depends on the company meeting an unusually ambitious plan.
Our interpretation is balanced but positive:
- The business performed better than the stock reaction suggests.
- AI demand, customer visibility and cash generation strengthened.
- The Q4 margin guide and flat sequential software outlook prevented a clean victory.
- The investment story has moved from proving AI demand to proving physical delivery and financial discipline.
A small post-earnings decline is not, by itself, a buy signal. The stronger signal would be Broadcom delivering $21.7 billion of Q4 AI revenue while keeping operating margin close to 66%, converting profit into cash and limiting balance-sheet exposure from AI financing.
What Should AVGO Investors Watch Next?
| Metric or event | Current benchmark | Why it matters |
| Q4 total revenue | About $34.8B | Tests whether 93% growth is achievable |
| Q4 AI semiconductor revenue | About $21.7B | Requires almost 30% sequential growth |
| Q4 gross margin | About 73% | Measures the cost of the XPU and memory mix |
| Q4 operating margin | About 66% | Tests management’s operating-leverage argument |
| Q4 software revenue | About $8.7B | Shows whether VMware can resume sequential growth |
| Q4 capital expenditure | About $1.4B | Reveals the cost of relieving supply constraints |
| AI financing exposure | $29B maximum on the first disclosed backstop | Tests balance-sheet discipline |
| FY2027 AI outlook | About $115B | Requires almost 100% annual growth |
| FY2028 AI outlook | About $230B | Tests the full 30-gigawatt deployment thesis |
Broadcom currently plans to report its Q4 and full-year FY2026 results on December 9, 2026, after the US market closes, according to the earnings call.
Broadcom Q3 Earnings Analysis: The Bottom Line
Broadcom’s Q3 earnings were stronger than the nearly flat AVGO stock reaction implies. AI semiconductor revenue reached $16.7 billion, margins beat guidance, free cash flow touched $13.665 billion, and management raised the FY2027 AI outlook to $115 billion while introducing a $230 billion FY2028 target.
The market’s hesitation also makes sense. Q4 guidance sat on either side of consensus depending on the data provider, operating-margin guidance was slightly below expectations, software momentum was not perfect, and the long-term AI targets require an enormous amount of physical infrastructure and customer financing.
The most important shift is this: Broadcom no longer needs to prove that customers want its AI chips. It now needs to prove that those chips can be delivered, deployed and paid for at the scale management has promised.
Demand has passed the test. Delivery and economics are the next exams.