Post Highlights
-
Q2 AI revenue:
-
Implied non-AI semiconductor + software:
-
If software stays flat:
-
Cash dividends: $3.086 billion ($0.65 per share quarterly)
-
Stock repurchases: $7.850 billion
Broadcom reported Q1 FY2026 results on March 4, 2026, for the quarter ending February 1, 2026. This article assumes you already know how Broadcom's two-segment business is structured how the semiconductor side makes money from networking ASICs and custom AI accelerators, how the VMware software side works on subscription economics, and what Hock Tan's acquisition playbook actually does to the businesses he buys. If you are starting fresh, read the business model overview first.
Broadcom's $51.6B revenue splits cleanly into two businesses : custom silicon and enterprise software
The quarter in one sentence
The semiconductor segment grew 52% year over year. The infrastructure software segment grew 1%.
That gap is the story of Q1 FY2026, and it is the lens through which every other number in this report should be read.
Q1 FY2026 headline results
| Metric | Q1 FY2026 | Q1 FY2025 | YoY change |
|---|---|---|---|
| Revenue | $19.311B | $14.916B | +29% |
| GAAP net income | $7.349B | $5.503B | +34% |
| Non-GAAP net income | $10.185B | $7.823B | +30% |
| Adjusted EBITDA | $13.128B | $10.083B | +30% |
| Adjusted EBITDA margin | 68% | 68% | flat |
| GAAP diluted EPS | $1.50 | $1.14 | +32% |
| Non-GAAP diluted EPS | $2.05 | $1.60 | +28% |
| Free cash flow | $8.010B | $6.013B | +33% |
| FCF margin | 41% | 40% | +1pp |
The GAAP vs non-GAAP gap is large and worth understanding before reading any further. The $2.836 billion difference between GAAP net income ($7.349B) and non-GAAP net income ($10.185B) comes primarily from three items. Amortization of acquisition-related intangible assets was $1.969 billion in Q1 this is the cost of VMware and prior acquisitions spreading through the income statement over years. Stock-based compensation was $2.176 billion. Restructuring charges were $116 million. None of these items affect cash flow directly. The adjusted EBITDA of $13.128 billion and FCF of $8.010 billion are the cleaner read on what the business actually generated in cash terms.
Segment breakdown: two businesses, one result
| Segment | Q1 FY2026 | Q1 FY2025 | YoY change | Share of revenue |
|---|---|---|---|---|
| Semiconductor solutions | $12.515B | $8.212B | +52% | 65% |
| Infrastructure software | $6.796B | $6.704B | +1% | 35% |
| Total | $19.311B | $14.916B | +29% | 100% |
The semiconductor segment at 65% of revenue growing 52% year over year is doing most of the lifting. Without it, the 29% headline growth number would look very different. Infrastructure software at $6.796 billion, up 1% from $6.704 billion in Q1 FY2025, barely moved.
That Q1 FY2025 comparison matters. By Q1 FY2025 (the quarter ending February 2025), Broadcom had owned VMware for over a year. Price increases and subscription conversion were already underway. This is not a tough comp issue. This is the first clean apples-to-apples read on VMware under the new model, and it grew 1%.
I will come back to what that means. First, the semiconductor story, because it deserves its own section.
The semiconductor side: AI revenue is accelerating, not plateauing
Q1 FY2026 AI revenue: $8.4 billion (+106% year over year)
That number is the headline inside the headline. AI revenue alone custom AI accelerators (XPUs) and AI networking silicon grew 106% year over year to $8.4 billion in a single quarter. That is 67% of Broadcom's entire semiconductor segment revenue coming from AI.
Hock Tan said on the earnings call: "Q1 AI revenue of $8.4 billion grew 106% year-over-year, above our forecast, driven by robust demand for custom AI accelerators and AI networking."
Two components drive the AI semiconductor number.
Custom AI accelerators (XPUs). Google's TPU and Meta's MTIA are the most publicly known examples of chips where Broadcom has had design involvement. These are chips built for one hyperscaler's specific training workload, not general-purpose GPUs. Once a hyperscaler deploys a training cluster built around one of these chips clusters with hundreds of thousands of processors switching to a different architecture requires rebuilding everything. That process takes 3-5 years minimum. The architectural lock-in is why XPU revenue is sticky in a way that most chip revenue is not.
AI networking silicon. Broadcom's Tomahawk and Jericho product lines dominate Ethernet-based data center fabric. As AI training clusters scale more GPUs, more servers, more interconnects the switching infrastructure scales in direct proportion. Every time a hyperscaler adds capacity to its AI cluster, it needs more Broadcom networking chips. This demand is structural, not discretionary.
Backing out the AI revenue from the semiconductor segment:
| Component | Q1 FY2026 | Q1 FY2025 (est.) | YoY |
|---|---|---|---|
| AI semiconductor | $8.4B | ~$4.1B | +106% |
| Non-AI semiconductor | ~$4.1B | ~$4.1B | ~flat |
| Total semiconductor | $12.515B | $8.212B | +52% |
The non-AI semiconductor business broadband chips, wireless connectivity, storage controllers, enterprise networking is roughly flat year over year. All of the semiconductor segment's growth came from AI. The non-AI semiconductor business is essentially flat. If hyperscaler AI capex pulled back hard, there is no comparable growth engine sitting underneath it. If AI capex spending by hyperscalers slowed materially, the semiconductor segment would lose its primary growth engine with nothing comparable to replace it from the non-AI side.
For context on how Broadcom's semiconductor business fits within the broader AI infrastructure chain and its dependency on TSMC's manufacturing:
TSMC spent $40.9B on capex in 2025: this is why no competitor has caught up
The infrastructure software side: 1% growth and what it means
Infrastructure software generated $6.796 billion in Q1 FY2026, up from $6.704 billion in Q1 FY2025. Growth of $92 million on a $6.7 billion base.
This is the VMware retention signal that matters most in this report, and I want to be precise about what we know and what we do not.
What 1% growth tells us:
Broadcom acquired VMware in November 2023. By Q1 FY2025 (the quarter ending February 2025), Broadcom had been running VMware for over a year implementing price increases, eliminating most of the product portfolio, and converting perpetual license customers to multi-year subscriptions. The Q1 FY2025 software revenue base of $6.704 billion was already the "Hock Tan model" in operation.
Growing 1% against that base means total software revenue is essentially flat. Broadcom raised prices on retained customers and converted them to subscriptions. Some customers left. The net result of higher prices on fewer customers is... approximately the same dollar amount.
What 1% growth does not tell us:
It does not tell us the segment is broken. Subscription conversion creates revenue recognition timing issues revenue from multi-year contracts spreads over the contract term rather than landing in one period. If Broadcom signed significant new multi-year contracts in recent quarters, some of that revenue is deferred. The subscription backlog, not the in-quarter revenue, would tell you more about future software revenue. Broadcom does not report that granular breakdown publicly.
It also does not tell us whether this is temporary or structural. One quarter at 1% is data, not a trend.
What the Track Record entry says:
Finovian's published analytical stance is that full-year FY2026 infrastructure software revenue growth above 5% confirms the VMware switching cost assumption is holding. Above 10% confirms the subscription conversion is working as designed. Below 5% would suggest attrition is more significant than the model allows.
Q1 at 1% is below the floor on a quarterly basis. The measurement is full-year, so Q1 is one data point. But it is not the number you want to see if you believe the software retention thesis is clean.
Q2 FY2026 will give a sharper read. Two consecutive quarters of sub-5% growth would be harder to attribute to subscription timing effects.
Margins and FCF: the financial structure held exactly as expected
Adjusted EBITDA came in at $13.128 billion, representing 68% of revenue flat year over year as a percentage. That 68% margin held despite significant revenue mix shift (semiconductor growing faster than software, and software typically carrying higher software-equivalent margins on a non-GAAP basis).
Free cash flow was $8.010 billion at 41% of revenue. Capital expenditures were $250 million, which is less than 1.3% of revenue. The fabless model doing its job Broadcom designs, TSMC builds. Broadcom keeps the margin without carrying a fab on its balance sheet.
| Margin metric | Q1 FY2026 | Q1 FY2025 | Change |
|---|---|---|---|
| Adjusted EBITDA margin | 68% | 68% | flat |
| FCF margin | 41% | 40% | +1pp |
| Capex as % of revenue | 1.3% | 0.7% | +0.6pp |
The GAAP gross margin was $13.157 billion, but this number is heavily affected by the $1.462 billion in acquisition intangible amortization sitting in cost of revenue. Non-GAAP gross margin was $14.868 billion (77% of revenue), which is the cleaner view of Broadcom's pricing power relative to its actual cost to produce.
Q2 FY2026 guidance: $22 billion is a significant step up
| Metric | Q2 FY2026 guidance | Q2 FY2025 actual | YoY |
|---|---|---|---|
| Revenue | ~$22.0B | ~$14.9B | +47% |
| Adjusted EBITDA margin | ~68% | "" | flat |
| AI semiconductor revenue | ~$10.7B | "" | "" |
$22.0 billion in Q2 revenue is a 14% sequential jump from Q1's $19.3 billion. A 14% sequential jump in one quarter is not normal operating growth. Something structurally shifted. And Hock Tan's comment that AI revenue will reach $10.7 billion in Q2 up from $8.4 billion in Q1 means AI semiconductor revenue alone would grow 27% in a single quarter sequentially.
If Q2 comes in at $22 billion with $10.7 billion in AI revenue, the math runs like this:
- Q2 AI revenue: ~$10.7B
- Implied non-AI semiconductor + software: ~$11.3B
- If software stays flat: ~$6.8B software + ~$4.5B non-AI semi
The AI business is now large enough that it sets the pace for the entire company. At $10.7 billion in Q2, the AI semiconductor run rate annualizes to roughly $42-43 billion in a company that generated $63.9 billion in total revenue for all of FY2025.
The 47% year-over-year growth guidance for Q2 is the highest quarterly growth rate Broadcom has guided in recent memory. Hock Tan said: "Our AI revenue growth is accelerating." He did not hedge that.
Balance sheet: the debt is manageable, not comfortable
| Item | Feb 1, 2026 | Nov 2, 2025 |
|---|---|---|
| Cash and equivalents | $14.174B | $16.178B |
| Short-term debt | $2.252B | $3.152B |
| Long-term debt | $63.805B | $61.984B |
| Total debt | ~$66.057B | ~$65.136B |
| Net debt | ~$51.883B | ~$48.958B |
| Goodwill | $97.801B | $97.801B |
| Intangible assets, net | $30.302B | $32.273B |
Cash fell from $16.178 billion to $14.174 billion in Q1. The reason is visible in the cash flow statement: Broadcom returned $10.9 billion to shareholders in the quarter $3.086 billion in dividends and $7.850 billion in stock repurchases while also refinancing debt (raised $4.474 billion in new borrowings, repaid $3.650 billion).
Net debt of $51.883 billion against a Q1 annualized FCF run rate of approximately $32 billion puts net debt at roughly 1.6x annualized FCF. That ratio has been declining since the VMware close in November 2023 when it was significantly higher.
The $97.801 billion in goodwill sits unchanged quarter over quarter it will stay there until something triggers an impairment test. Goodwill impairment flows directly through the income statement. It doesn't touch cash flow, but it goes straight through the income statement if something triggers an impairment test. Worth keeping an eye on. The goodwill balance is the accumulated premium Broadcom paid above book value across its acquisition history primarily VMware. If the software segment underperforms what Broadcom paid for it, that number becomes a problem.
The $30.302 billion in net intangible assets is declining at roughly $2 billion per quarter as acquisition intangibles amortize. That amortization shows up as a non-cash charge in the income statement, widening the GAAP vs non-GAAP gap. By FY2029-2030, a meaningful portion of the VMware intangible amortization will be fully written off, which would make GAAP earnings look substantially better without any change in the underlying business.
Shareholder returns: $10.9 billion returned in one quarter
Broadcom returned $10.9 billion to shareholders in Q1 FY2026 alone.
- Cash dividends: $3.086 billion ($0.65 per share quarterly)
- Stock repurchases: $7.850 billion
For context, $10.9 billion in shareholder returns in one quarter is more than many S&P 500 companies generate in total annual free cash flow. Broadcom's FCF in the same quarter was $8.010 billion meaning it returned more cash than it generated in the quarter, funding the difference by drawing down cash and refinancing debt.
The Board also authorized a new $10 billion share repurchase program through December 31, 2026. That authorization sits on top of any amounts already remaining from prior programs.
The $0.65 quarterly dividend translates to $2.60 annualized per share. At a stock price of approximately $300 (as of late March 2026), that is roughly a 0.9% dividend yield. The dividend yields under 1% at current prices. The buybacks are where the capital return actually lives.
What this report means for the XPU thesis
Hock Tan disclosed in December 2024 that three hyperscaler customers are developing next-generation XPUs with Broadcom, with the serviceable addressable market from those three customers alone reaching $60-90 billion by FY2027 assuming each deploys a one-million-chip cluster.
Q1 FY2026's $8.4 billion in AI revenue is the progress report against that thesis.
8.4 billion in one quarter is roughly 10% of the low end of the $60-90 billion three-year SAM estimate. The pace is consistent with the trajectory Hock Tan outlined. Q2's guided $10.7 billion would push cumulative H1 AI revenue to $19.1 billion roughly 21-32% of the cited three-year window in just two quarters.
Demand is visible. The variable is timing XPU cluster deployments slip, revenue recognition slips with XPU clusters take years to design and get into production. Revenue recognition follows physical deployment. If a hyperscaler delays a cluster buildout for budget reasons, architecture changes, or supply chain issues Broadcom's AI revenue quarter can shift. The $60-90 billion SAM number is a useful frame for the size of the opportunity, but the quarterly pacing is genuinely hard to model from the outside.
Current performance links
For how the macro environment hyperscaler capex cycles, interest rates, and geopolitical risk affects Broadcom's near-term outlook:
How macro conditions affect Broadcom : hyperscaler capex, interest rates, and the VMware renewal cycle
For why NVIDIA and Broadcom are not competing for the same decision, and how the XPU model creates architectural lock-in:
Broadcom's $51.6B revenue splits cleanly into two businesses custom silicon and enterprise software
The number I cannot stop thinking about in this report is not $8.4 billion in AI revenue. It is $92 million.
That is how much infrastructure software revenue grew year over year. $6.796 billion in Q1 FY2026 versus $6.704 billion in Q1 FY2025. Ninety-two million dollars of growth on a nearly $7 billion base, in a quarter where the overall company grew 29%. That is 1.4% growth against a comparison period where Broadcom already owned VMware, had already raised prices, had already cut the product portfolio. This is the first clean read on the VMware retention thesis, and it is... not convincing yet.
To be fair: I am not ready to call the VMware bet broken on one quarter of data. Subscription conversion timing is real. Multi-year contract deferrals are real. And Broadcom has not provided enough granular data on bookings or backlog to know whether the 1% in-quarter growth reflects weak demand or just revenue recognition timing. It genuinely could be either.
But here is the thing that concerns me more than the 1% number itself. Broadcom raised prices dramatically on VMware customers after the acquisition. If you raise prices on retained customers and your revenue still only grows 1%, either a significant number of customers left, or the subscription timing effects are masking a better underlying story. Both interpretations are open right now. FY2026 full-year infrastructure software revenue will close that debate.
The semiconductor side is a different story entirely. $8.4 billion in AI revenue growing 106% year over year and management guiding $10.7 billion in Q2 is the kind of acceleration that changes what this company is. At the Q2 pace, Broadcom's AI semiconductor business alone would be a top-30 S&P 500 company by revenue if it were standalone. Run the math and that's where it lands. 106% growth compounded does that.
The FY2026 full-year infrastructure software revenue growth rate is the number to track. My published stance is that above 5% growth confirms the VMware switching cost assumption is holding at scale. Above 10% confirms the subscription conversion is working as designed.
Q1 at 1.4% is one quarter. But it sets up Q2 as the more telling data point. If Q2 infrastructure software is also at or below 3-4% year over year, the full-year path to 5% requires a significant acceleration in the second half that would need explaining.
Two quarters of sub-5% growth would be the signal I take seriously.
On the AI side, the signal I am watching is whether the three named hyperscaler XPU customers are actually deploying clusters on the timeline Hock Tan described in December 2024. Q1's $8.4 billion is consistent with the trajectory. Q2's $10.7 billion guidance is consistent with acceleration. If AI revenue stalls at $9-10 billion per quarter for multiple quarters rather than continuing to grow that would suggest deployment timelines are slipping and the $60-90 billion SAM is farther out than the 2027 framing implied.
Both signals are worth watching. Right now, only one of them is showing up in the numbers.
FAQs
What was Broadcom's revenue in Q1 FY2026?
Broadcom reported Q1 FY2026 revenue of $19.311 billion, up 29% year over year from $14.916 billion in Q1 FY2025. The quarter ended February 1, 2026, and results were reported on March 4, 2026.
How much was Broadcom's AI revenue in Q1 FY2026?
Broadcom's Q1 FY2026 AI revenue was $8.4 billion, up 106% year over year. This includes revenue from custom AI accelerators (XPUs) and AI networking silicon. Hock Tan confirmed this came in above the company's own forecast. For Q2 FY2026, Broadcom guided AI semiconductor revenue at approximately $10.7 billion, which would be a 27% sequential increase from Q1.
What happened to Broadcom's infrastructure software revenue in Q1 FY2026?
Infrastructure software primarily VMware generated $6.796 billion in Q1 FY2026, up 1.4% year over year from $6.704 billion in Q1 FY2025. The Q1 FY2025 comparison already reflected Broadcom's ownership of VMware including post-acquisition price increases and subscription conversion. The 1.4% growth rate is the first clean apples-to-apples read on whether VMware customer retention is holding under the new pricing model.
What is Broadcom's Q2 FY2026 guidance?
Broadcom guided Q2 FY2026 revenue at approximately $22.0 billion, a 47% increase year over year and roughly 14% sequential growth from Q1's $19.3 billion. Adjusted EBITDA margin is guided at approximately 68% of projected revenue. AI semiconductor revenue is expected to reach $10.7 billion in Q2.
What is Broadcom's free cash flow margin?
Broadcom generated $8.010 billion in free cash flow in Q1 FY2026, representing 41% of revenue. Cash from operations was $8.260 billion and capital expenditures were $250 million (less than 1.3% of revenue). The low capex reflects Broadcom's fabless model TSMC handles semiconductor manufacturing, so Broadcom does not carry fab capital costs.
Why is Broadcom's GAAP net income so different from non-GAAP?
The $2.836 billion gap between GAAP net income ($7.349B) and non-GAAP net income ($10.185B) in Q1 FY2026 comes from three main items. Amortization of acquisition-related intangible assets was $1.969 billion this is the cost of VMware and prior acquisitions spreading through the income statement over time. Stock-based compensation expense was $2.176 billion. Restructuring and other charges were $116 million. These items are real costs but do not represent cash leaving the business in the quarter they are recognized, which is why FCF of $8.010 billion is a more direct measure of cash generation.
What does Broadcom's $97.8 billion goodwill balance mean?
Goodwill of $97.801 billion is the accumulated premium Broadcom paid above book value across its acquisition history primarily the $69 billion VMware acquisition. Goodwill does not amortize automatically; it is tested for impairment. If acquired businesses underperform the value Broadcom paid for them, goodwill impairment flows directly through the income statement. It is not a cash charge but it affects reported net income. The goodwill balance is essentially unchanged quarter over quarter, meaning no impairment trigger has been reached.
What is an XPU and how does Broadcom make money from it?
An XPU is a custom AI accelerator designed for one company's specific workload rather than a general-purpose GPU sold at market price to any buyer. Google's TPU and Meta's MTIA are examples where Broadcom has had design involvement. Hyperscalers build custom chips because a workload-specific design produces better cost per computation at the scale they operate. Broadcom provides the chip design engineering resources. Once a hyperscaler deploys a cluster built around a Broadcom-designed chip, switching to a different design partner requires rebuilding the cluster architecture a 3-5 year process minimum. That architectural lock-in makes XPU revenue sticky in a way most semiconductor revenue is not.
How much did Broadcom return to shareholders in Q1 FY2026?
Broadcom returned $10.9 billion to shareholders in Q1 FY2026: $3.086 billion in cash dividends ($0.65 per share) and $7.850 billion in stock repurchases. The Board also authorized a new $10 billion share repurchase program through December 31, 2026. The $10.9 billion in returns exceeded Q1's free cash flow of $8.010 billion Broadcom funded the difference by drawing down cash and refinancing debt.
What is Broadcom's net debt position?
As of February 1, 2026, Broadcom had cash and equivalents of $14.174 billion against total debt of approximately $66.057 billion (short-term $2.252B plus long-term $63.805B), resulting in net debt of approximately $51.883 billion. Against a Q1 annualized FCF run rate of roughly $32 billion, net debt is approximately 1.6x FCF. The ratio has been declining since the VMware acquisition close in November 2023.