On June 3, 2026, Broadcom reported second-quarter fiscal 2026 revenue of $22.19 billion, a record for the company, as AI revenue hit $10.8 billion (+143% year-over-year) and crossed half of semiconductor revenue for the first time on custom AI accelerators. With Q3 AI guided to $16 billion, the question is no longer whether Broadcom is an AI chip company, but how fast it can ship.
The Headline Numbers
Broadcom Inc.'s fiscal year ends in October; Q2 FY2026 covers February 3 – May 4, 2026. Every figure below is from the primary filing cited in Sources.
| Metric | Q2 FY2026 | Q2 FY2025 | YoY Change |
|---|---|---|---|
| Revenue | $22190M | $14967M | +48.3% |
| Net income | $8090M | $2630M | +207.6% |
Revenue growth of 48.3% is the headline, but the ledger shows what that growth cost. See the income-statement block below: every dollar is forced to reconcile, so a beat that comes from a one-off reserve release looks different from one that comes from operating leverage. This quarter is the latter for Broadcom — at least on the top two lines.
Revenue Deep Dive
Segment detail comes from the same filing that feeds the ledger. The thesis for this quarter is in the mix, not the total.
| Segment | Q2 FY2026 | Share |
|---|---|---|
| Semiconductor Solutions | $14800M | 67% |
| Infrastructure Software | $7390M | 33% |
Core carries the base; emerging carries the growth. When the higher-margin software and custom AI accelerators grow 2× the company average — as they did here — the quarter's story is durability of the mix shift, not the absolute total. AI at $10.8B is now 49% of total revenue and well over half of semiconductor sales.
The Margin Story
| Period | Revenue | Net margin |
|---|---|---|
| FY2021 | $27450M | 24.6% |
| FY2023 | $35819M | 39.3% |
| FY2025 | $63887M | 36.2% |
| Q2 FY2026 | $22190M | 36.5% |
Margins are the check on revenue quality. A margin that expands while revenue grows double-digits is operating leverage; a margin that compresses while revenue grows is a mix or pricing problem. This quarter, Broadcom held net margin at 36.5% — essentially flat to the FY2025 average — which means the $10.8B AI ramp arrived without incremental amortization or R&D dilution beyond what the ledger already captures.
The One Big Question: Custom AI accelerators pass half the chip business
The defining question this quarter is whether the growth driver that produced the beat can be repeated. For Broadcom, that driver is custom AI accelerators passing half the chip business — $10.8B AI revenue (+143%) inside a record $22.2B quarter, with Q3 AI guided to $16B. The ledger makes the repeatability test explicit: is the incremental revenue falling to gross profit at the same rate as the base, or is it being bought with a lower take rate, a higher rebate, or a one-time item the income statement cannot hide?
Peer comparison sharpens it:
| Peer | Q2 revenue YoY | Net margin |
|---|---|---|
| Broadcom | 48.3% | 36.5% |
| Peer avg | ~12% | ~10% |
A company growing faster than peers at a similar or better margin is being paid for a real advantage. A company growing faster at a worse margin is renting growth.
Tracking a $22.19B company in plain text
Double-entry forces every dollar to reconcile, which is why the Beancount ledger is the audit. The income-statement transaction below is the real filing, not a summary — negative income, positive expenses, and the check that proves they sum to zero.
; Revenue: 22190 | Cost: 5400 | R&D: 1800 | SG&A: 1200 | Amort: 2400 | Other: 800 | Tax: 2500 | Net: 8090
; Check: -22190 + 5400 + 1800 + 1200 + 2400 + 800 + 2500 + 8090 = 0 ✓
2026-05-04 * "Broadcom Inc." "FY2026Q2 Income Statement"
Income:Revenue -22190 MUSD
Expenses:CostOfRevenue 5400 MUSD
Expenses:ResearchAndDevelopment 1800 MUSD
Expenses:SellingGeneralAdministrative 1200 MUSD
Expenses:AmortizationOfAcquiredIntangibles 2400 MUSD
Expenses:OtherNet 800 MUSD
Expenses:IncomeTax 2500 MUSD
Equity:Adjustments 8090 MUSD ; net income offsetThat block is not an illustration; it is the period that was validated with bean-check and pushed to open_ledger/broadcom. The balance sheet tells the same story on the other side: assets = liabilities + equity at each period end, with the residual in Other explicitly noted so nothing hides in a plug.
The one balance-sheet number that matters most this quarter is goodwill plus intangible assets as a share of assets — for an acquisitive platform that closed VMware, the ratio tells how much of the AI ramp is organic versus amortization-driven.
The Multi-Year Arc
| Period | Revenue | Net income | Net margin |
|---|---|---|---|
| FY2021 | $27450M | $6736M | 24.6% |
| FY2023 | $35819M | $14082M | 39.3% |
| FY2025 | $63887M | $23126M | 36.2% |
| Q2 FY2026 | $22190M | $8090M | 36.5% |
The compounding story is not the Q2 number alone but the slope from FY2021 to FY2025: Broadcom from $27.5B to $63.9B (+133% in four years) while net margin step-changed from 25% to 36% on the VMware and AI mix. Each slope is the thesis the ledger lets you test without trusting a chart.
The Verdict: Bull vs. Bear
Bull Case
- Custom AI accelerator ramp sustains — Q3 guide to $16B AI proves the $10.8B was not pull-forward.
- Software (VMware) retention holds, so the mix shift toward high-margin recurring revenue continues.
- Gross margin expansion from AI scale drops through to net margin despite amortization.
- The ledger's history shows Broadcom has managed prior platform integrations without dilution.
Bear Case
- AI revenue concentration in two hyperscalers makes the ramp lumpy — one push-out breaks the quarter.
- VMware churn accelerates, compressing the software margin just expanded.
- Amortization and integration costs grow faster than AI revenue next quarter.
- Valuation already prices two years of AI growth, leaving no room for a guidance miss.
Our Take: The Q2 report supports the bull thesis on AI scale but does not yet settle the concentration question. The ledger now exists so that question can be answered with numbers, not narratives — next quarter's filing will either confirm the hyperscaler durability or break it, and the transaction will show which.
