Broadcom Q3 FY2026 Earnings — The Numbers Are Already Guided; the Print Is Whether FY2027 AI Gets Raised
The Call
The quarter itself is close to arithmetic: Broadcom has guided total revenue ($29.4B) and AI-semiconductor revenue ($16.0B) precisely and has narrowly beaten both for two consecutive quarters, so a modest beat on each is the base case, not a genuine unknown. What actually moves the stock is the same binary that drove the -13.6% after-hours reaction at Q2 — whether FY2027 AI semiconductor guidance finally gets RAISED above the reaffirmed '>$100B' floor — now complicated by two new diversification-away developments (MediaTek's exclusive TPU v9 upgrade win, Marvell's $120B Google warrant deal) that give management fresh reason to stay conservative even as the quarter itself beats.
| Metric | Our Estimate | Company Guidance | Consensus |
|---|---|---|---|
| RevenueGAAP +34.1% QoQ, +86.5% YoY vs Q3 FY2025's $15,952M. Guide issued 2026-06-03; consensus sits almost exactly on the guide. | 29,750 USD M | 29,400 USD M | 29,435 USD M |
| Gross profitGAAP Revenue x GAAP gross margin estimate, itself derived from the non-GAAP GM guide via the historical GAAP/non-GAAP cost-of-revenue bridge. | 20,236 USD M | — | — |
| Gross marginGAAP Derived from the ~74% non-GAAP guide less the anchored COGS-side bridge ($1,779M). Roughly flat to slightly down vs Q2's 69.5% GAAP GM — the mix-dilution effect shows up at similar magnitude on both bases. | 68.0% | — | — |
| Operating profitGAAP Derived from the ~67% non-GAAP operating-margin guide less the anchored COGS-side and opex-side bridges ($1,779M + $2,540M). | 15,614 USD M | — | — |
| Operating marginGAAP | 52.5% | — | — |
| Net incomeGAAP GAAP operating profit less anchored net interest expense ($850M) and a ~6% GAAP effective tax rate, both back-solved from the Q2 FY2026 print. | 13,878 USD M | — | — |
| Diluted EPSGAAP Net income / ~4,870M estimated diluted shares. No reliable third-party GAAP-EPS-specific consensus was found to cross-check against — the tracked Street consensus for AVGO (yfinance/earnings-pages) is itself non-GAAP; see the non-GAAP diluted EPS businessMetric ($3.37 est vs $3.24 consensus) for the figure the market actually prices the print against. This is the largest single-line confidence gap in the ladder — see breaksIf. | 2.85 USD | — | — |
- → AI semiconductor revenue lands at or below the $16.0B guide (rather than above it) — the clearest signal that CoWoS-allocation share is eroding toward MediaTek/Marvell faster than the base case assumes, or that a contracted gigawatt delivery (Meta, OpenAI, or Anthropic) is slipping.
- → FY2027 AI semiconductor guidance is reaffirmed at '>$100B' again rather than raised — a repeat of the exact Q2 dynamic that drove the -13.6% after-hours reaction, now with two new competitive-diversification developments (MediaTek's exclusive Triggerfish win, Marvell's $120B Google warrant deal) giving management fresh reason for continued caution on the multi-year number even if the quarter itself beats.
- → Non-GAAP gross margin prints below ~74% — would mean AI-hardware mix dilution is running ahead of the guide, compressing both the non-GAAP and GAAP net-income lines below this call even if revenue is in line.
- → The 10-Q or call quantifies the AI XPV Apollo/Blackstone residual-value guarantee at a materially larger figure than the disclosed ~$29B initial exposure, or confirms the reported $60-100B second-transaction structure — a balance-sheet/counterparty-concentration risk the framework only started pricing on 2026-08-14 that the market has likely not yet fully absorbed.
- → Infrastructure Software revenue lands far above or far below its recent ~9% YoY trend — either outcome means the residual math behind the $13,450M non-AI-semi-plus-software estimate is wrong, since that estimate rests on an assumed Apple-seasonal skew rather than a software re-acceleration.
- → Management discloses VMware renewal-cohort churn above 10% ahead of the late-2026/2027 renewal cliff — the largest un-stress-tested risk in the framework, given ~30,000 cumulative Nutanix migrations already disclosed.
Behaviour forecast, not arithmetic, held to lower confidence than every line above. The general up-cycle lesson already in this pipeline (STX FQ1 FY2027 -9.3%, AMD Q3 2026 -3.1%, and NVDA's own FQ2 FY2027 guide today built deliberately above a mechanical step) argues for expecting management to signal acceleration rather than repeat caution. But AVGO-specific evidence cuts the other way since the Q2 print: MediaTek's exclusive win on the upgraded 'Triggerfish' TPU v9 variant (confirmed 2026-06-22) and Marvell's $120B cumulative-revenue warrant deal with Google (signed 2026-07-29) are both new diversification-away-from-Broadcom data points that did not exist when Broadcom chose to reaffirm rather than raise at Q2 — giving management a fresh, concrete reason to stay conservative on the specific FY2027 number even as the quarter itself likely beats. The two effects pull in opposite directions, so this call sits closer to a coin flip than the confident beats posted elsewhere in this framework.
Evidence (10)
Positioning Suggestions
→ The framework's own anchor ($368.45, 2026-08-21) already sits below the bear-case multiple (~23.0x spot vs. 24x bear-case FY2027E) — much of the sell-the-news risk from a Q2-repeat (beat-but-reaffirm) outcome is already priced, so treat a reaffirm-not-raise print as a potential entry point rather than an automatic reason to trim.
→ Watch MRVL's own Aug 27, 2026 print (six days before AVGO's) for incremental Google-TPU-program commentary — a Google-specific update from Marvell could re-price AVGO ahead of its own print.
→ Avoid sizing up aggressively into the print until the 10-Q's specific AI XPV residual-value-guarantee disclosure is read; it is the single most-underpriced new risk in the framework and is not resolved by the operating-line beat/miss.
Predictions
8AVGO Q3 FY2026 AI semiconductor revenue exceeds the $16.0B guide
A clear beat (>$16.3B) alongside a raised FY2027 number is the strongest incremental-buy signal available this cycle.
AVGO raises FY2027 AI semiconductor revenue guidance to an explicit figure above $105B (not merely reaffirmed at '>$100B')
If reaffirmed again rather than raised, expect a near-term negative reaction; given the framework's spot multiple already sits below the bear case, treat that reaction as a add opportunity rather than a reason to trim.
AVGO Q3 non-GAAP gross margin prints at or above the ~74% guide
AVGO's 10-Q or earnings call quantifies new detail on the AI XPV Apollo/Blackstone residual-value guarantee beyond the disclosed ~$29B initial exposure
AVGO Infrastructure Software (VMware) revenue grows above 12% YoY in Q3 FY2026
Management names an additional XPU customer beyond the existing six programs plus the two disclosed ~$6B customers
A new named customer would be the cleanest incremental-bull surprise in the print — consider adding on confirmation.
AVGO stock closes lower the trading day after the print, even if AI semiconductor revenue beats $16.0B
Management directly addresses the MediaTek Triggerfish / Google TPU v9 diversification reporting on the call
Key Questions
- Does FY2027 AI semiconductor guidance get raised above the reaffirmed '>$100B' floor, or reaffirmed again as at Q2?
- Does the ~74% non-GAAP gross-margin guide hold, and does management call the AI-hardware mix dilution transient or ongoing?
- Does the 10-Q or call quantify the AI XPV Apollo/Blackstone residual-value guarantee beyond the disclosed ~$29B exposure, or address the reported $60-100B second-transaction structure?
- Does management address the MediaTek Triggerfish / Google TPU v9 diversification news, and does Broadcom's own training-class design assignment stay intact?
- Does Infrastructure Software show any early renewal-cohort churn signal ahead of the late-2026/2027 VMware renewal cliff?