Broadcom Market Analysis
Broadcom Raises Its 2027 AI Outlook to $115 Billion — While Google Diversifies and EU Scrutiny Advances
Last Updated: September 8, 2026
- Fundamental Momentum: Strong — Q3 delivered record revenue, operating profit and free cash flow, with AI semiconductor growth remaining the dominant operating driver
- Evidence Balance: Mixed-Positive — reported AI and cash-flow evidence strengthened, while Google supplier diversification, a slightly softer-than-consensus Q4 revenue guide and advancing EU VMware scrutiny keep the outlook two-sided
- Evidence Strength: High — the core financial evidence is supported by Broadcom and SEC filings, while the two principal counter-risks are supported by Marvell’s 8-K and an EU General Court order
- Risk Level: Medium — the main risks are execution against a much larger 2027–2028 AI ramp, customer and supplier concentration, competition in custom silicon, and potential VMware licensing remedies in Europe
- Time Horizon: 3–15 months — the next hard checkpoint is Q4 FY2026, followed by evidence that the 2027 AI deployment and supply plan is converting into reported revenue
- Key Catalyst: Q4 AI semiconductor delivery and the first evidence that secured 2027 capacity is being absorbed on customer deployment schedules
- Thesis Evidence: Strengthening — the prior Q3 operating test was passed, but the stronger thesis now depends on a substantially higher long-duration execution bar
Key Questions
Did Broadcom pass the $16 billion Q3 AI test from the prior analysis?
Yes, although the scale of the beat should be kept in proportion. Broadcom reported $16.7 billion of Q3 FY2026 AI semiconductor revenue, up 221% year over year and 54% sequentially. That cleared the $16 billion threshold highlighted in the prior analysis by about 4.4%. The quarter therefore validates the direction of the prior AI-growth thesis, but the more important evidence is not a $700 million beat by itself. Broadcom immediately raised the next operating bar: management now expects AI semiconductor revenue to reach $21.7 billion in Q4, roughly 30% above Q3 on a sequential basis.
The Q3 test was passed, but the thesis did not become easier to evaluate.
Broadcom converted the prior $16 billion checkpoint into reported revenue, then moved the debate outward to whether a much larger 2027–2028 AI ramp can be supplied, deployed and monetized on schedule.
What changed in the 2027–2028 AI outlook, and how much of it is already proven?
On the September 2 earnings call, CEO Hock Tan raised Broadcom’s fiscal 2027 AI semiconductor revenue outlook to approximately $115 billion, from the prior formulation of more than $100 billion, and said the company has line of sight to approximately $230 billion in fiscal 2028. Reuters reported that Broadcom has secured supply to support the 2027 ramp and that demand currently exceeds the company’s outlook. Those are material changes because they extend the growth thesis beyond the next quarter and into a multi-year infrastructure cycle.
The distinction between evidence and forecast is critical. Q3 revenue and free cash flow are reported facts. The 2027 and 2028 figures are management outlooks that depend on customer data centers being built, powered and brought online, on advanced packaging, wafers and memory arriving as planned, and on Broadcom retaining enough design scope as customers diversify suppliers. The stronger interpretation is therefore not that $230 billion is already secured revenue; it is that Broadcom has presented a much larger, supply-backed roadmap whose credibility now has to be tested quarter by quarter.
Does Google’s expanded Marvell deal weaken Broadcom’s custom-silicon position?
It strengthens the countercase without proving displacement. Marvell disclosed in an August 19 8-K that it entered an expanded commercial agreement with Google covering custom semiconductor products attached to the TPU ecosystem, including AI inference accelerators, storage controllers, network-interface controllers, memory-interface controllers and near-memory compute. Google also received a warrant for up to 58.97 million Marvell shares, with most vesting tied to future custom-product purchases.
This is directly relevant because Google supplier diversification was one of the variables identified in the prior Broadcom analysis. Reuters reported that analysts generally viewed the Marvell agreement as an expansion of Google’s supplier pool rather than a direct replacement of Broadcom. That is the more defensible reading today: the deal confirms that Google is willing to allocate meaningful custom-silicon work outside Broadcom, but it does not demonstrate that Broadcom has lost its existing TPU-related role. The analytical test is now narrower — whether Marvell remains additive around the TPU ecosystem or begins taking materially larger scope from Broadcom over subsequent design generations.
Did the EU VMware risk remain only a trade-group complaint?
No. The prior article treated the July pressure from European cloud groups as several steps removed from formal regulatory action. That description is now stale and should be corrected. On August 3, the EU General Court rejected Broadcom and VMware’s request for interim relief in Case T-280/26 R, which concerns a European Commission request for information under Article 18(3) of Regulation 1/2003. Reuters reported that Broadcom had sought to suspend the Commission’s demand for U.S. legal documents related to VMware and that the court rejected the challenge.
This does not establish a violation, fine or remedy. It does establish that the risk is no longer only an external complaint asking regulators to act: the Commission is using formal evidence-gathering powers, and Broadcom is already litigating the scope of those requests. At the same time, Broadcom’s Q3 infrastructure software revenue still grew 29% year over year, so there is not yet reported evidence that the European dispute has impaired the segment’s aggregate revenue trajectory.
Since our last update, at a glance:
- Q3 AI semiconductor revenue cleared the prior $16 billion checkpoint and continued rapid sequential growth
- Q3 consolidated revenue, operating profit and free cash flow reached company records
- Q4 AI semiconductor guidance was raised again, while total Q4 revenue guidance came in slightly below the Reuters-reported analyst consensus
- Management raised its fiscal 2027 AI semiconductor outlook to about $115 billion and described a path to roughly $230 billion in fiscal 2028
- Google materially expanded its custom-silicon relationship with Marvell, confirming that supplier diversification is real even though displacement of Broadcom has not been established
- EU VMware scrutiny progressed from external industry pressure to a Commission information request already tested in the EU General Court
- Infrastructure software still posted strong reported growth, so the regulatory risk has not yet translated into visible segment-level deterioration
Previous Thesis Check: Did the prior framework survive the next evidence cycle?
Prior test: the previous Broadcom analysis identified Q3 AI semiconductor revenue versus the $16 billion guide as the principal operating checkpoint, with Google chip-supplier diversification and EU VMware scrutiny as important counter-risks.
Observed since then: the Q3 AI threshold was exceeded; Broadcom then raised its longer-dated AI outlook. Separately, Google expanded its Marvell relationship, and the EU competition process advanced to a formal Commission information request that Broadcom challenged in court.
Assessment: supported and strengthened on the core AI operating mechanism, but with stronger counterevidence. The revenue thesis survived its hard Q3 test. The prior interpretation of EU risk as mainly a complaint-stage issue did not survive and is corrected here. Google diversification also moved from a monitoring possibility to a confirmed commercial fact, although current evidence does not show that Marvell has replaced Broadcom in Google’s core custom-AI relationship.
Methodology note: this is not a forecast hit rate or an investment-return track record. It records whether previously stated analytical mechanisms, catalysts and falsification tests remained consistent with later verified evidence.
Key Facts
Q3 FY2026 Results — Reported September 2
- Total net revenue: $29.591 billion, up 86% year over year
- GAAP operating income: $15.955 billion; non-GAAP operating income: $20.095 billion
- GAAP diluted EPS: $2.68; non-GAAP diluted EPS: $3.32
- Cash from operations: $14.197 billion; capital expenditures: $532 million
- Free cash flow: $13.665 billion, equal to 46% of revenue
- Semiconductor Solutions revenue: $20.839 billion, up 127% year over year and representing 70% of consolidated revenue
- Infrastructure Software revenue: $8.752 billion, up 29% year over year and representing 30% of consolidated revenue
AI and Q4 Outlook
- Q3 AI semiconductor revenue: $16.7 billion, up 221% year over year and 54% quarter over quarter
- Q4 AI semiconductor revenue expectation: $21.7 billion, up 236% year over year
- Q4 consolidated revenue guidance: approximately $34.8 billion, up 93% year over year
- Q4 non-GAAP operating income guidance: approximately 66% of projected revenue
- Reuters-reported earnings-call outlook: approximately $115 billion of AI semiconductor revenue in fiscal 2027 and approximately $230 billion in fiscal 2028; these are forward-looking management estimates, not reported revenue
Competition, Software and Regulation
- Marvell and Google entered an expanded custom-semiconductor agreement on July 29, disclosed in Marvell’s August 19 8-K; the covered programs attach to the TPU ecosystem and include AI inference, storage, networking and memory-related silicon
- Google received a warrant for up to 58,970,907 Marvell shares at an exercise price of $206.58, with most vesting tied to future custom-product purchases
- Reuters reported that analysts viewed the Marvell expansion as supplier diversification rather than evidence that Google had replaced Broadcom
- On August 3, the EU General Court issued an order in Broadcom and VMware International v Commission, Case T-280/26 R, concerning a Commission Article 18(3) information request; Broadcom did not obtain the requested suspension
- Broadcom introduced VMware Private AI Cloud and VMware AI Factory on August 31; these launches expand the software product roadmap but do not yet establish incremental financial contribution
The Harder Test Moved From One Quarter to a Multi-Year Deployment Cycle
The prior article was built around a relatively clean question: could Broadcom deliver approximately $16 billion of AI semiconductor revenue in Q3? It did. The analytical problem is now less binary. Broadcom’s reported Q3 numbers demonstrate that custom accelerators and networking are already converting into very large revenue and cash flow. The new 2027–2028 outlook, however, pushes the thesis into a phase where physical deployment becomes as important as design wins. Power availability, data-center construction, advanced packaging, memory supply and customer timing can all delay revenue even when semiconductor demand exists on paper.
The same evidence also argues against a one-sided interpretation. Google’s Marvell agreement proves that a major customer is actively broadening its supplier base, while the EU VMware matter has progressed beyond the complaint-stage framing in the previous article. Neither development breaks Broadcom’s current operating momentum. Together, they make the future thesis more dependent on execution and competitive share than the Q3 headline growth rate alone suggests.
Valuation Context
Broadcom’s valuation context is increasingly a duration question rather than a simple current-quarter multiple question. The strongest evidence supporting a premium valuation is now reported: Q3 produced record revenue, profit and free cash flow, and AI semiconductor revenue continued to scale rapidly. The harder part of the valuation case is that a growing share of the narrative depends on management’s 2027 and 2028 revenue outlook, which is necessarily more sensitive to deployment timing, supply availability and competitive share than the already-reported Q3 results.
The post-earnings share-price reaction reinforces that distinction. Reuters reported that Broadcom’s $34.8 billion Q4 revenue guide was slightly below the analyst consensus it cited, and shares fell 2.7% in the following session despite the Q3 beat and raised long-term AI outlook. That reaction is not evidence that the business thesis failed; it is evidence that expectations were already demanding. This analysis does not set a proprietary Broadcom share-price target, and third-party targets are not used here as a substitute for testing the underlying assumptions.
Current Market Data
Broadcom trades on Nasdaq under ticker AVGO. At 2:49 p.m. ET on September 8, 2026, Charles Schwab’s public market-data page showed AVGO at approximately $368.34, with a 52-week range of $289.96–$495.00 and market capitalization near $1.7 trillion. This is a dated intraday observation rather than a permanent price reference; the live TradingView chart below may reflect newer movement after this written analysis was completed.
Scenario Analysis
Constructive
Q4 AI semiconductor revenue meets or exceeds the company’s current guide, customer data-center deployments stay on schedule, and the supply Broadcom says it has secured converts into a visible 2027 revenue ramp. Google’s Marvell relationship remains additive around the TPU ecosystem rather than displacing Broadcom’s highest-value custom-accelerator and networking scope, while VMware software growth remains resilient without material EU remedies.
Central
AI semiconductor revenue continues to grow rapidly but with quarter-to-quarter lumpiness as customers sequence capacity, power and system deployments. Marvell wins meaningful adjacent Google programs without materially displacing Broadcom’s existing role, and the EU VMware process remains an active legal and regulatory overhang rather than a quantified financial impairment. Infrastructure software continues to grow, but at a much slower rate than semiconductors.
Adverse
Customer infrastructure delays or supply bottlenecks cause the 2027 AI ramp to fall materially behind management’s current outlook, while Google or other hyperscalers allocate a larger share of future custom-silicon work to competing suppliers. At the same time, EU proceedings produce interim measures or licensing changes that reduce VMware economics, creating pressure on both sides of Broadcom’s semiconductor-and-software model.
Unlock the MatrixPro24 Analytical View
Create a free MatrixPro24 account to access our complete Analytical View, including scenario analysis, monitored catalysts, and a full analytical assessment.
Sources
- Broadcom — Third Quarter Fiscal Year 2026 Financial Results — September 2, 2026
- SEC — Broadcom Form 8-K — September 2, 2026
- Reuters — Broadcom raises AI chip forecast as Big Tech keeps writing bigger checks — September 2, 2026
- Marvell — Form 8-K describing expanded Google custom-silicon agreement — August 19, 2026
- Reuters — Marvell gives Google option to buy stake in custom AI chip deal — August 19, 2026
- General Court of the European Union — Broadcom and VMware International v Commission, T-280/26 R — August 3, 2026
- Reuters — Broadcom loses court bid over EU antitrust information request — August 3, 2026
- Broadcom — VMware Private AI Cloud announcement — August 31, 2026
- Charles Schwab — AVGO market data — September 8, 2026 observation
About MP24 Analyst X
Published by MP24 Analyst X. Read our Editorial and Content Policy for the publication’s research, verification and editorial-accountability framework.
MP24 Analyst X is the public-facing pseudonym used for research and editorial work. The publication process emphasizes source verification, evidence separation, falsification and transparent monitoring without presenting the byline as a disclosed credentialed identity.
Disclaimer
This analysis is for informational and educational purposes only and does not constitute personalized financial or investment advice or a recommendation to buy, sell, or hold any financial instrument. Market observations are dated where relevant, and the live TradingView chart may reflect newer price movement. Past performance is not indicative of future results.
