Broadcom: A Value Investing Deep Dive

Moat

The table below sorts the moat into three categories: "present / partly present / weak." Here I stress in particular: facts, inferences and opinions must be kept separate. Facts come from financial statements, products and news; "whether the moat is durable" is inference and opinion.

Supporting evidence. As of FY2025 the company held about 19,000 patents and about 2,170 patents pending; FY2025 R&D spend was about 11 billion dollars; of the roughly 33,000 employees in 2025, about 57% were in R&D roles, with a global voluntary attrition rate of about 4.1%. These figures show that Broadcom's moat comes mainly not from brand advertising but from IP, engineering capability, years of collaboration with key customers, and embedding in mission-critical infrastructure. Reuters further shows that in AI custom accelerators and Ethernet interconnect, Broadcom has formed deep partnerships with Google, Meta, Anthropic, OpenAI and others, and claims visibility into more than 100 billion dollars of AI chip revenue by 2027, which further reinforces the stickiness of its "co-design plus shared roadmap" model.

Is the moat widening, stable, or narrowing. My judgment is: the company-level moat is broadly stable, the AI-related moat is widening somewhat, and the VMware software moat is "strong but increasingly contested." The widening part is in AI interconnect, open Ethernet and custom XPU collaboration; the more contested part is in VMware's bundling, price increases and partner-ecosystem reshaping. The Tomahawk Ultra, Tomahawk 6 and other AI networking products Broadcom launched or emphasized in 2025 show its position strengthening in "open Ethernet replacing closed interconnect"; but the European cloud industry body CISPE has continued through 2025 and 2026 to raise legal and antitrust challenges to VMware licensing and ecosystem changes, a reminder that the deeper the software moat, the easier it is to trigger customer backlash and regulatory attention. I give moat strength 4/5.

A few key judgments. Replicating Broadcom's capability in AI custom chips and Ethernet interconnect is not just a matter of spending money; it also requires years of customer collaboration, advanced-packaging and manufacturing coordination, system-level validation and software compatibility, typically several years and resources in the billions of dollars or more. The company's pricing power in an inflationary environment is present, because FY2025 gross margin recovered to about 67.8% and the software side is shifting to a subscription model; but VMware's price increases also prove that raising prices is not without cost. The company still maintains high profitability and high free cash flow in downturns, which shows that a substantial part of the high margin is a structural advantage rather than a mere cyclical bonus; yet the current super-cycle in the AI business clearly carries elements of cyclicality and a capex surge.

Management and Capital Allocation

Whether management can be trusted. The combination of Chairman Henry Samueli and CEO Hock Tan has historically been labeled not as "great storytellers" but as "good at making trade-offs." Over many years Broadcom expanded from its original semiconductor business into a "semiconductor plus enterprise infrastructure software" platform through M&A; the FY2025 proxy statement also explicitly attributes FY2025's record revenue, operating cash flow and free cash flow to AI semiconductors and VMware software. The proxy further discloses that Hock Tan's leadership was extended by the board through FY2030, with new PSUs and follow-on stock ownership requirements; this helps incentive alignment but also means the scale and concentration of compensation deserve close shareholder scrutiny. My conclusion is: management is broadly credible and clearly long-term oriented, but the governance style leans assertive and must be judged against results.

Ownership and shareholder alignment. As of February 24, 2026, Henry Samueli held about 85.56 million shares, about 1.8%; Hock Tan held about 0.908 million shares; all current directors and officers together held about 89.467 million shares, 1.9%. This is not a typical "high-management-ownership" company, and the CEO's personal stake in particular is not high; but the company has strict ownership guidelines, requiring the CEO to hold 6x base salary and other officers 3x, and it discloses that as of the record date all directors and officers met the relevant requirements. My view is: Broadcom's interest alignment relies more on compensation structure, long-term equity and Samueli's presence as founding chairman than on an exceptionally high CEO personal stake.

Whether capital allocation is excellent. This is the part of Broadcom most worth studying carefully and also the easiest to evaluate emotionally. At the factual level, the capital allocation scorecard is impressive: FY2025 returned 13.6 billion dollars to shareholders, of which 11.1 billion dollars was dividends and 2.5 billion dollars buybacks; the company raised its FY2026 target annual dividend to 2.60 dollars per share, the 15th consecutive year of increasing the annual dividend. In FY2026 Q1, the company paid another 3.09 billion dollars in dividends and executed 7.85 billion dollars of buybacks, and added a 10 billion dollar buyback authorization in March 2026.

But capital allocation also has a cost. The VMware acquisition closed on November 22, 2023, with Broadcom paying VMware shareholders about 30.79 billion dollars in cash and issuing 544 million Broadcom shares with a fair value of about 53.398 billion dollars. As a result, the FY2024 balance sheet expanded significantly, with FY2024 year-end assets of 165.6 billion dollars, including 97.87 billion dollars of goodwill and 40.58 billion dollars of intangibles; at the end of FY2025, goodwill was still as high as 97.80 billion dollars and intangibles 32.27 billion dollars. In other words, this is a company that funds large M&A with strong cash flow, yet carries a great deal of "prepaid future cash flow" on its balance sheet. My assessment of capital allocation is not "perfect" but: historically broadly excellent, with VMware a large bet that looks operationally successful so far but whose long-term return still needs further validation. I give management and capital allocation 4/5.

Specific judgments on buybacks and M&A. Broadcom's buybacks do not belong to the textbook Buffett style of "buying back aggressively only when undervalued and fully restrained otherwise"; they are more a pragmatic style of "returning capital to shareholders when cash flow is very strong and there is no better large use." By outcome, the FY2024 to FY2025 buyback prices were far below the current price, but this does not automatically prove the stock was undervalued at the time. On M&A, the path of CA, Symantec Enterprise and VMware shows that Broadcom is not an organic pure-growth company but an "M&A-integration composite enterprise"; if you dislike a model that drives growth through integration, that alone is a strong enough reason not to buy.

Financial Quality and Owner Earnings

Key financial table.

Note: FY2021 to FY2022 share counts are adjusted for the 10-for-1 split in 2024; free cash flow is estimated as operating cash flow minus capex. The raw data above come from the 2023 and 2025 annual reports, with the 2024 split based on the company's June 2024 announcement.

How to read this table. First, Broadcom's growth is not "burning cash for scale" but high-margin, low-capex, strong-cash-conversion growth: FY2021 to FY2025 revenue CAGR was about 23.5%, while capex stayed at only about 1% of revenue. Second, the sharp drop in FY2024 net margin does not mean cash generation collapsed; FY2024 operating cash flow was still close to 20 billion dollars and free cash flow 19.41 billion dollars, with GAAP profit heavily affected by VMware consolidation, intangible amortization and tax swings. Third, the company's most important financial strength is: accounting profit is often noisy, but cash flow is very good over the long run.

Is the profit real cash profit or accounting profit. Broadcom's GAAP profit and cash flow have shown a clear long-term divergence, but on verified data this looks more like noise from acquisition accounting and taxes than visible earnings manipulation. FY2025 operating cash flow was 27.54 billion dollars, clearly higher than net income of 23.13 billion dollars; the reasons include 8.20 billion dollars of amortization, the depreciation portion within 5.74 billion dollars of depreciation/amortization, 7.57 billion dollars of stock-based compensation, and tax and working-capital items. FY2025 also had a one-time tax benefit, with the effective tax rate even negative, which is a reminder for investors not to treat FY2025 GAAP EPS as "normalized earnings." My judgment is: Broadcom should be tracked more on free cash flow, share count and tax-normalized earnings than on trailing P/E alone. PwC issued unqualified opinions on both the FY2025 and FY2023 financial statements.

Returns on capital, leverage and survivability. Broadcom's ROE has limited reference value under an M&A-driven model, but a rough 2025 calculation on average shareholders' equity is still about 31%, and ROA on average total assets about 14%. More meaningful is ROIC: using a normalized 15% tax rate, Broadcom's FY2023 ROIC was about 27%, fell to about 13% in FY2024 after VMware consolidation, and recovered to about 17% in FY2025, showing that post-acquisition returns are repairing but have not yet returned to pre-acquisition heights. On leverage, FY2025 year-end cash was 16.18 billion dollars and debt principal 67.12 billion dollars, with net debt of about 50.94 billion dollars; on a rough GAAP EBITDA basis, net debt/EBITDA was about 1.5x and EBIT/interest coverage about 7.9x. By FY2026 Q1, cash was 14.17 billion dollars and total debt 66.06 billion dollars, with leverage still manageable. Conclusion: the balance sheet is not "obsessively clean," but it is sound given the company's cash flow capability.

Working capital, inventory, receivables and share count. FY2025 accounts receivable rose by 2.72 billion dollars, and in FY2026 Q1 climbed again from 7.15 billion dollars to 8.46 billion dollars; the company attributes this mainly to higher billings and timing of collections. Inventory rose to 2.27 billion dollars in FY2025 and again to 2.96 billion dollars in FY2026 Q1, showing the company is supporting higher shipments and program ramps, but inventory remains modest relative to revenue. On share count, the real big change came from the 544 million shares newly issued for the VMware acquisition; afterward, FY2024 and FY2025 repurchased 67 million and 16 million shares respectively, offsetting part of the dilution. The fact is that Broadcom is not "zero dilution"; the opinion is that its dilution comes more from strategic M&A than from runaway employee share issuance.

Owner Earnings analysis. Following an "owner earnings" approach, I use two measures. Base measure: treat TTM free cash flow directly as owner earnings. Starting from FY2025 free cash flow of 26.91 billion dollars, adding FY2026 Q1 free cash flow and subtracting FY2025 Q1 free cash flow, TTM free cash flow is about 28.9 billion dollars. Conservative measure: treat stock-based compensation as partly a real cost, so deduct about 50% of TTM SBC from TTM free cash flow, giving conservative owner earnings of about 24.7 billion dollars. Because Broadcom's capex is extremely low, treating all capex as maintenance capex is already conservative enough; working-capital swings are already reflected in operating cash flow. On the current market cap of about 1.95 trillion dollars, Broadcom corresponds to about 69x TTM free cash flow, or about 79 to 81x conservative owner earnings. This is not a cheap valuation.

Valuation and Margin of Safety

Owner Earnings Discount Method

The table below shows three scenarios I deliberately skew conservative. I stress here: this is not "the truth" but a way to put a good company into rigorous assumptions and see whether the price still leaves you any room.

Note: Starting owner earnings are estimated from FY2025 and FY2026 Q1 / FY2025 Q1 cash flow data; intrinsic value per share is estimated on about 4.735 billion shares. The conservative scenario here already assumes Broadcom remains a strong company, the neutral scenario already clearly counts AI growth, and the optimistic scenario requires owner earnings to compound rapidly over the next decade.

My interpretation. If a company's current price struggles to leave room even under many optimistic scenarios, then it is not "safe" in the value-investing sense. Broadcom's most fragile valuation assumption right now is that the market assumes it can extend AI-related high growth at fairly high quality for a long time, while VMware software does not weaken cash flow quality under customer backlash and regulatory pressure. Put differently, the current price looks like it is discounting in advance a path where "almost everything goes right." My judgment is: the margin of safety is insufficient.

Relative Valuation Method

Broadcom's current relative valuation is hard to call cheap. Based on verified current market data, Broadcom's trailing GAAP P/E is about 83.7x and P/B about 24x; Nvidia is about 54.1x, Marvell about 61.8x, Oracle about 32.6x, Qualcomm about 21.0x and AMD about 135.8x. It should be noted that GAAP P/E is noisy under acquisition amortization and tax disturbances, and AMD, Marvell and others are not necessarily "comparable"; but even so, Broadcom remains in the expensive zone. Reuters also provides more comparable forward valuations: Marvell's 12-month forward P/E is about 27.25x, while Broadcom's is about 38.39x. Broadcom's quality is indeed higher than many peers, but high quality does not mean price can be ignored.

More important is cash flow relative yield. By this report's estimate, Broadcom's TTM free cash flow yield is about 1.4% to 1.5%; meanwhile, on May 18, 2026, the U.S. 10-year Treasury yield was about 4.61% and the Moody's Aaa corporate bond yield about 5.65%. This does not mean bonds are necessarily better than Broadcom, but it does mean: the price you pay for Broadcom today has already taken away many years of future high growth in advance. If future growth turns out merely "decent" rather than "very strong," the return will be swallowed by a very expensive starting point.

Asset and Liquidation Value Method

On an asset basis, Broadcom is not cheap and does not even have much of a "hard-asset floor." At the end of FY2025, total assets were 171.09 billion dollars, of which goodwill was 97.80 billion dollars and intangibles 32.27 billion dollars, totaling 130 billion dollars; shareholders' equity was 81.29 billion dollars. In other words, Broadcom's tangible net assets are roughly negative. This does not mean the company is in danger; it means there is almost no "liquidation discount" to pick up here, and its value comes almost entirely from future cash flow. For a long-term business owner, this is not a bad thing; but for a value investor who emphasizes downside protection, it means you must care more about price and competitive position than about book assets.

Intrinsic Value Range and Buy/Sell Ranges

Combining owner-earnings DCF, relative valuation and the asset method, I offer the following ranges:

At the current price of about 411 dollars, Broadcom trades at roughly a 47% to 142% premium to the fair value range I give; and still about a 25% to 47% premium to the optimistic value range. My clear conclusion is: this is a textbook candidate for "good company, bad price."

Risks, Bear Case and Opportunity Comparison

The most important risks. First, competitive risk: Nvidia is still the AI platform definer, Marvell competes head-to-head in custom chips and networking, and large customers such as Google are strengthening their bargaining power and multi-supplier strategies. Second, technology substitution and customer in-house design risk: Broadcom itself acknowledges that part of the competition comes from customers' internal resources. Third, regulatory and customer-relationship risk: VMware continues to face antitrust complaints and litigation pressure from CISPE in Europe, and Broadcom has recently even gone to court with EU regulators over procedural matters tied to related investigations. Fourth, leverage and post-M&A aftermath risk: although leverage is manageable, the company still carries debt on the order of 66 billion dollars. Fifth, overvaluation risk: this is the most realistic risk right now, requiring no deterioration in fundamentals, since multiple compression alone is enough to cause years of low returns. Sixth, supply-chain and geopolitical risk: most of the company's products and materials are manufactured and sourced in Asia, and its 10-K explicitly flags risks such as China-Taiwan relations.

The strongest bear case. The strongest bear logic is actually simple: Broadcom may be a very good company, but the current stock price has already bundled in "an AI surge plus high-margin VMware renewals plus multiples that never come down" all at once. If 2027 AI demand does not boom the way the market imagines today, or key customers such as Google/Meta/OpenAI slow their programs or shift toward multiple suppliers, Broadcom's owner earnings may not be able to sustain a high compound growth rate near 18% over a decade. At the same time, if VMware's pricing adjustments keep triggering customer migration, partner departures and regulatory constraints, the software segment's "high margin plus high stickiness" assumption will also be discounted. By then Broadcom's fundamentals may still be good, but today's buyer could still bear a permanent capital loss of 40% to 60%.

Which facts would overturn the investment judgment. If the following facts appear over the next 12 to 24 months, I will clearly admit the original judgment needs revision: first, AI revenue keeps growing at high quality, and Broadcom's multi-customer custom ASICs, once in volume production, prove that owner earnings really can keep expanding at mid-to-high double digits; second, VMware customer renewal rates, private-cloud-stack adoption and regulatory outcomes turn out better than market fears; third, the company maintains a high free cash flow rate, low capex and manageable leverage even while growing fast. Conversely, if the following facts appear, I will admit I was wrong and turn more cautious: first, AI growth slows materially after Q2-Q4; second, the VMware partner ecosystem keeps deteriorating and triggers substantive remedies; third, net debt fails to come down while stock-based compensation keeps pushing real costs higher.

Comparison with other opportunities. Against its strongest peers, Broadcom's advantage is a "more balanced cash flow and business mix" rather than unbeatable single-point technology; compared with Nvidia, it looks more like a composite that "sells the shovels, sells the routing and sells the tollbooths" of AI infrastructure; compared with Marvell, it is larger in scale, stronger in cash and thicker in software base. Against the index, Broadcom's total shareholder return clearly beat the S&P 500 over the past 5 years, with the proxy disclosing a five-year TSR of 1,082%; but past returns cannot substitute for today's expected returns. Against risk-free yields, Broadcom's roughly 1.5% free cash flow yield is below both the 10-year Treasury and Aaa corporate bond yields. Therefore, if your alternatives are a broad-based index or high-grade bonds, AVGO is clearly worth tying up capital only if you are very confident about AI delivery over the next decade. If I could hold only 5 assets, I do not think it qualifies at the current price; if there is a clear pullback into the 160 to 220 dollar zone, it earns a place back on the candidate list.

Open questions and information limitations. This report has three main limitations: first, the latest complete single-customer revenue share and the breakdown of the top few customers were not obtained as full financial footnote text in the fragments verified this time; second, strictly comparable, same-basis peer EV/EBITDA, P/FCF and ROIC data were not all obtained, so the relative valuation section leans more on Broadcom's own free cash flow yield and current P/E position; third, as of May 20, 2026, Broadcom had not yet disclosed FY2026 Q2, so this report uses the latest disclosed data from FY2025 and FY2026 Q1.

Checklist and Final Investment Conclusion

Investment Checklist

The judgments in this checklist combine the earlier analysis of business, competition, cash flow, leverage, governance and valuation.

【Final Rating】 Watch.

【One-Sentence Investment Thesis】 Broadcom is an excellent, cash-rich infrastructure platform company, but the current price demands near-flawless delivery for years, leaving an insufficient margin of safety.

【Core Bull Case】

• Strong position in AI custom ASICs and Ethernet interconnect, with FY2025 AI semiconductor revenue of 20.2 billion dollars and FY2026 Q1 AI revenue of 8.4 billion dollars; the cycle remains hot.

• Infrastructure software upgrades the business from "pure-cycle chips" to "chips plus more recurring software," with FY2025 software revenue of 27 billion dollars.

• Extremely strong free cash flow capability, with FY2025 free cash flow of 26.9 billion dollars and very light capex.

• A strong historical track record in management integration and capital allocation, with the annual dividend raised for 15 consecutive years.

【Core Bear Case】

• The current valuation is too high, at about 69x TTM free cash flow and about 84x trailing GAAP P/E, very hard to call "having a margin of safety."

• VMware licensing and ecosystem restructuring still carries risks of customer churn, litigation and EU regulation.

• The AI business depends heavily on a few large customers and large programs; once the pace slows, the stock will be very sensitive to downward fundamental revisions.

• Tangible net assets on the books are negative, meaning downside protection rests mainly on future cash flow rather than hard assets.

【Key Assumptions】 For Broadcom's investment logic to hold, at minimum these conditions must be met: AI owner earnings keep growing at high double digits for years; VMware renewals and private-cloud platform adoption are not materially disrupted by regulation and customer backlash; stock-based compensation and acquisition amortization do not keep eroding real "per-share" returns; leverage stays at a manageable level and continues to decline.

【Fair Buy Price】 I think a more reasonable buy price is 160 to 220 dollars per share, derived from this report's fair value range of 170 to 280 dollars with an added 20% to 30% margin of safety. If you only accept the approach of "extremely high quality, without demanding high returns," you can widen the watch ceiling to 220 to 260 dollars, but that no longer belongs to typical value investing.

【Target Holding Period】 If bought at a reasonable price in the future, I think Broadcom is suited to a holding period of 5 to 10 years or more, because its value is determined mainly by long-term cash flow and customer embedding rather than the volatility of a quarter or two.

【Expected Annualized Return】 Based on this report's scenario assumptions, my expected annualized returns at the current price are roughly: conservative scenario about -1% to 0%, neutral scenario about 4% to 5%, optimistic scenario about 9% to 10%. These estimates are built on TTM owner earnings, a payout rate of around 40%, and exit valuations after 10 years of about 25x, 30x and 35x owner earnings respectively. In other words, Broadcom is not "without return," but only when the optimistic scenario clearly materializes is the return enough to compensate for today's high price.

【Maximum Loss Risk】 The worst case is not corporate bankruptcy, but "good company plus bad price" layered with "growth slowdown." If AI growth comes in below expectations, VMware software margins decline, and valuation multiples revert to a more normal range, Broadcom's stock could fully see a permanent capital loss of 40% to 60%. Because its tangible net assets are negative, once valuation returns to a more ordinary cash flow multiple, the downside will hurt.

【Tracking Metrics】 Going forward I will keep tracking these: AI semiconductor revenue growth; the number of hyperscaler customers added or lost; VMware Cloud Foundation subscription renewals, partner feedback and regulatory progress; infrastructure software revenue and gross margin; the free cash flow rate; stock-based compensation as a share of revenue; net debt/EBITDA; buyback prices and buyback size; diluted share count; changes in inventory and accounts receivable.

【Signals That Trigger Re-evaluation】 If any of the following occurs, I will immediately rebuild the model: AI revenue or management's AI visibility is materially revised down; key customers such as Google, Meta, OpenAI and Anthropic show signs of program delays or order shifts; VMware faces substantive regulatory remedies or large-customer litigation spreading; free cash flow begins to run persistently and materially below net income; stock-based compensation and dilution turn higher again; deleveraging stalls.

【Final Recommendation】 If you think like someone acquiring a whole business, Broadcom deserves respect and deserves long-term tracking; but if you demand "room to spare" the way a value investor does, today's price is not restrained enough. For this kind of company, the hardest part is not understanding the company but accepting the discipline of "right company, not yet buying." My recommendation is: put Broadcom on a high-priority watchlist, focus on AI revenue delivery, the VMware ecosystem reaction and the window for valuation to come down; until there is a better price, stay patient.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.