Monumental Numbers, Impossible Price

Broadcom Inc — one of the largest semiconductor companies in the world, expanded heavily into infrastructure software. All figures in US dollars; Broadcom's fiscal year ends in early November (FY2025 ended 2 November 2025). The valuation tool shows figures in cents (477 = $4.77).

Broadcom Inc is one of the largest semiconductor companies in the world, and it has expanded heavily into infrastructure software. Its semiconductors serve computing and networking, and custom AI accelerators now account for a huge amount of the business. It holds some manufacturing in house, but it is primarily a fabless designer. In software, it sells virtualisation, infrastructure and security software to large enterprises, financial institutions and governments.

The revenue record is astounding. Growth of 14.9% (FY2021), 21.0% (FY2022), a lacklustre 7.9% (FY2023), then 44.0% (FY2024) and 23.9% (FY2025). From FY2021 to FY2025 turnover grew by around 132% — over four years, an average of roughly 33% a year. Remarkable levels. It is interesting that FY2025 was lower than FY2024, and worth noting that these growth rates are not very stable: they are mostly massive figures, but not stably in the same direction.

Gross margins have oscillated too — from 61.4% (FY2021) up to 68.9% (FY2023), landing at 67.8% in FY2025 — a surprising amount of movement, but always very high, as you would expect of a company of this nature. FY2025 operating profit was monumental at $26.1 billion, producing pre-tax profit of $23.3 billion and reported earnings per share of 477 cents. As always with these very large super-tech companies, the figures are monumental.

Six-year financial history

Balance sheet highlights

The cash jump (from $9.3bn to $16.2bn, up 73%) is the standout — driven by $27.5bn of operating cash flow, partly offset by −$11.1bn of dividends and debt repayments. Receivables rose 62% to $7.1bn on higher billings. Encouragingly, they used the cash to cut long-term debt from $66.3bn to $62.0bn (down −6.5%), and total liabilities fell −8.3%. Equity rose 20% to $81.3bn.

The income statement shows exactly the kind of growth noted above. Product sales rose from $35.0bn to $44.8bn, and subscriptions and services from $16.6bn to $19.0bn:

Statement of operations

The operating leverage is striking: revenue up 23.9%, but operating income up 89% and net income nearly quadrupling, as the FY2024 VMware integration costs and amortisation fell away. Selling, general and administrative costs actually dropped −15.1%. This is the year the VMware acquisition stopped being a drag and started paying off.

Research and development spending was broadly flat around $4.9–5.3bn for years, then jumped 77% in FY2024 when VMware was consolidated, and rose a further 17.9% in FY2025. So the FY2025 rise to $11.0bn is a continuation of the VMware step-change, not a fresh surge.

R&D expense by fiscal year

Expected R&D for the coming year (FY2026). Broadcom does not guide a headline R&D dollar figure, but the direction is clearly up. In Q1 FY2026 (quarter ended 1 February 2026), R&D rose 32% year on year — a faster rate than the full-year FY2025 increase — as the company pours investment into custom AI accelerators and AI networking. On that Q1 run-rate, full-year FY2026 R&D would plausibly land in the region of $13–14bn. The strategic logic is the AI-accelerator roadmap: Broadcom guided Q1 FY2026 AI semiconductor revenue to roughly double year on year to $8.2bn, and that kind of custom-silicon design work is R&D-intensive. So I expect R&D to keep climbing in absolute terms, even if it stays roughly flat as a percentage of revenue (it was 17% of revenue in FY2025, down from 18%).

The company is growing very fast, in revenue terms and — especially in the last twelve months — on earnings per share. But the valuation rests on extrapolating a historical four-year earnings-per-share growth average of around 33.5%, and I do not think it will continue at quite that pace. Some will find that a ridiculous thing to doubt when the record is this strong, but the growth is not stable year to year.

They might well see explosive growth over the next three years, but I do not foresee sustained earnings-per-share growth faster than the low-thirties percent. To justify the market's valuation you would have to hope for something like doubling earnings per share again — a continued growth rate around 36% — which is not likely. So I do not think it is worth investigating further: on its historical trend it is overvalued, quite heavily so.

I am using a growth rate of 20%, which is lower than the historical average but matches the two-year growth figure, applied to the current reported earnings per share of 477 cents with a valuation multiple of 9. That gives a formula valuation of 12,020.40 cents (around $120) per share against a current price of 37,671 cents (around $377) — a value opportunity of −68.1%. I set a 20% risk factor and a probability of 60%, because I am reasonably sure this is an overvalued share, or at least one that will struggle to surprise investors to the upside. Even if you used the full 33.5% historical growth rate, the shares would still screen as heavily overvalued.

VALUATION SNAPSHOT

Market capitalisation and valued market capitalisation are computed from the current share price and the 4,757.58m shares in issue so the table is internally consistent.

This gives an actual valuation of 12,020.40 cents against a current share price of 37,671 cents — a value opportunity of −68.1%. On these inputs the shares look dramatically overvalued. I should acknowledge the obvious counter: the market is paying a trailing price-to-earnings ratio near 79x precisely because it is capitalising an AI-accelerator growth story that could, for a few years, run faster than my 20% assumption. The custom-silicon order book and the AI-revenue doubling guidance are significant bull arguments. But my framework deliberately refuses to extrapolate a volatile, high-thirties growth rate over the long run, and on that basis I am comfortable calling it a Sell.

• The AI-accelerator story runs hotter for longer: Q1 FY2026 AI semiconductor revenue was guided to roughly double year on year to $8.2bn. If custom silicon for the hyperscalers keeps compounding at that pace, my 20% growth assumption will look far too cautious — the single biggest risk to a Sell.
• Explosive near-term growth is plausible: I said it myself — they might well see explosive growth over the next three years. A couple more years at 30%+ earnings growth would close a good part of the valuation gap.
• The VMware software annuity: Subscriptions and services revenue of $19.0bn gives a sticky, recurring earnings base that deserves a premium multiple against ordinary chip cyclicality.
• Formidable cash generation: $27.5bn of operating cash flow, cash up 73% to $16.2bn, and long-term debt cut by $4.3bn in a year — the balance sheet is strengthening rapidly.
• Momentum: Mega-cap AI names have repeatedly stayed more expensive for longer than value frameworks suggest they should.