Broadcom & the Annuity

Broadcom has executed one of the cleanest AI pivots in semiconductors — but the share price now embeds most of it. The custom-silicon franchise is genuine: Q2 AI revenue grew 143% to $10.8 billion, the latest in a relentless two-year beat-and-raise, and the line-of-sight to more than $100 billion of AI chip revenue in fiscal 2027 rests on multi-year, gigawatt-scale customer commitments rather than spot demand. Faircurve credits that annuity. What it will not do is pay twice for it. At 26x forward earnings — a premium to the higher-quality NVIDIA, and a ~50% premium on EV/Sales — the stock is priced for the ramp to persist and broaden, and a reverse-DCF shows the current enterprise value already requires roughly 23% annual revenue growth sustained for five years. That is achievable, not assured, and the customer concentration that powers the upside is the same concentration that defines the downside. The risk-reward is two-sided and close to balanced: a base-case fair value of ~$460 sits just below spot.

Three signals will settle whether the annuity deserves the multiple. First, the conversion of development-stage hyperscale customers — the named fourth-to-sixth, including OpenAI and Anthropic — from design into volume production, which is what turns a serviceable market into booked revenue. Second, the trajectory of infrastructure software, where persistent single-digit growth would undercut the 'de-risked by software' thesis. Third, any change in hyperscaler capex commentary, the upstream variable custom-XPU demand ultimately tracks. A fourth-and-fifth customer reaching scale pushes the read toward the bull case; the first sign of an AI-order push-out confirms the base.

The transcript arc is a two-year crescendo. ‘AI’ mentions roughly double into fiscal 2026 while ‘VMware / software’ and the broad ‘hyperscaler’ language fade — the story narrows from a diversified conglomerate to a custom-silicon supplier to a handful of named buyers. Two sub-surface signals matter. First, ‘new / named customers’ and ‘bookings / backlog’ climb as the roster widens from three hyperscalers to six (Google, Meta, Anthropic, OpenAI and two others) and as gigawatt-scale, multi-year commitments replace a bounded ‘$60–90B serviceable market.’ Second, the recurring metrics show AI revenue compounding from $3.1B to $10.8B a quarter while the margin and free-cash-flow lines rise through the mix shift — the operating leverage is real. The concentration the first signal reveals is the same concentration the bear case rests on.

Across eight quarters the AI guide was raised every single time — from a $3.5 billion next-quarter number to $16 billion, and from a bounded ‘$60–90 billion serviceable market’ to a ‘line-of-sight above $100 billion’ in fiscal 2027, backed by gigawatt-scale, multi-year commitments from six named customers. That cadence is the bull case in one data series. The same record carries the bear’s warning: management has repeatedly stressed the ramp ‘will not be linear’ and concedes ‘quarterly variability,’ and the entire step-change rests on a customer roster narrow enough to name. The arc is one of accelerating conviction built on deepening concentration.

At 26x forward earnings Broadcom screens below the peer mean of 33x — but that mean is inflated by AMD (41x) and Marvell (50x), smaller, lower-margin, higher-beta challengers carrying speculative AI-share premiums. Against the two large-cap quality anchors — NVIDIA at 17.4x and Qualcomm at 23.7x — Broadcom sits at a clear premium, and on EV/Sales it trades at 14.1x, roughly 50% above NVIDIA’s 9.4x despite comparable margins. The premium is earned: a 69% adjusted EBITDA margin, a 46% free-cash-flow margin and a software annuity none of the silicon peers possess. But it is a premium to the best-in-class comparable, not a discount — the valuation is rich relative to quality, not cheap.

Broadcom’s own discounted-cash-flow model returns roughly $300 on an 11.3% cost of capital and 3.0% terminal growth — well below spot, as is typical for a name whose value is front-loaded into a steep near-term ramp; it carries only a 20% weight as a cash-flow discipline. FMP’s independent levered-DCF is lower still at $194. Both confirm the direction: little of Broadcom’s value is supported by discounted cash flows at a reasonable discount rate — the case rests on the multiple the market assigns the growth.

At a $2.31 trillion enterprise value, a 10% annual return over five years requires the enterprise value to compound to roughly $3.73 trillion by fiscal 2032. Held to a normalized ~8x EV/Sales — a premium the franchise can plausibly sustain given 50%-plus margins, but well below today’s 14x — that implies about $466 billion of revenue, a ~23% compound annual growth rate off the fiscal-2027 base (about 29% off fiscal-2026). Consensus already has growth decelerating from +59% in FY27 to +29% in FY28 and toward the mid-teens thereafter; sustaining a 23% average for five years requires the AI-silicon annuity to keep compounding well past the current six-customer cohort. On earnings, the same math leaves the stock capitalizing close to the full FY27 consensus at a normalized multiple — little margin of safety. The price is not irrational; it leaves the durability of the annuity, not its existence, as the open question.

Faircurve Equities · Independent Single-Name Research · AVGO · 4 Jun 2026 · Issue No. 017

Sources: Financial Modeling Prep (quote, profile, statements, financial-estimates, key-metrics, DCF). Broadcom Q2 FY2026 Form 8-K (filed 3 Jun 2026). Last 8 quarterly earnings-call transcripts. Peer data: NVDA, AMD, MRVL, QCOM (FMP).

Methodology: Forward FY+1 (calendarized ~2027) peer comparison. Consensus-anchored base case. 40/40/20 blend of Forward P/E, Forward EV/EBITDA and a Faircurve DCF. CAPM cost of capital. Reverse-DCF on implied revenue growth.

Disclaimer: Research, not investment advice. Not a recommendation to buy or sell any security. The author held no position in AVGO at publication.