# Broadcom Inc. (AVGO) — InvestMoat Analysis _Last analyzed: July 19, 2026_ _Asset class: equity · Canonical page: https://investmoat.com/stocks/avgo_ ## Scores | Dimension | Score (0–100) | | --- | --- | | Moat durability | 67 | | Growth trajectory | 90 | | Valuation | 75 | | **Composite** | **80** | | **Recommendation** | **Accumulate** | Scores are computed deterministically from this asset’s data by the InvestMoat formula (see https://investmoat.com/llms.txt for methodology). Scores are not directly comparable across asset classes. ## Key stats - **Ticker:** AVGO - **Market Cap:** $1.83T ## Moat Dual-moat machine — irreplaceable custom silicon for AI hyperscalers and VMware's stranglehold on enterprise virtualization infrastructure. Apple renewed its supply agreement through 2031 (>$30B, spanning FBAR wireless filters and custom AI silicon, July 2026), joining Meta (through 2029) and Google (through 2031) as long-dated contracted customers. Moat statuses unchanged in the July 2026 review: the Q2 sell-off was a multiple event, not moat damage — though the CISPE EU escalation (five trade groups urging interim measures, July 15) now poses a live regulatory threat to VMware pricing power. ### The Dual-Moat Architecture Broadcom's moat is built on **Proprietary Silicon and Enterprise Infrastructure Lock-In**: - **Custom AI Accelerator Dominance:** Broadcom designs custom AI XPUs for six confirmed hyperscaler customers. Q2 FY2026 AI revenue hit $10.8B (+143% YoY), with Q3 guided to $16.0B (+200%+ YoY) and full-year FY2026 guidance raised to $56B (~+180% YoY); management reaffirmed $100B+ for FY2027. Meta is extended through 2029 (multi-GW MTIA rack-level supply); Google is formalized through 2031 (future TPU generations, AI networking, rack-scale components); Apple renewed through 2031 (>$30B covering FBAR filters and custom AI silicon, July 2026). Anthropic is tripling compute to 3GW+ before year-end 2026. The OpenAI 10GW inference program remains stalled on an $18B financing snag (Broadcom requiring a Microsoft 40% off-take that Microsoft has not agreed to), with first Jalapeno chips now expected mostly in 2027. - **Ethernet Networking Monopoly for AI:** Broadcom's Tomahawk and Jericho switch chip families dominate the hyperscale data center networking market. As AI clusters require ever-faster intra-cluster networking at lower cost than InfiniBand, Broadcom's Ultra Ethernet ecosystem is positioned to capture the AI networking buildout. Broadcom is now selling fully assembled 'Ironwood Racks' (Google's 7th-gen TPU) directly to AI firms including Anthropic. - **VMware: The Enterprise Infrastructure Tax:** VMware's vSphere and vCenter run the virtualization layer for ~70% of the Fortune 500. VMware Cloud Foundation (VCF) is described by management as the 'permanent abstraction layer' between AI software and physical silicon that 'cannot be disintermediated or replaced.' Q2 FY2026 infrastructure software revenue was $7.18B (+9% YoY) with ARR growth of 17% YoY (a modest deceleration from Q1's 19%) at 93% software gross margins. The regulatory front is heating up: after Broadcom terminated most of its European VMware Cloud Service Provider partnerships (January 2026), CISPE filed an EU antitrust complaint (March 2026), and on July 15 2026 five cloud trade groups jointly urged the EU Commission to impose interim measures suspending Broadcom's partner-program terminations. - **Hock Tan's Capital Allocation Engine:** CEO Hock Tan has executed the most disciplined semiconductor M&A strategy of the past decade — acquiring CA Technologies, Symantec enterprise security, and VMware, then aggressively cutting costs, raising prices, and migrating customers to subscription contracts. In Q1 FY2026, Broadcom returned $10.9B to shareholders ($3.1B dividends + $7.8B buybacks). The board authorized a new $10B buyback through 2026. **Moat verdict:** Broadcom is a net AI beneficiary with two compounding engines: custom XPU co-development for six hyperscaler customers (AI revenue +143% YoY to $10.8B in Q2 FY2026; FY2026 guided to $56B, ~+180% YoY; FY2027 $100B+ reaffirmed) and VMware's stranglehold on enterprise virtualization (ARR +17% YoY at 93% gross margins). The strongest AI-resilient moats are systemOfRecord (vCenter as compute-inventory authority), businessLogic (custom silicon embedded in hyperscaler training pipelines), and talentScarcity (scarce custom ASIC design teams). Long-dated contracts — Meta through 2029, Google through 2031, and now Apple through 2031 (>$30B, July 2026) — meaningfully reduce medium-term customer concentration risk. July 2026 review: all ten moat statuses unchanged — the ~25% drawdown from the June 3 ATH ($495) was a multiple event (Q3 AI guide of $16B vs ~$17.2B whisper; $1.3T chip-sector sell-off June 5), not moat damage. Three risks require monitoring: (1) AI-capex digestion — the whisper-miss reaction proved expectations fragility, and the OpenAI $18B financing snag remains unresolved with Jalapeno chips slipping to 2027; (2) the CISPE EU complaint has escalated — five cloud trade groups jointly urged interim measures against Broadcom's VCSP terminations on July 15 2026; an adverse interim decision would directly weaken the bundling moat and warrant a status downgrade; (3) customer concentration among a handful of hyperscalers. Reassess after Q3 FY2026 results (early September 2026). ## Growth AI XPU revenue grew 143% YoY to $10.8B in Q2 FY2026 (reported June 3), beating the $10.7B guide; Q3 is guided to $16.0B (+200%+ YoY) and full-year FY2026 AI guidance was raised to $56B (~+180% YoY) from the prior ~$46B trajectory. Management reaffirmed $100B+ AI chip revenue for FY2027, up from $20B in FY2025. Total Q3 revenue guided to ~$29.4B (+84% YoY). Apple renewed its supply agreement through 2031 (>$30B); Meta extended through 2029; Google through 2031; Anthropic tripling to 3GW+ by year-end. VMware infrastructure software ARR growth decelerated modestly to 17% YoY (from 19%) on $7.18B quarterly revenue (+9% YoY). Note: the Q3 AI guide of $16B landed below the street's ~$17.2B whisper, triggering a 12% one-day sell-off — an expectations problem, not a demand problem. - **Revenue CAGR estimate:** 30–50% - **Primary type:** TAM expansion - **Margin trend:** stable - **Key risk (moderate):** AI-capex digestion risk partially materialised in June–July 2026: the Q3 AI guide of $16B (below the street's ~$17.2B) triggered a 12% one-day drop and a $1.3T global chip-sector sell-off, and the stock sits ~25% below its June 3 ATH ($495) amid insider selling and AI fatigue — proof the multiple, not just the fundamentals, is the exposure. The OpenAI $18B financing snag remains unresolved (Broadcom requires a Microsoft 40% off-take; OpenAI's compute head calls that structure 'likely unworkable'; Jalapeno chips slip mostly to 2027) — if the 10GW program formally dies or another top-6 customer pushes out delivery by 2+ quarters, the FY2027 $100B+ target is at risk. Separately, an EU interim measure on the CISPE complaint (five trade groups urged action July 15 2026) forcing VCSP partner reinstatement or pricing rollback would stall VMware ARR growth below 10% - **Drivers:** - AI Custom Silicon (XPU) — +143% YoY to $10.8B in Q2 FY2026; Q3 guided $16.0B (+200%+ YoY); FY2026 guidance raised to $56B (~+180% YoY) (accelerating) - Infrastructure Software (VMware/VCF) — +9% revenue YoY to $7.18B in Q2; +17% ARR YoY (vs +19% in Q1) (stable) - AI Networking (Tomahawk/Jericho) — Included in ~$15.0B Q2 semiconductor revenue (+~79% YoY); Tomahawk 6 ramping at 102.4 Tbps (accelerating) - **Score derivation:** Base 92.5 (CAGR midpoint 40% of 30–50% range: FY2026 AI revenue guided to $56B, ~+180% YoY; FY2027 $100B+ reaffirmed June 3 2026) + 2.7 trajectory (2 of 3 drivers accelerating: AI XPU +143% YoY accelerating toward +200% guided; AI networking accelerating; VMware stable) + 0 margin (stable) + 3 TAM expansion (AI XPU TAM toward $200B+ by 2027; 6 hyperscaler programs plus Apple contracted through 2031) − 5 moderate risk (customer concentration; OpenAI $18B financing snag unresolved with Jalapeno chips slipping to 2027; Q3 whisper-miss shows expectations fragility) = 93 ## Valuation At ~$370 (July 18, 2026), Broadcom trades ~24% below the $490 base case — the most attractive entry since April, courtesy of the post-earnings de-rating rather than any fundamental deterioration. The stock fell ~25% from its June 3 ATH ($495) after the Q3 AI guide of $16B missed the street's ~$17.2B whisper despite Q2 beating and FY2026 AI guidance being raised to $56B. The forward P/E has compressed to ~20× (non-GAAP NTM) — now below the semiconductor industry median (~34×) — and the PEG sits well under 0.5. The market is pricing meaningful AI-capex digestion risk that the order book ($56B FY2026 guided, $100B+ FY2027 reaffirmed, Apple contracted through 2031) does not yet show. **Fair value:** $490 — Multi-year customer commitments — Apple through 2031 (>$30B, added July 2026), Meta through 2029, Google through 2031, Anthropic tripling to 3GW+ by year-end — provide forward visibility that few semiconductor names can match. The Q3 FY2026 print (early September 2026) is the next catalyst: guidance is ~$29.4B revenue (+84% YoY) and $16.0B AI revenue (+200%+ YoY), and after June's lesson the setup matters more than the print — a beat-and-raise against a now-reset multiple has asymmetric upside. Two open items to monitor: (1) the OpenAI $18B financing snag — still unresolved, Jalapeno chips slipping to 2027; formal resolution either way removes an overhang or confirms the first major program casualty; (2) the CISPE escalation — five cloud trade groups urged EU interim measures on July 15; an adverse interim decision would hit VMware pricing power directly. Sell-side consensus remains Buy with an average target ~$501. | Multiple | Value | Note | | --- | --- | --- | | Trailing P/E (GAAP) | ~62× | TTM GAAP EPS ~$5.94 | | Forward P/E (NTM) | ~20× | non-GAAP NTM EPS est. ~$18.5 | | PEG Ratio | ~0.45× | fwd P/E ÷ ~45% EPS CAGR | | Price / Sales (NTM) | ~15× | ~$120B NTM revenue est. | | Price / FCF | ~55× | ~$33B FCF (TTM est.) | The June–July de-rating flipped AVGO from premium-priced to relatively cheap: forward P/E compressed from ~36× (May) to ~20×, now ~41% below the semiconductor industry median (~34×) even as forward EPS estimates rose on the $56B FY2026 AI guide and the guided Q3 revenue step-up to ~$29.4B (+84% YoY). The PEG of ~0.45 is deep GARP territory. The multiple compression reflects sector-wide AI fatigue ($1.3T chip-sector wipeout June 5, insider selling, 'great isn't good' whisper-number dynamics) rather than an AVGO-specific demand problem — Q2 actually beat and full-year guidance went up. The risk is that the market is early on capex digestion; the opportunity is that the order book says it isn't. _(as of July 2026)_ ## Price scenarios ### Bear — $200 EU antitrust ruling forces VMware pricing rollback; hyperscaler XPU customers accelerate in-house ASIC programs; AI semiconductor demand disappoints expectations. - EU Commission issues adverse ruling on CISPE complaint, ordering VMware pricing rollback and partner program reinstatement, cutting software ARR growth below 5% - Google, Meta, or Anthropic materially reduces custom XPU orders or accelerates fully in-house ASIC design, collapsing the path to $100B AI chip revenue in FY2027 - Semiconductor industry downcycle hits networking and broadband segments while AI chip ramp slows from supply or execution challenges - GAAP amortization burden from VMware acquisition weighs on reported earnings longer than expected, compressing the stock's re-rating ### Base — $490 AI chip revenue reaches the guided $56B in FY2026 and $100B+ in FY2027; VMware ARR sustains mid-to-high-teens growth; 27× forward non-GAAP P/E on FY2027 EPS of ~$18. - VMware Cloud Foundation (VCF) adoption grows across enterprise base, sustaining 15%+ ARR growth at 93% software gross margins - AI semiconductor revenue reaches the guided $56B in FY2026 and $100B+ in FY2027 as contracted customers (Meta through 2029, Google and Apple through 2031) scale to multi-GW deployments - Broadcom sustains 68%+ adjusted EBITDA margins while returning $10B+ per quarter to shareholders - Networking segment benefits from Ultra Ethernet Consortium standardization, cementing Tomahawk as the AI cluster switch of choice ### Bull — $650 AI revenue exceeds $100B in FY2027 ahead of schedule; sovereign AI programs and new enterprise verticals become XPU customers; VMware becomes the de facto private cloud AI platform. - AI XPU market expands beyond the 6 known customers; sovereign AI programs (US DoD, EU agencies) and enterprise verticals become incremental XPU customers - VMware Cloud Foundation becomes the dominant enterprise private cloud platform for AI workloads, with ARR scaling to $15B+ at 75%+ margins - Broadcom's custom silicon expertise and rack-scale integration prove cost/performance superior to NVIDIA GPUs for inference at scale, capturing >30% of hyperscaler AI accelerator spend - Dividend doubles within 3 years as free cash flow exceeds $40B annually, triggering multiple expansion --- InvestMoat is an open-source research and education framework. Nothing here is financial advice. Past performance does not guarantee future results.