Broadcom's custom AI accelerators are running critical infrastructure for six hyperscalers (including OpenAI), with AI revenue projected to eclipse $100B in FY27. Simultaneously, the massive VMware integration has ballooned gross margins to ~70%. Yet the stock plunged 12%+ on its latest record-breaking earnings because of a single risk: what happens when six customers control the demand, and a $64B debt pile restricts the balance sheet?
Gray line = Broadcom's actual price trajectory (split-adjusted, running from $269 low in July 2025 → $495 pre-earnings peak → $360.45 today following a "sell the news" reaction); colored paths = synthesized scenario midpoints forward, probability-weighted (base 50% · bull 25% · bear 25%). Mid-year marks. Wall Street 12-month consensus is ~$525 (Strong Buy dominant among 50 analysts).
The consensus targets are built on high conviction in AI spending. If you disagree, adjust the dials. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of the scenario cards all update live.
Hock Tan's empire rests at the intersection of custom silicon and enterprise software. Depending on which facet you emphasize, the valuation looks entirely different. Synthesized expert perspectives with 12-month targets.
Broadcom is the quiet toll-collector on the AI hyperscaler buildout. OpenAI's new "Jalapeño" chip brings their major custom silicon roster to six, completely derisking reliance on Google's TPU. AI revenue is rocketing toward $100B in FY27, backed by multi-gigawatt power deployment pacts through 2029. Paying 23x forward earnings for a company capturing the most lucrative custom ASIC cycles in history is an absolute steal compared to Nvidia.
The beauty of Hock Tan’s M&A playbook is the sheer cash conversion. Q2 FY26 generated $10.3B in free cash flow alone — a staggering 46% margin. The VMware integration is yielding exactly as promised, driving overall company gross margins toward 70%. Even with $64B in debt on the balance sheet, a ~$27B trailing free cash flow run-rate pays it down while funding a $10B buyback and a growing dividend.
This multiple relies entirely on AI spend, but six hyperscalers dictate the entire custom silicon order book. If Meta, Google, or OpenAI suddenly moderate their CapEx in 2027, Broadcom's growth profile shatters. Meanwhile, their $64B debt pile restricts the aggressive M&A roll-up strategy that drove past decades. The EV/EBITDA multiple is nearly 58x trailing due to the VMware deal; priced for perfection.
Investors obsessed with AI silicon are ignoring the stabilizing ballast of Infrastructure Software. VMware and cybersecurity offerings now operate at software-level margins (85-90%+), insulating the company from inevitable semiconductor hardware downcycles in broadband or server storage. Broadcom isn't just an AI chipmaker; it's a deeply entrenched enterprise IT platform with unmatched networking IPs (Tomahawk/Jericho).
Despite the June 2026 post-earnings selloff to ~$360, sell-side desks overwhelmingly view the drop as an entry point. The consensus implies +46% upside.
Selection of sell-side 12-month targets out of ~50 firms covering Broadcom (AVGO). Following the Q2 FY26 earnings release, many desks reiterated "Strong Buy" ratings, interpreting the 12% drop as a valuation reset rather than a structural failure. Firms and specific targets are synthesized for illustrative purposes.
Synthesized scenario midpoints (mid-year). Returns shown vs. today's $360.45. These are illustrative frameworks, varying wildly based on how hyperscaler AI spending evolves through the decade.
Hock Tan's financial machine is built on acquiring high-margin IP and slashing bloat. Notice the gargantuan gap between Free Cash Flow (olive) and Capex (clay) — Broadcom is a capital-light software/design powerhouse, but the massive debt pile (slate) from the VMware acquisition must be serviced.
Unlike typical hyperscalers building massive AI datacenters, Broadcom's capital expenditure (clay) is virtually zero compared to its revenue. It licenses intellectual property and designs custom chips built elsewhere. Free cash flow (olive) acts as a relentless gravity, pulling down the $64B VMware acquisition debt (slate) while repurchasing shares. Figures illustrative; 2026/2027 include projected AI hyperscaler scale-up.
Targets are just a forward EPS estimate multiplied by a sentiment-driven multiple. The base case targets ~$660 by 2031, which requires Broadcom's earnings to scale gracefully to $30 over the next five years (implying a 22x exit multiple).
Adjusted (non-GAAP) split-adjusted EPS. The gray bars reflect historical actuals; the olive bars trace the base-case scenario assuming custom AI ASIC growth normalizes from its +143% explosive trajectory into sustained mid-teens growth by 2031. Note that in 2024, Broadcom enacted a 10-for-1 forward stock split; past EPS figures here are split-adjusted to align.
Reported year-over-year Q2 FY26 growth rates. The divergence between the "Core" software/networking business and the "Frontier" AI ASIC business is what fuels the debate.
The underlying core legacy businesses (broadband, server storage, basic networking) are heavily cyclical and often post single-digit or negative growth depending on the macro environment. However, VMware's software strength (olive) and the astonishing explosion of bespoke AI accelerator design (clay) for hyperscalers are masking legacy cyclicality entirely.
Is Broadcom the indispensable AI foundry, or a debt-heavy roll-up at the peak of a cyclical bubble?
The true existential threat sits squarely in the "hot" upper-right corner: what if the AI hyperscalers simply stop spending?
If Meta, Google, or OpenAI decide AI datacenter returns aren't justifying the spend, orders for Broadcom's custom silicon will evaporate.
As hyperscalers gain chip design expertise, they may eventually bypass Broadcom's design services to capture more margin internally.
Complete bans on shipping advanced AI networking or custom chips to certain regions, or disruptions in Taiwan (TSMC), paralyzing the supply chain.
A "higher for longer" rate environment makes refinancing the $64B debt pile costlier, eating into the free cash flow meant for buybacks.
Broadcom's strategy of hiking VMware prices and bundling software alienates enterprise IT departments, pushing them toward open-source or cloud-native alternatives.
Global regulators block any future sizable software or hardware acquisitions, neutralizing Broadcom's historic growth engine.
Standard down-cycles in non-AI broadband, set-top boxes, and legacy enterprise storage slightly drag on overall top-line growth.
TSMC CoWoS packaging capacity remains tight, limiting how fast Broadcom can ship completed custom silicon to eager customers.
Hover the dotted terms in the metrics, or scan the desk's working definitions here.