NVIDIA vs Broadcom

The widest gap in this series: 26 points. Both companies sell the picks and shovels of the AI buildout, and Broadcom is growing at a rate that would top almost any other card. The distance between them is the distance between owning the platform and buying into it.

The success line is 65. Demerits (section D) are subtracted; a demerit total of 15 or more flags a company for BlitzFailing watch. Scores as of the 2026 Q4 review (October 21, 2026).

Speed — elite against historic

Broadcom’s speed card (20) would win most pairings: Q3 FY2026 revenue of $29.59B grew +85.5%, AI semiconductor revenue of $16.7B grew +221% for a thirteenth straight quarter of AI acceleration, and Q4 guidance implies roughly +93%. NVIDIA still takes the section 23–20, because its growth is at ten times the revenue base — trailing revenue of about $303B at +83%, with Q2 FY2027 up +106% — and its productivity is in a class of one: about $7.2M of revenue per employee across roughly 42,000 staff (A3: 5 vs 3), against Broadcom’s elite but acquisition-swollen $1.94M. Both score 9 or 10 on growth itself; the rubric separates them on what the growth is made of.

Moat — CUDA against the bundle

The moat gap, 52 vs 36, is the largest in this series and it is one item wide at the top: network effects, 16 vs 7. NVIDIA’s B1 cites 7.5 million CUDA developers, twenty years of accumulated libraries and framework-standard status — the anchor table’s own example of the 15–18 band. Broadcom’s scorecard states flatly that it has no true network effect; its 7 is bundle lock-in (over 90% of VMware’s top 10,000 customers buying the VCF bundle, per-core subscription conversion, price rises of 100–300% absorbed without mass exit — powerful, but a toll, not a flywheel). Distribution follows the same shape (14 vs 9): NVIDIA sells into every hyperscaler, sovereign AI program and neo-cloud simultaneously, while Broadcom runs a direct-plus-channel motion. Even gross margin separates them at this altitude — 75% GAAP at $300B of revenue (B4: 12, a perfect score) against Broadcom’s 69.1% record quarter (10).

Destruction and demerits — the completed blitz, and its price

NVIDIA is the only company in the index with a perfect creative-destruction card (15): the CPU-centric data center has been replaced inside the AI domain, industry capex has irreversibly reorganized around accelerators, and each silicon generation has cut inference cost per token by up to ten times on the company’s figures. Broadcom’s 10 describes a slower, contested reshuffle — it is re-pricing the virtualization order it acquired rather than replacing an industry’s physics, with Gartner expecting only about a third of VMware workloads to leave by 2028. Demerits are modest on both cards but differently sourced: NVIDIA’s −3 is geopolitics (China data-center revenue guided to zero under export licensing, plus two DOJ inquiries), while Broadcom’s −5 combines acquisition leverage ($59.4B of borrowings, −3) with the VMware pricing litigation cluster (−2). Twenty-six points is what the rubric sees between the platform and its largest custom-silicon challenger — Broadcom’s second blitz is real, underway, and still graded as a challenger’s.

How to read this

Both companies passed the same two gates before scoring: real product-market traction (G1), and a ten-year market ceiling of at least ten times current revenue (G2). The Blitz Index then grades speed (25), moat (60) and creative destruction (15), and subtracts BlitzFailing demerits (up to −30) for debt, revenue quality, governance and regulatory cost. The model is frozen at v1.0; past scores are never silently rewritten. Our own pre-registered tests found the score measures business quality — it does not predict stock returns.