AVGO
Broadcom
Summary
What they do:
Design custom AI accelerator chips (Google TPU, Meta MTIA, plus 3+ additional hyperscaler programs), networking ASICs (Tomahawk switches), and run VMware's enterprise software — sitting at Layer 06 as the dominant custom silicon design house and the networking fabric that connects AI clusters.
Why they matter:
Broadcom is the only company designing custom AI chips for multiple hyperscalers simultaneously, with a $73B AI backlog, 5 confirmed ASIC customers, and AI semiconductor revenue doubling YoY — while also controlling the Tomahawk switching silicon that routes data inside virtually every major data center.
Recent performance:
Q1 FY2026 revenue $19.3B (+29% YoY), AI semiconductor revenue $8.4B (+106% YoY). Q2 guide $22B (+47% YoY). Stock at ~$400, market cap ~$1.9T. Next earnings June 2026.
Our Verdict
The dominant custom AI chip designer with 106% AI revenue growth, $73B backlog, and line-of-sight to $100B+ AI revenue by 2027 — at ~35x forward earnings, one of the few mega-cap AI stocks where growth still justifies the multiple.
Structural trends
Structural
87
/ 100
Moat
9/10
Custom ASIC dominance — 5+ hyperscaler programs, $73B backlog, largest design force globally
AI Exp.AI Exposure
Pure Play~65% AI
Play Type
EstablishedAI Growth
~106%
Rel. Value
48
FAIRPriceLIVE
$380.78
+0.27%
Live via Yahoo Finance · refreshes every 5 min
Market Cap
$1.8T
P/E Ratio
74.1
P/S Ratio
26.4x
52W High
$414.61
52W Low
$161.61
52W Chg
135.6%
Beta
1.25
Broadcom is the architect of custom silicon for the world's largest AI companies. When Google needs a next-generation TPU optimized for transformer training, or Meta needs an inference accelerator tuned for Llama's specific architecture, they contract Broadcom to design the chip. Broadcom's engineers write the RTL, manage the physical design, coordinate with TSMC for fabrication, and deliver a finished chip design that the hyperscaler deploys at million-unit scale. This is not contract manufacturing — it is intellectual partnership at the bleeding edge of silicon design.
The custom ASIC business has exploded. Q1 FY2026 AI semiconductor revenue hit $8.4B (+106% YoY), driven by accelerating demand from Google, Meta, and at least three additional confirmed hyperscaler programs (widely believed to include ByteDance, OpenAI/Microsoft, and a fifth undisclosed customer). Management has line-of-sight to AI chip revenue exceeding $100B in 2027. The AI backlog stands at $73B — the largest confirmed AI order book in the semiconductor industry.
Broadcom's second pillar is networking silicon. Tomahawk switch ASICs sit inside virtually every major data center switch, routing packets at 51.2+ terabits per second. As AI clusters scale from hundreds to thousands of GPUs, the network fabric connecting them becomes critical — and Broadcom's Tomahawk and Jericho ASICs are the dominant merchant solution. Networking revenue compounds alongside AI cluster deployments.
The third pillar is VMware, acquired for $61B in 2023. VMware's virtualization and cloud management software generates ~$6.8B per quarter in recurring revenue with high retention. While VMware growth is flat, it provides stable free cash flow that funds semiconductor R&D and reduces overall business volatility.
Q1 FY2026 total revenue was $19.3B (+29% YoY). Semiconductor solutions revenue was $12.5B (+52% YoY). Adjusted EBITDA was $13.1B at a 68% margin. The company employs ~7,000 chip engineers — the largest custom ASIC design force in the world outside of the hyperscalers themselves.
Supply Chain Dependencies
Upstream Suppliers
The Catch
Broadcom's custom ASIC dominance rests on a fragile paradox: its best customers are also building the capability to replace it. Google has hundreds of chip engineers who learn from each TPU generation designed with Broadcom. Meta is investing billions in in-house silicon design. The very act of designing custom ASICs for hyperscalers trains those customers to eventually design their own. This dynamic is 3–5 years from materializing — current backlog is secure — but the long-term question is whether Broadcom remains the architect or becomes the contractor. Meanwhile, customer concentration is extreme: an estimated 50–60% of AI semiconductor revenue comes from 2 customers. And the $100B 2027 AI revenue target, while management-confirmed, requires every hyperscaler ASIC program to ramp on schedule simultaneously — a feat of execution coordination that has never been attempted at this scale.
If They Win
If Broadcom executes on the $100B AI revenue target — if all 5+ ASIC programs ramp as guided, if Tomahawk networking captures the AI fabric opportunity, if VMware provides the stable cash flow base — Broadcom becomes the semiconductor industry's second trillion-dollar-revenue company after NVIDIA. AI semiconductor revenue of $100B+ at 70%+ EBITDA margins generates $70B+ in annual EBITDA. Total company revenue approaches $130-140B. The custom ASIC model proves superior to merchant GPU for inference-heavy workloads, and hyperscalers deepen their Broadcom dependencies rather than pulling design in-house. Market cap reaches $3T+ as the market recognizes Broadcom as the critical complement to NVIDIA — NVIDIA for training, Broadcom for custom inference and networking. The VMware software base provides downside protection that pure-play semiconductor companies lack. Broadcom becomes the most important semiconductor company you've never heard of.
Others in The Finished Chip
Not financial advice. All scores generated via AI algorithms using public data.