Broadcom is the dominant enabler of custom AI silicon, holding ~60% of the AI server compute ASIC design market through 2027 (Counterpoint Research). It now serves 6 hyperscaler XPU customers -- Google (TPU), Meta (MTIA), ByteDance, OpenAI (Titan), Anthropic (via Google TPU), and one undisclosed -- with each of the three original customers planning 1-million-XPU clusters by 2027. AI semiconductor revenue has followed a steep trajectory: $12.2B in FY2024 to ~$20B in FY2025 (+63%) to a projected ~$40-50B in FY2026 (~100-150% YoY).
The $73B AI order backlog (deliverable over 18 months) provides strong near-term revenue visibility. CEO Hock Tan has stated 'clear line of sight' to $100B+ in annual AI chip revenue by FY2027. For NVIDIA, Broadcom's XPU business represents the primary channel through which custom silicon threatens GPU market share: every dollar hyperscalers spend on Broadcom-designed ASICs is a dollar not spent on NVIDIA GPUs. However, Broadcom's networking portfolio (Tomahawk switches, Jericho routers, optical components) also grows alongside GPU deployments, creating a partially hedged relationship..
Competitive pressure is real but bounded
Custom ASICs and AMD offer cheaper alternatives for specific workloads, but only a handful of companies can afford multi-billion-dollar chip programs. The competitive threat is structural but limited in scope.
What fraction of Broadcom's $73B AI backlog converts to actual revenue vs. being revised or deferred? Backlog ≠ guaranteed revenue.
Stub: Deep-dive into each of Broadcom's 6 hyperscaler XPU customer programs. Google TPU (largest, ~$22B FY2026 revenue, TPU v7 Ironwood); Meta MTIA (4 generations in 24 months, RISC-V, 25x compute increase); OpenAI Titan (TSMC 3nm, Samsung HBM4, 10GW deployment); ByteDance (undisclosed scale); Anthropic (via Google TPU, 1GW->3GW+). Each program's maturity, volume trajectory, and displacement of NVIDIA GPU purchases..
Competitive pressure is real but bounded
Custom ASICs and AMD offer cheaper alternatives for specific workloads, but only a handful of companies can afford multi-billion-dollar chip programs. The competitive threat is structural but limited in scope.
Stub: Broadcom's AI networking business -- Tomahawk 6 (102.4 Tbps, world's first), Jericho 4, optical DSPs, lasers, PCIe components -- represented ~33% of Q1 FY2026 AI revenue ($2.8B) and is growing to ~40% in Q2. ~$20B of the $73B AI backlog is non-accelerator networking content. Key question: does Broadcom's networking revenue grow regardless of GPU vs ASIC mix, partially hedging against NVIDIA GPU displacement?.
Competitive pressure is real but bounded
Custom ASICs and AMD offer cheaper alternatives for specific workloads, but only a handful of companies can afford multi-billion-dollar chip programs. The competitive threat is structural but limited in scope.
Stub: TSMC CoWoS advanced packaging is the primary supply bottleneck for both NVIDIA GPUs and Broadcom ASICs. NVIDIA secures ~60% of CoWoS capacity; Broadcom ~15% (150K wafers: 90K Google, 50K Meta, 10K OpenAI). CoWoS oversubscribed through 2026 with >85% locked by major customers.
TSMC ramping from ~40K wafers/month (2024) to 140-150K by 2026. Supply constraints may actually gate ASIC deployment speed, giving NVIDIA time advantage..
Competitive pressure is real but bounded
Custom ASICs and AMD offer cheaper alternatives for specific workloads, but only a handful of companies can afford multi-billion-dollar chip programs. The competitive threat is structural but limited in scope.
Stub: Broadcom holds ~60% of AI ASIC design market vs Marvell's ~8% (projected 2027, Counterpoint). Marvell claims 20+ ASIC design wins but lost lead partner status for AWS Trainium to AIchip (Taiwan). Google-MediaTek alliance emerging as potential threat to Broadcom's dominance.
Competitive pressure is real but bounded
Custom ASICs and AMD offer cheaper alternatives for specific workloads, but only a handful of companies can afford multi-billion-dollar chip programs. The competitive threat is structural but limited in scope.