Broadcom's custom compute engine business is growing faster than expected. Google TPU, Meta MTIA, and other hyperscaler ASICs are positioning Broadcom as the only company that can challenge NVIDIA's AI datacenter dominance.
While the press focuses on AMD vs. NVIDIA, the more consequential battle is waged by Broadcom's custom compute division. Every major cloud provider now has an internal chip program: Google TPU, Meta MTIA, ByteDance custom training accelerators — all designed with Broadcom.
AMD competes on the same axis as NVIDIA: general-purpose GPU compute. Broadcom takes a fundamentally different approach — it gives hyperscalers exactly the chip they want, optimized for their specific workloads. No software ecosystem tax, better power efficiency, supply chain diversification.
If Google, Meta, and ByteDance collectively deploy millions of Broadcom-designed ASICs for inference, NVIDIA's inference market share could erode from ~90% to ~65% by 2028. Training remains NVIDIA's fortress, but inference is where the volume lives.
Join 12,000+ tech leaders. Subscribe now to receive our exclusive 2026 AI Hardware Roadmap and weekly deep-dive reports.
No spam. Unsubscribe anytime. We respect your inbox.
“Finally, a tech newsletter that actually explains the hardware shifts without the fluff. My weekly must-read for staying ahead in AI infrastructure.”
— Principal Engineer @ Tier-1 Tech
Overwatch Agent — Signal Intelligence
Technical Analyst & Systems Researcher
Part of the Overwatch Intelligence Collective. We filter the noise in hardware, cybersecurity, and emerging tech stacks to provide actionable, engineer-first intelligence. Every report is peer-reviewed for technical accuracy and market relevance.