Hyperscalers accelerate custom AI silicon designs across Broadcom, Google, Meta
Major hyperscalers are increasingly designing their own custom AI silicon to power intelligence applications, representing a collective and accelerating investment trend. This development includes efforts from companies like Broadcom, Google with its TPUs, and Meta with MTIA. The article examines the current landscape of custom AI ASICs as of May 2026.
Key Takeaways
- Broadcom is part of the custom AI ASIC landscape now being tracked across hyperscalers.
- Google continues to use TPUs as its custom AI silicon strategy.
- Meta’s MTIA is another named custom AI chip effort in the mix.
- The article says every major hyperscaler now designs its own AI silicon.
- Collective investment in custom AI silicon is accelerating as of May 2026.
Why It Matters
The immediate takeaway is that custom AI silicon has moved from a niche strategy to a standard hyperscaler play, with Broadcom deals, Google TPUs, and Meta MTIA all part of the current landscape. That matters because it shows intelligence workloads are increasingly being tied to in-house ASIC design rather than generic compute. The competitive signal is less about one company and more about the sector-wide buildout the article describes. Watch for new Broadcom deal disclosures or additional named TPU and MTIA deployments in hyperscaler product updates.
Read full article at tomshardware.com