Meta Speeds Up Custom AI Chip Production to Cut GPU Costs
Curated by the Inblix editorial team
Meta is accelerating its plan to produce its latest AI chips, aiming to start manufacturing in September to reduce reliance on expensive GPUs from Nvidia and AMD. The chips, part of its MTIA program, are designed with Broadcom and will be made by TSMC, with components from Samsung, Sandisk, and Sumitomo. One chip passed testing in about six weeks, signaling rapid progress. Meta uses a modular approach to adapt to fast-evolving AI needs. While these chips will help save on GPU purchases, Meta still plans heavy spending on third-party chips. The company is investing billions in data centers and power deals, targeting 7 gigawatts of compute this year and double that next. Meta joins a trend of tech giants like OpenAI, Amazon, and Google developing custom silicon to cut costs and boost performance. Why it matters: This move highlights how AI leaders are racing to control their hardware supply chains, challenging Nvidia’s dominance and reshaping the economics of AI infrastructure.
💡 Key Takeaways
- Meta will start making its latest MTIA AI chips in September to lower GPU costs amid a component shortage.
- The chips, designed with Broadcom and built by TSMC, passed testing in six weeks and use a modular design for adaptability.
- Meta plans to double its compute capacity to 14 gigawatts next year, spending tens of billions on infrastructure.
Keep reading: See related articles below for more coverage on this topic.
Get smarter about AI
The sharpest AI news, curated daily. Delivered free to your inbox.