Meta Tests In-House AI Training Chips to Reduce Nvidia Dependence
2 min read Meta is testing an in-house chip for AI training to reduce dependence on Nvidia. The chip, made with TSMC, is in a small-scale trial, with potential expansion if successful. Unlike previous Meta chips, which only ran AI models, this one focuses on training. Meta has struggled with past chip design efforts but hopes this move will cut some of its $65 billion capital expenditure, largely spent on Nvidia GPUs. March 12, 2025 14:09
Meta is piloting its own AI training chip in a small deployment, aiming to cut reliance on Nvidia GPUs. Developed in partnership with TSMC, the chip is designed for AI-specific workloads and could scale up if successful.
Key Takeaways
✅ AI Training Focus – Unlike previous Meta chips, this one is built for training AI models.
✅ Cutting Costs – Meta plans to spend $65B on capex this year, much of it on Nvidia GPUs—own chips could save billions.
✅ Mixed Track Record – Previous in-house chip projects failed, but this move signals a fresh push toward self-sufficiency.
If successful, Meta could reshape AI infrastructure and reduce dependence on external chipmakers.
User Comments (0)
No comments added yet.
Comments will not be approved to be posted if they are SPAM, abusive, off-topic, use profanity, contain a personal attack, or promote hate of any kind.
More »
Editor's Choice
Repobase
Mootion Storyteller
Vsub
Create Studio
ChatGPTOpenAI
Midjourney
Writesonic
Jasper
Hitpaw
Fliki
AI Agents