Meta (Nasdaq: meta) just do A bold decision in AI ratifying his first internal training chip in order to reduce dependence on Nvidia (Nasdaq: NVDA) and lower the high -high infrastructure costs. The company has burned money on AI, projecting up to 65 billion dollars in capital expenditure for 2025. Now, it changes speed, aimed at fueling its recommendation engines and generative AI with personalized silicon. The chip is currently in small -scale deployment, and if everything is going well, Meta could accelerate production quickly. Partnership with Taiwan Semiconductor Manufacturing (NYSE: TSM), the technology giant seeks to optimize efficiency and gain greater control of its AI battery.
Meta played this game before the previous personalized inference chip does not cut, forcing a shopping trip of several billion dollars on the NVIDIA GPUs. But things are different this time. The company’s MTIA chip for inference has already proven a success in Facebook and Instagram recommendation systems. Now he attacks the formation, the true power of AI. Chris Cox, Meta Products director, calls for a walk, crawl, execute processes consuming that they are still in the early stages, but moving in the right direction. If this new chip delivers, it could shake up the material landscape of AI, which questions Nvidia’s workforce on the market.
The stakes are high, just like skepticism. The AI industry is at the crossroads of the scale of large models with more GPU is no longer the only game in town. Meta’s personalized chip strategy is a bet on efficiency, but will it pay? Investors look closely. If Meta realizes this, it could point out a seismic lag in AI computer science. Otherwise, it could eventually double on Nvidia.
This article appeared for the first time on Gurufocus.