Meta Platforms (META) is quietly testing a secret weapon against Nvidia (NVDA). The company has been developing its own AI training chip, a move that could change up the AI hardware market and slash its sky-high costs.
The company is testing its first in-house AI training chip. The goal is clear—cut massive infrastructure costs. At the same time, Meta is pushing deeper into artificial intelligence, solidifying its position in the race for AI dominance. This exclusive report from Reuters reveals that Meta has already begun small-scale deployment of the chip. If all goes well, the company plans to ramp up production and take another step toward AI independence.
Meta Aims to Reduce AI Costs with Custom Silicon
AI development isn’t cheap. Meta has projected up to $119 billion in total expenses for 2025, with $65 billion dedicated to AI infrastructure. Meta has been pouring billions into Nvidia’s GPUs, effectively renting the keys to its AI future. But now, it’s trying to take control of its own destiny. By designing its own AI training chip, Meta hopes to cut costs and optimize performance for tasks like recommendation systems and generative AI.
Meta’s Chip Faces High Stakes Testing
The chip, part of Meta’s Training and Inference Accelerator (MTIA) series, is being manufactured by Taiwan’s TSMC (TSM). According to Reuters, Meta recently completed a crucial “tape-out” phase—an expensive first step in chip development.
But success isn’t guaranteed. If the chip flops, Meta will be forced to go back to square one, costing millions and months of delays. If the test fails, Meta will need to diagnose issues and start over, a process that can take months and cost tens of millions of dollars.
Meta Aims to Integrate Chip in AI Training by 2026
Meta wants to fully integrate this chip into its AI training systems by 2026, according to Reuters. If it pulls this off, it could loosen Nvidia’s grip on the AI market and set a new standard for Big Tech’s AI ambitions. Chief Product Officer Chris Cox described the effort as a “walk, crawl, run situation,” acknowledging past setbacks. While Meta scrapped an earlier chip, its new inference chip has been a success, powering recommendations across Facebook and Instagram.
If this training chip works, Meta could reduce its reliance on Nvidia’s costly GPUs and take a major step toward controlling its AI infrastructure.
Is META Stock a Good Buy Now?
Analysts remain bullish about META stock, with a Strong Buy consensus rating based on 44 Buys, three Holds, and one Sell. Over the past year, META has increased by more than 20%, and the average META price target of $764.61 implies an upside potential of 23.4% from current levels.
Questions or Comments about the article? Write to editor@tipranks.com