{ }
001122334455554433221100
001122334455554433221100

Meta tests in-house AI training chip to reduce reliance on Nvidia

Meta is testing its first in-house AI training chip, aiming to reduce reliance on external suppliers like Nvidia and cut infrastructure costs. The dedicated accelerator chip, produced in collaboration with TSMC, is part of a long-term strategy to enhance AI capabilities, with plans for broader deployment if initial tests succeed. Meta's goal is to utilize these chips for training AI systems by 2026, starting with recommendation systems and expanding to generative AI products.

Meta tests in-house AI training chip to reduce reliance on Nvidia

Meta is testing its first in-house AI training chip, aiming to reduce reliance on external suppliers like Nvidia and cut infrastructure costs. The dedicated accelerator chip, produced in collaboration with TSMC, is part of a long-term strategy to enhance AI capabilities, with plans for broader deployment if initial tests succeed. Meta's goal is to utilize these chips for training AI systems by 2026, starting with recommendation systems and expanding to generative AI products.

Machinary offers a groundbreaking, modular, and customizable solution that provides advanced financial news and statistical analysis. Our platform goes beyond traditional quantitative analysis, offering users a comprehensive understanding of real-time market dynamics, event detection, and risk analysis.

Address

Newsletter

© 2025 by Machinary.com - Version: 1.0.0.0. All rights reserved

Layout

Color mode

Theme mode

Layout settings

Seems like the connection with the server has been lost. It can be due to poor or broken network. Please hang on while we're trying to reconnect...
Oh snap! Failed to reconnect with the server. This is typically caused by a longer network outage, or if the server has been taken down. You can try to reconnect, but if that does not work, you need to reload the page.
Oh man! The server rejected the attempt to reconnect. The only option now is to reload the page, but be prepared that it won't work, since this is typically caused by a failure on the server.