Meta Starts Producing Its Own Artificial Intelligence Chips in September
Meta is putting its own chips into production in September to reduce high costs and increase artificial intelligence capacity. Details are in our news.
Meta begins production of its own artificial intelligence chips in September to reduce GPU costs and combat increasing component shortages.
According to news based on an internal document, preparations have been completed for the mass production of the new generation chips designed within the scope of the Meta Training and Inference Accelerator (MTIA) program.
Meta, which collaborates with Broadcom on chip design, will work with Taiwanese giant TSMC in the production phase. This breakthrough is seen as a module of the company’s strategy to reduce its dependence on classic hardware manufacturers such as Nvidia and AMD.
- Meta will start producing its own MTIA artificial intelligence chips as of September.
- The production process will be carried out with design partner Broadcom and production partner TSMC.
- With this strategic move, the company aims to reduce high GPU costs and increase computing capacity.
Meta Provides Flexibility with Its Modular Chip Design
Meta adopts a modular approach in the chips it develops under the MTIA program.
Considering the rapidly changing nature of artificial intelligence technologies, the company uses chiplet technology to ensure that chips can adapt to future requirements more quickly.
This modular structure allows each new generation chip to be built on the previous one.
The company plans to commission 7 gigawatts of computing capacity in 2026.
Huge Investments Support the Quest for Efficiency
Meta is making a huge infrastructure investment to train and run artificial intelligence models.
The company predicts a capital expenditure of between 125 billion and 145 billion dollars for 2026, and uses a significant part of this budget to increase its data center capacity. The distribution and development of the Muse Spark series artificial intelligence models are at the center of this intensive infrastructure work.
Industry Giants Are Developing Their Own Chip Solutions
Not only Meta, but also technology giants such as Amazon, Google and OpenAI are turning to their own chip solutions to reduce dependence on Nvidia. While OpenAI is working with Broadcom on an inference processor, companies such as Anthropic are also evaluating similar initiatives. This shows that the fight for hardware dominance in the world of artificial intelligence is intensifying.
How do you think technology giants designing their own chips will affect Nvidia’s dominance in the market in the coming years? Share your ideas with us in the comments section.
