Reuters reports that Meta is testing its first “home-made” chip for training artificial intelligence systems. The move is essential for developing its silicon solutions and reducing its dependence on external suppliers such as Nvidia.
The Facebook parent company is currently installing the chip on a trial basis. If the tests are successful, production will be expanded.
Meta’s long-term plan to reduce its enormous infrastructure costs involves developing its chips. The company is also actively investing in artificial intelligence tools.
According to the source, the new chip is a “special accelerator” designed exclusively for AI tasks and is more energy-efficient than commonly used graphics processing units (GPUs).
Meta has forecasted $114-119 billion in spending for 2025, of which $65 billion is primarily for capital expenditures on AI infrastructure.
