Meta plans to start production of its latest AI-specific chips in September to reduce GPU costs amid a component shortage. These chips, developed under the Meta Training and Inference Accelerator (MTIA) program, will help the company optimize its AI workloads and deployments.
Meta is set to commence production of its new AI chips in September, as revealed in an internal memo. The first chip reportedly passed testing in six weeks, indicating readiness for deployment soon.
These chips are part of the Meta Training and Inference Accelerator (MTIA) program, which employs a modular design approach to adapt to the rapidly evolving AI landscape. Broadcom is collaborating with Meta on the chip design, while TSMC will handle manufacturing.
By developing its own AI chips, Meta aims to cut costs associated with purchasing GPUs from companies like Nvidia and AMD. The company plans to leverage these chips for various AI tasks including model training for ranking algorithms, inference, and broader AI applications.
Meta has announced a substantial capital expenditure plan of $125 billion to $145 billion for 2023, primarily focusing on AI initiatives. This includes deals for data center capacity and significant investments in compute power to support advanced AI models, such as its Muse Spark series.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
Meta plans to start production of its latest AI-specific chips in September to reduce GPU costs amid a component shortage. These chips, developed under the Meta Training and Inference Accelerator (MTIA) program, will help the company optimize its AI workloads and deployments.