Meta’s new AI chips will begin production in September

In an effort to reduce GPU costs amid an unprecedented component shortage, Meta is on track to start making the latest versions of its AI-specific chip in September, according to Reuters reportedciting an internal memo.

At least one chip went through the testing phase in about six weeks, the memo said. Meta is working with Broadcom on the chip design, but will use Taiwan’s TSMC for production. It also buys RAM from Samsung, storage from Sandisk and fiber optic equipment from Sumitomo Electric, according to the report.

Meta detailed the four new chips, developed under the Meta Training and Inference Accelerator (MTIA) program, in March, some of which are currently in use or will be this year or next year. The company is taking a modular approach to designing these chips and anticipates that their needs will change as AI rapidly evolves by the time the chips are in production.

“Each MTIA generation builds on the previous one, uses modular chiplets, integrates the latest AI workload insights and hardware technologies, and deploys at a shorter cadence,” the company wrote at the time.

The chips are expected to help the company save on purchasing GPUs from chipmakers like Nvidia and AMD, although it still expects to spend a lot of money on those providers as well, Reuters reports. Meta plans to use the MTIA chips for training models for its ranking and recommendation algorithms, broader AI workloads, and inference focused on its applications. The social media company is done has been producing its own AI chips since 2023.

Meta has spent a huge amount of money securing enough computing capacity to power its various AI efforts. The company said this in April expected capital expenditures between $125 billion and $145 billion this year, much of which will go to AI efforts.

The company has signed data center and power deals around the world and spent tens of billions to secure computing capacity to train and deploy its new Muse spark range of AI models. It plans to deploy 7 gigawatts of computing power this year and then double that, according to Reuters, which cited the memo.

Last year it also signed a deal with ARM to secure computing capacity for its recommendation systems, in addition to a multi-billion dollar deal with AMD for its Instinct GPUsand a multibillion-dollar deal with Amazon to use the cloud giant’s CPUs for AI-related needs.

Meta isn’t the only company trying to stem the flow of capital to Nvidia. OpenAI last month unveiled an inference processor it is building with Broadcom, and Anthropic is said to be considering developing its own chips with Samsung. Amazon and Google are both developing their own chips for AI training and inference, and there are a host of startups building in the space to meet skyrocketing demand.

Meta declined comment.