Meta plans to put its own MTIA chip into production in September and is aiming to roughly double the computing capacity across its data centres, Reuters reported on Thursday, citing people familiar with the matter. Meta declined to comment on the specifics, and both the timing and the capacity target come from anonymous sources rather than any public disclosure.
The MTIA chip belongs to Meta’s in-house silicon line, the Meta Training and Inference Accelerator, which the company has been scaling up as part of a record spending push and a wider effort to cut its reliance on Nvidia. If the September timeline holds, it would extend a programme that has already moved fast this year: Meta unveiled four new MTIA chips in March, the 300, 400, 450, and 500, and said it would ship them on a roughly six-month cadence rather than the annual pace common across the industry.
Why the MTIA chip matters to Meta’s Nvidia bill
Meta remains one of Nvidia‘s biggest customers, buying large numbers of GPUs to train its Llama models and to run recommendation systems for more than 3 billion daily users. Every workload it can shift onto its own silicon is one it does not have to buy at Nvidia’s margins. So far, the MTIA chip has largely handled inference, the day-to-day job of serving predictions once a model has been trained. The MTIA 300 is already in production for ranking and recommendation work, while the 450 and 500, built for generative image and video inference, are slated for mass deployment through 2027.
A training-capable MTIA chip would be a harder test. Training frontier models is the workload where Nvidia’s hardware and its CUDA software have proved stickiest, and where in-house alternatives from Google and Amazon have taken years to mature. Reuters’ report does not specify which MTIA generation enters production in September, nor how any doubling of capacity would be split between new chips and additional data-centre floor space.
A hedge across several chip suppliers
The chips are manufactured by TSMC and co-developed with Broadcom, whose partnership with Meta now runs through 2029 and covers several generations of custom silicon. Broadcom has said the newer MTIA parts will be among the first custom AI chips built on a 2-nanometre process.
The capacity ambition sits inside a large spending plan. Meta has guided 2026 capital expenditure to between $125 billion and $145 billion, with nearly all of the increase going toward data centres, GPUs, and custom silicon, and Mark Zuckerberg has floated eventual targets measured in gigawatts. That build-out has grown large enough that Meta is now looking to rent out spare compute to outside customers, a model long used by cloud providers. The company has also signed deals for Amazon’s Graviton5 chips and AMD accelerators alongside its standing Nvidia orders.
Custom silicon changes the underlying economics because a chip designed for the exact models Meta runs can cut power draw and unit costs at scale, provided the software stack keeps pace. Even so, the MTIA chip is unlikely to displace Nvidia outright. Analysts tend to frame it as a way to absorb growth and trim the GPU bill at the margins, not to replace Nvidia in the near term, and Meta keeps expanding its GPU commitments even as it ramps up custom parts. September, if the reporting holds up, would be the next milestone in that effort, with the training workload still the one Nvidia has yet to give up.