Overview
Meta is set to begin production of its latest AI-specific chips in September 2026, according to an internal memo seen by Reuters. The development marks a strategic move to lower GPU costs amid an unprecedented component shortage. The memo reveals that at least one chip completed its testing phase in approximately six weeks.
Chip Details and Partners
Design and Manufacturing
Meta has partnered with Broadcom for the chip design, while Taiwan Semiconductor Manufacturing Company (TSMC) will handle manufacturing. The company is also sourcing RAM from Samsung, storage from Sandisk, and fiber-optic equipment from Sumitomo Electric, as reported by Reuters.
MTIA Program and Modular Approach
The four new chips, developed under Meta’s Training and Inference Accelerator (MTIA) program, were announced in March 2026. Some are currently in deployment or will be deployed this year or next. Meta is taking a modular approach, designing chips with chiplets to adapt to rapidly evolving AI needs. The company stated: “Each MTIA generation builds on the last, using modular chiplets, incorporating the latest AI workload insights and hardware technologies, and deploying on a shorter cadence.”
Impact on GPU Spending
The MTIA chips are expected to help Meta save on buying GPUs from Nvidia and AMD, though the company will still spend heavily with those providers. Meta intends to use the chips for training models for its ranking and recommendation algorithms, broader AI workloads, and inference for its applications. Meta declined to comment on the report.
Meta’s AI Infrastructure Investments
Capital Expenditure and Compute Plans
Meta has been spending massively to secure compute capacity. In April 2026, the company said it expects capital expenditures between $125 billion and $145 billion this year, a large portion dedicated to AI. The company has struck data center and power deals globally, spending tens of billions. According to the memo, Meta plans to deploy 7 gigawatts of compute this year and double that in 2027.
Meta also struck a deal with ARM last year for compute for its recommendation systems, in addition to multibillion-dollar deals with other companies for AI needs. Meta is not alone; OpenAI last month announced an inference processor built with Broadcom, and Anthropic is considering a partnership with Samsung. Other tech giants also develop their own AI chips, and there is a growing market to meet skyrocketing demand.