Observed Signal · Jul 9, 2026 · Production Start · Source: techcrunch · Impact: 4/5 · Sentiment: Positive
Meta to Begin MTIA AI Chip Production in September
Meta plans to begin production of its latest AI-specific chips in September, according to a Reuters-cited internal memo. The chips are part of Meta’s Meta Training and Inference Accelerator (MTIA) program and use a modular chiplet approach; at least one chip completed testing in about six weeks. Meta worked with Broadcom on the chip design and will have Taiwan’s TSMC manufacture the chips, while sourcing RAM from Samsung, storage from Sandisk, and fiber-optic equipment from Sumitomo Electric. Meta expects the MTIA chips to reduce some GPU purchases from vendors like Nvidia and AMD and will use them for training and inference across ranking, recommendation and broader AI workloads. The article is dated 2026-07-09.
Meta — a major platform — moving its own AI chip production into manufacturing reduces reliance on third-party GPUs, affects chip demand dynamics (Nvidia/AMD), and expands Meta's capacity for AI training/inference; this is a significant infrastructure development for AI at scale.
Track Meta Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Meta plans to start making the latest versions of its AI-specific chips in September 2026, per an internal memo reported by Reuters.
- Meta worked with Broadcom on the chip design and will use Taiwan’s TSMC to manufacture the chips.
- Meta is sourcing RAM from Samsung, storage from Sandisk, and fiber-optic equipment from Sumitomo Electric for the chip production.
- The chips are part of Meta’s Meta Training and Inference Accelerator (MTIA) program and use a modular chiplet approach.
- Meta expects the MTIA chips to help reduce some GPU purchases from vendors like Nvidia and AMD and to be used for training and inference for ranking, recommendation and other AI workloads.
Connected Companies & Entities
12 Entities mapped“In a bid to lower its GPU costs amid an unprecedented component shortage, Meta is on track to start making the latest versions of its AI-spe...”
“In a bid to lower its GPU costs amid an unprecedented component shortage, Meta is on track to start making the latest versions of its AI-spe...”
“Meta is working with Broadcom on the chip design, however it will use Taiwan’s TSMC to manufacture them....”
“Meta is working with Broadcom on the chip design, however it will use Taiwan’s TSMC to manufacture them....”
“It is also buying RAM from Samsung, storage from Sandisk, and fiber optic equipment from Sumitomo Electric, according to the report....”
“The chips are expected to help the company save on buying GPUs from chipmakers like Nvidia and AMD, although it still expects to spend plent...”
“The chips are expected to help the company save on buying GPUs from chipmakers like Nvidia and AMD, although it still expects to spend plent...”
“Meta ... a multi-billion dollar deal with Amazon to use the cloud giant’s homegrown CPUs for AI-related needs....”
“OpenAI last month unveiled an inference processor that it is building with Broadcom, and Anthropic is said to be considering developing its ...”
“Amazon and Google both develop their own chips for AI training and inference, and there’s a host of startups building in the space to meet s...”
“OpenAI last month unveiled an inference processor that it is building with Broadcom, and Anthropic is said to be considering developing its ...”
“Title: Meta’s new AI chips will begin production in September (article published on TechCrunch)....”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Meta Unveils Custom AI Chips Amid Nvidia, AMD Partnerships
Meta revealed four custom in-house AI chips in its MTIA (Meta Training and Inference Accelerator) family as part of a rapid data-center expansion. Meta has deployed MTIA 300 (for training smaller models used in ranking, recommendations and ad delivery) and completed testing of MTIA 400, which is optimized for generative-AI inference and is slated for near-term deployment; MTIA 450 and MTIA 500 are planned to be operational in 2027. Meta said the chips are manufactured by Taiwan Semiconductor and that one data-center rack will hold 72 MTIA 400 chips. Meta framed the custom silicon as a way to improve price/performance, diversify silicon supply and hedge against vendor price changes, while noting concerns about securing high-bandwidth memory (HBM). The company has also signed large multi-year deals for Nvidia and AMD GPUs to preserve options.
Meta Publishes MTIA Inference-Optimized Chip Roadmap
Meta published a technical blog outlining a multi‑generation MTIA roadmap that includes four inference‑optimized chips (MTIA 300/400/450/500) shipped or planned within roughly two years. The post emphasizes inference-first design (optimized for GenAI inference and recommendation workloads) and transparent specs/timelines, endorsing a shift from a GPU‑only strategy toward purpose‑built silicon. The roadmap leverages chiplet modularity to enable faster iteration and mixed process nodes, and Meta highlights software compatibility (PyTorch/Triton/vLLM) and a CUDA‑compatible stack from partners. The update reinforces the business case that custom inference silicon can expand margins for advertising and cut costs for high‑token GenAI usage. The article notes supply‑chain and competitive implications for Broadcom, Nvidia, AMD, HBM suppliers, TSMC, Arista, and inference chip startups, and references Meta’s prior acquisition and collaboration with Rivos.
Meta, Broadcom Agree on 1GW Custom AI Chip Deal
Meta and Broadcom announced a multi-year partnership extending through 2029 for the design of Meta's custom in-house AI accelerators. Meta committed to an initial deployment of 1 gigawatt of its Training and Inference Accelerators (MTIA) and plans to scale to multiple gigawatts in 2027 and beyond. Broadcom said MTIA chips will be the first AI silicon using a 2-nanometer process. Broadcom CEO Hock Tan informed Meta he will not stand for reelection to Meta’s board; Tracey Travis will also leave the board. The agreement follows Broadcom’s recent deals with Google and Anthropic and comes as Meta pursues large-scale AI investments while expanding its data center footprint.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.