Meta Platforms plans to begin manufacturing its in-house artificial intelligence chip in September as part of a strategy to double its computing capacity to 14 gigawatts in 2027, according to an internal company memo reviewed by Reuters, in a move designed to strengthen its AI infrastructure while reducing reliance on external chip suppliers.

According to Reuters, the chip, code-named Iris, forms part of Meta's four-generation Meta Training and Inference Accelerators (MTIA) programme, which aims to deliver custom silicon for the AI systems powering Facebook and Instagram. The memo said testing of the processor was completed in six weeks without major issues, marking faster-than-expected progress for an initiative that has struggled to gain momentum since it began more than five years ago. Meta declined to comment.

The memo said Iris is intended to complement, rather than replace, the large volumes of graphics processing units Meta buys from Nvidia and Advanced Micro Devices. It added that integrating the latest GPUs across Meta's infrastructure "has been a heavy lift, and it has cost us time," highlighting one reason the company is investing in proprietary hardware.

Reuters reported that Meta is developing the chip with Broadcom and will use Taiwan Semiconductor Manufacturing Co to manufacture it. The company revealed Iris under its technical name in March alongside three other AI processors and plans to introduce a new chip roughly every six months through 2027, a faster cadence than the annual release cycles common across the semiconductor industry.

The internal memo showed Meta expects to deploy seven gigawatts of computing infrastructure during 2026 before doubling capacity to 14 gigawatts next year. The company is expected to spend as much as $145 billion on AI infrastructure this year, representing a significant share of the more than $700 billion Big Tech companies are projected to invest in artificial intelligence.

Reuters reported that Meta has secured long-term supply agreements to support that expansion, including deals with Samsung Electronics for memory chips, Sandisk for flash storage and Sumitomo Electric for fibre-optic equipment. Sandisk declined to comment, while Samsung Electronics and Sumitomo Electric did not respond to requests for comment.

Morgan Stanley analysts said rising demand for AI chips and memory has pushed component prices sharply higher, creating concerns about "chipflation" across the wider economy.

---