Meta Platforms plans to begin manufacturing a new artificial intelligence chip in September as it works to double its computing capacity to 14 gigawatts in 2027. The chip, code-named Iris, is part of a four-generation project under Meta’s Training and Inference Accelerator program. Meta is designing the processors internally to support the AI systems used across Facebook, Instagram, and its other services, according to an internal memo reviewed by Reuters .

Meta Platforms plans to begin manufacturing a new artificial intelligence chip in September as it works to double its computing capacity to 14 gigawatts in 2027.

The chip, code-named Iris, is part of a four-generation project under Meta’s Training and Inference Accelerator program. Meta is designing the processors internally to support the AI systems used across Facebook, Instagram, and its other services, according to an internal memo reviewed by Reuters .

Meta completed bug testing for Iris in six weeks and found no major problems, according to the memo .

The rapid testing represents positive progress for Meta’s custom-chip program, which has faced difficulties since the company began working on in-house processors more than five years ago.

Meta declined to comment on the previously unreported testing results and September production schedule.

Meta developed Iris around its own AI workloads and is working with Broadcom on the chip’s design.

Taiwan Semiconductor Manufacturing Company will manufacture the processor. Meta expanded its custom-chip partnership with Broadcom earlier this year under an agreement covering several generations of AI processors.

Building custom chips could help Meta lower its computing expenses and reduce its dependence on processors supplied by Nvidia and AMD.

However, Meta does not plan to replace those companies’ graphics processors entirely. Iris will work alongside the large number of Nvidia and AMD GPUs already used in its data centers.

The internal memo said adopting the latest GPUs across an operation as large as Meta had required substantial work and cost the company time.

Forrester analyst Mike Gualtieri said major AI companies increasingly need their own processors to compete on the cost of running AI models instead of relying completely on external chip suppliers.

Meta publicly revealed the technical version of Iris in March alongside three other planned AI processors.

The company intends to develop and deploy four generations of custom chips within two years, covering ranking, recommendation and generative AI workloads. It plans to introduce approximately one chip every six months through 2027, considerably faster than the annual or longer cycles commonly used by chip developers.

Meta plans to operate seven gigawatts of computing infrastructure in 2026.

The company added one gigawatt during the first half of the year and expects to add another 2.5 gigawatts by the end of 2026, according to the memo. One gigawatt is approximately enough electricity to power 800,000 homes.

Meta then plans to double its total computing capacity to 14 gigawatts in 2027.

The expansion forms part of Meta Compute, an initiative created to manage the company’s data centers, infrastructure and supplier relationships as it pursues more advanced AI systems.

Meta expects to spend as much as $145 billion on AI infrastructure in 2026.

That amount represents a significant portion of the more than $700 billion that major technology companies are collectively expected to spend on AI during the year.

Meta has also secured several long-term supply agreements to support its data-center expansion.

The agreements include Samsung Electronics for memory chips, Sandisk for flash storage and Sumitomo Electric for fiber-optic equipment. Sandisk declined to comment, while Samsung and Sumitomo Electric did not respond to Reuters’ requests for comment.

Long-term supply agreements have become increasingly important as technology companies compete for limited supplies of memory chips, AI processors and other data-center components.

Rapidly growing demand has pushed prices high enough for Morgan Stanley analysts to describe the wider effect as “chipflation,” warning that rising semiconductor costs are becoming a broader economic concern.

Meta shares initially fell after Reuters published the report. They later recovered after the company announced developer access to an AI coding model that competes with offerings from OpenAI and Anthropic, and were trading 4.6% higher during late-afternoon trading.

Iris represents Meta’s latest attempt to gain greater control over the hardware powering its AI products while reducing the cost and complexity of relying exclusively on Nvidia and AMD.

Get the latest tech news, telecom insights, and product launches wherever you prefer.

Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.