Meta is preparing for one of its biggest artificial intelligence infrastructure upgrades yet. According to an internal company memo reviewed by Reuters, the company plans to begin manufacturing its next-generation AI chip, codenamed Iris, in September 2026.
The chip is part of Meta's long-term Meta Training and Inference Accelerator (MTIA) programme, under which the company is developing its own AI processors instead of relying entirely on third-party suppliers. The goal is to improve the AI systems powering Facebook, Instagram and other Meta products while reducing computing costs.
The reported production follows six weeks of successful testing, during which engineers found no major issues with the chip.
Why is Meta building its own AI chip?
Running advanced AI models requires enormous computing power, and buying large numbers of graphics processing units (GPUs) from companies such as Nvidia and AMD has become increasingly expensive.
Meta believes custom-designed chips can perform many AI tasks more efficiently while giving the company greater control over its infrastructure.
Trending Stories
The Iris processor has been designed with help from Broadcom, while Taiwan Semiconductor Manufacturing Co. (TSMC) will manufacture the chip. Industry analysts say nearly every major AI company is now pursuing custom silicon to reduce costs and avoid dependence on external chip suppliers.
Meta plans to double computing capacity
The Reuters memo shows Meta expects to deploy around 7 gigawatts (GW) of computing infrastructure by the end of 2026 before doubling that figure to 14GW in 2027.
For comparison, 1GW of electricity can power roughly 800,000 homes. Such a rapid expansion highlights the enormous energy requirements of modern AI systems. Meta also expects to spend up to $145 billion this year on AI infrastructure, making it one of the biggest investors in artificial intelligence globally.
The AI hardware race is accelerating
Meta's chip strategy reflects a broader industry trend. Google has its Tensor Processing Units (TPUs), Amazon has Trainium and Inferentia, while Microsoft, OpenAI and Anthropic are also exploring custom AI hardware.
According to the memo, Meta plans to introduce a new AI chip roughly every six months through 2027, a much faster release cycle than the industry's traditional annual updates. The company has also secured long-term supply agreements with Samsung Electronics for memory chips, Sandisk for storage and Sumitomo Electric for fibre-optic equipment as it expands its global data centre network. With AI demand continuing to grow, owning more of its hardware stack could become one of Meta's biggest competitive advantages in the years ahead.

&imwidth=800&imheight=600&format=webp&quality=medium)
&im=FitAndFill=(700,400))
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
)
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)