Meta is reportedly preparing to take a major step toward controlling its own AI infrastructure by beginning production of a new custom AI chip, code named “Iris,” in September 2026. The initiative is designed to reduce the company’s long-term dependence on Nvidia’s GPUs while supporting Meta’s rapidly expanding AI ambitions.
What is Meta building?
According to an internal memo reported by Reuters, Meta’s new chip is part of its Meta Training and Inference Accelerator (MTIA) program.
The chip is being:
– Designed by Meta engineers.
– Manufactured by TSMC.
– Developed in partnership with Broadcom.
– Built to handle both AI training and inference workloads across Meta’s products, including Facebook, Instagram, WhatsApp, and Meta AI.
Why Meta Wants Its Own AI Chip:
Meta currently spends tens of billions of dollars purchasing AI hardware from Nvidia and, to a lesser extent, AMD.
Building its own silicon could help Meta:
– Reduce its dependence on Nvidia.
– Avoid supply shortages that have affected AI chip availability over the past several years.
– Lower long-term AI infrastructure costs.
– Gain greater control over hardware optimization.
– It’s not replacing Nvidia overnight
– Improve performance for Meta-specific AI models.
Despite the headlines, Meta is not abandoning Nvidia. Reports indicate the new Iris chips are expected to complement, rather than completely replace, Nvidia GPUs. Meta will continue buying large numbers of Nvidia accelerators while gradually expanding the role of its own chips as they mature.
Massive AI infrastructure expansion: The chip project is part of an even bigger investment.
Meta reportedly plans to:
– Increase computing capacity from 7 gigawatts in 2026 to 14 gigawatts by 2027.
– Spend between $125 billion and $145 billion on AI infrastructure this year.
– Introduce a new generation of AI chips roughly every six months through 2027.
Meta isn’t alone. Major AI companies are increasingly designing their own processors to reduce reliance on Nvidia, including:
* Microsoft
* Google (TPUs)
* OpenAI (reported custom chip efforts)
* Amazon (Trainium and Inferentia)
* Apple
* DeepSeek in China
If Meta’s Iris chips perform well, they could significantly reduce the company’s AI operating costs over time and strengthen its control over the entire AI technology stack from hardware to models and applications. However, Nvidia is expected to remain a critical supplier for the foreseeable future, particularly for cutting-edge AI training workloads.
For Summaries Or Deep-Dives On Any Tech Trends, Contact Us At DawentsIT Or Visit Www.Dawentsit.Com. Subscribe To DawentsIT For More Tech News And Videos. Stay Inspired, And We Deliver! Read more on http://www.dawentsit.com/blog-2/ #Technology #DawentsIT #TechnologyNews #Tech #Meta #Iris #AIChip #AIInfrastructure #GPU #AllThingsTechnologyNews #AllFactorsTechnology #WeInspireToDeliver
