BY COMFORT OGBONNA
Meta Platforms is preparing to begin manufacturing a new artificial intelligence chip in September as the company pushes forward with an ambitious plan to dramatically expand its computing infrastructure and reduce its dependence on outside chip suppliers. The effort represents a major step in Meta’s long-term strategy to strengthen its AI capabilities while lowering the enormous costs associated with developing and running advanced AI systems.
According to an internal company memo, Meta’s newest AI chip, code-named “Iris,” is part of the company’s Meta Training and Inference Accelerators (MTIA) program, a multi-generation initiative focused on designing high-performance AI processors in-house. The project includes four generations of custom-built chips that will power the artificial intelligence technologies behind Facebook, Instagram, and other Meta products.
The company plans to begin production of the Iris chip in September after completing an extensive testing process that lasted only six weeks. The tests reportedly uncovered no major technical issues, an encouraging result for Meta after years of investing in custom silicon development. The relatively smooth testing phase signals significant progress for an initiative that has faced numerous challenges since it began more than five years ago.
Meta has partnered with Broadcom to assist with the chip’s design, while Taiwan Semiconductor Manufacturing Company (TSMC), the world’s largest contract chipmaker, will manufacture the processor. By developing its own AI hardware, Meta aims to optimize performance for its specific software workloads while reducing its reliance on external suppliers such as Nvidia and Advanced Micro Devices (AMD), whose graphics processing units currently power much of the company’s AI infrastructure.
Although the new processor is expected to play an increasingly important role, Meta does not plan to replace GPUs entirely. Instead, the Iris chip will work alongside the vast number of Nvidia and AMD graphics processors already deployed across the company’s data centers. Executives believe the custom chip will improve efficiency while giving the company greater control over its rapidly expanding AI ecosystem.
The internal memo also noted that integrating the latest commercially available GPUs across Meta’s enormous infrastructure has proven both time-consuming and expensive. Developing proprietary chips tailored specifically to the company’s needs is expected to simplify deployment while delivering better performance for AI training and inference tasks.
Industry analysts view the move as part of a broader trend among the world’s largest technology companies. As artificial intelligence becomes increasingly central to future growth, major cloud providers and technology firms are investing heavily in custom-designed chips to reduce costs, improve efficiency, and gain greater independence from third-party manufacturers.
Meta first introduced the Iris processor under its technical MTIA branding earlier this year alongside three additional AI processors. The company now plans to release a new AI chip approximately every six months through 2027, a significantly faster development cycle than the annual release schedules followed by many semiconductor companies.
Alongside its chip development efforts, Meta is dramatically increasing its computing capacity. The company expects to operate approximately seven gigawatts of computing infrastructure this year. It has already added one gigawatt during the first half of the year and plans to bring another 5.5 gigawatts online before year-end. Looking ahead, Meta intends to double its computing capacity again, targeting a total of 14 gigawatts by 2027.
To put that scale into perspective, a single gigawatt of electricity can supply power to roughly 800,000 homes. Reaching 14 gigawatts would make Meta one of the world’s largest operators of AI-focused computing infrastructure, underscoring the immense energy requirements needed to support next-generation artificial intelligence.
The company’s aggressive expansion comes with an equally ambitious investment strategy. Meta expects to spend as much as $145 billion on AI infrastructure this year alone, accounting for a substantial share of the technology industry’s projected spending on artificial intelligence. Those investments include new data centers, networking equipment, advanced processors, storage systems, and other critical infrastructure needed to support increasingly sophisticated AI models.
To secure the components required for its expansion, Meta has signed long-term supply agreements with several major technology manufacturers. These partnerships include Samsung Electronics for memory chips, Sandisk for flash storage, and Sumitomo Electric for fiber-optic equipment used in large-scale data centers. Long-term contracts have become increasingly important as global demand for AI hardware continues to outpace supply.
The surge in demand for memory chips, graphics processors, and other AI-related components has created supply constraints across the technology industry, prompting companies to secure multi-year agreements to guarantee access to critical hardware. As businesses continue investing billions of dollars in artificial intelligence, competition for computing resources is expected to remain intense.
Meta’s latest investments highlight the company’s determination to become a leader in artificial intelligence by building not only powerful AI models but also the underlying hardware needed to support them. With custom-designed chips, expanding data centers, and record-breaking infrastructure spending, the company is positioning itself for the next phase of the global AI race.
Original article
Comments