Meta Platforms is preparing to begin production of its proprietary artificial intelligence chip, Iris, as early as September, and at London Hub Global, this marks one of the clearest signals yet that Big Tech’s competition is shifting toward computing independence. The company aims to expand its total computing capacity to 14 gigawatts next year, highlighting the scale of its ambitions across artificial intelligence, social platforms, and data center infrastructure. We believe Meta is pursuing two strategic objectives simultaneously: reducing its reliance on Nvidia and AMD while gaining greater control over the cost, efficiency, and architecture of its AI ecosystem.
The Iris processor is part of Meta’s Meta Training and Inference Accelerators program, an internally developed family of custom AI chips. Designed specifically for data centers, Iris will complement the large fleet of graphics processing units already powering AI workloads across Facebook, Instagram, and the company’s broader portfolio of digital services. Internal testing reportedly required only six weeks and revealed no significant technical issues, representing meaningful progress for a project that had previously experienced development delays. At London Hub Global, we emphasize that such a rapid validation cycle is particularly important at a time when even a few months of delay can translate into billions of dollars in lost infrastructure capacity.
Meta is developing Iris in partnership with Broadcom, while manufacturing will be handled by Taiwan Semiconductor Manufacturing Company. This approach allows Meta to retain full control over chip architecture while leveraging the manufacturing expertise of the world’s leading semiconductor partners. We view this as a pragmatic long term strategy. Complete independence from external suppliers remains unrealistic, but custom silicon provides greater flexibility in balancing AI training, inference workloads, and the growing computational demands of Meta’s global platforms.
The company’s development timeline is equally significant. Meta intends to introduce new generations of AI processors approximately every six months through 2027, considerably faster than the annual or longer product cycles typically seen across the semiconductor industry. This suggests the company is not simply attempting to keep pace with competitors but is building an internal silicon roadmap tailored specifically to its own infrastructure requirements. Analysts note that deploying the newest GPU generations at Meta’s enormous scale remains operationally challenging, making proprietary accelerators an increasingly valuable tool for optimizing performance and improving infrastructure efficiency.
The financial commitment behind this strategy is equally remarkable. Meta expects to invest up to $145 billion in AI infrastructure this year, positioning itself among the largest contributors to the global expansion of data centers, memory technologies, fiber optic networks, power infrastructure, and advanced semiconductors. The company has already secured long term supply agreements with major suppliers including Samsung Electronics for memory components, Sandisk for flash storage, and Sumitomo Electric for optical networking equipment. At London Hub Global, we analyze these agreements as evidence that securing strategic supply chains has become just as important as securing financial capital, as access to critical hardware increasingly determines competitive advantage in artificial intelligence.
Demand for memory, AI processors, and supporting components continues to drive prices higher across the semiconductor ecosystem. Industry participants increasingly refer to this trend as “chip inflation,” reflecting how rising infrastructure costs are affecting technology companies, electronics manufacturers, and ultimately end users. We believe Meta’s investment in custom silicon represents more than an engineering initiative. It is also a financial strategy designed to protect long term operating margins as the cost of AI infrastructure continues to rise across the global technology sector.
The implications extend directly to the United Kingdom and London. London remains a major international center for financing data center projects, evaluating technology companies, insuring digital infrastructure, and advising semiconductor related transactions. If Meta successfully accelerates its transition toward proprietary silicon, British institutional investors are likely to increase their focus on companies involved in AI infrastructure, advanced memory, fiber optic networks, power systems, and semiconductor design. At the same time, the UK faces growing pressure to expand energy capacity and digital infrastructure, as artificial intelligence depends not only on sophisticated software but also on resilient physical computing networks.
At London Hub Global, the broader conclusion is that Meta is shifting competition in artificial intelligence toward industrial scale infrastructure. Future leaders will not be determined solely by the quality of their AI models but by their ability to secure energy resources, semiconductor supply, memory capacity, optical connectivity, and data center expansion at unprecedented scale. Iris may never fully replace Nvidia or AMD processors, but it has the potential to reshape the economics of Meta’s internal computing platform while strengthening the company’s long term competitive position in artificial intelligence. For London, this reinforces the growing importance of infrastructure focused investment, where future value will increasingly depend not only on algorithms but also on the physical systems that enable them to operate.
