Meta to Begin Mass Production of In-House AI Chips in September, Reducing Reliance on Nvidia
Preparing for Global Semiconductor Shortages
$145 Billion Invested in AI Computing Infrastructure
Meta Platforms will begin mass production of its self-designed artificial intelligence (AI) chips starting in September 2026. By introducing new chips optimized for its own platforms, Meta aims to strengthen its computing infrastructure and respond to changes in the semiconductor supply chain.
Reuters reported on July 9 (local time) that it had obtained an internal memo stating Meta completed testing of its in-house AI chips in just six weeks and found no significant issues.
In March 2026, Meta unveiled four types of AI chips under the codename "Iris." These chips were custom-designed for the fourth-generation learning and inference accelerator (MITA) project. They are intended to enhance the AI performance powering social networking services such as Facebook and Instagram. With the accelerated pace of chip development, Meta plans to roll out new chips approximately every six months until 2027, whereas typical chip development takes more than a year. The chips are designed in collaboration with Broadcom and manufactured by Taiwan's TSMC.
Meta is expected to further accelerate chip development to hedge against rising prices caused by the global semiconductor shortage. CNBC analyzed, "As big tech companies race to expand their AI data centers, a surge in memory and chip prices—so-called 'chipflation'—has emerged as a risk to Meta’s profitability," adding, "By producing its own chips, Meta can also reduce its reliance on potential future competitors such as Nvidia and AMD."
Alongside in-house chip development, Meta is also dramatically expanding its investment in AI computing infrastructure. This year, Meta will build computing infrastructure with a capacity of approximately 7 gigawatts (GW), and next year will double that to 14 GW. Previously, Meta announced it would invest $145 billion (about 220 trillion won) solely in AI infrastructure. This amount accounts for 20% of the total projected capital expenditures of $700 billion by the entire big tech industry this year.
To secure its computing infrastructure supply chain, Meta has reportedly signed long-term supply agreements with Samsung Electronics for memory chips, with SanDisk for flash storage, and with Sumitomo Electric for optical fiber equipment.
Hot Picks Today
"Two Spend 500,000 Won in a Flash, Yet Wait Two Hours?" ... 'Dreamland' Shunned by Kids and Foreign Tourists
- The Era of Getting 200 Million Won After 30 Years Is Over... A Japanese Bank's "Perpetual Department Head" Who Just Clocked In Cries Out
- Spain Hit by Its Worst Tragedy... Lee Kangin, Moreno and Real Madrid Join Soccer World in Mourning
- They Said, “Make It Your Home for Life” - Then Suddenly Kicked 200 Passengers Off... What Happened to the Around-the-World Cruise?
- "130 Rice Bowls Sold Out in a Day"... 'Unfiltered Face' Adds 8 Million Won to Night Market Sales: China's Bungeunwoo
On this day, Meta also released its new AI coding model, "Muse Spark 1.1," as a paid API. Until now, Meta’s major generative AI models have centered around Llama, which is open source. After the release of Muse Spark 1.1, Meta’s previously declining stock price rebounded, closing at $631.50, up 4.7% from the previous day.
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.