Meta is finally ready to make its own AI chip at scale. According to an internal memo reviewed by Reuters, the company will start mass production of its new AI chip in September this year. The chip, known internally as Iris, is part of Meta's in house chip program called MTIA(Meta Training and Inference Accelerator.)
This is a big moment for Meta. The company has been trying to build its own AI chips since 2023, but the effort has struggled for years. Iris is the first chip from this program to move so smoothly through testing. It cleared bug testing in just six weeks without any major issues. For a chip built to run inside massive data centers, that speed is rare.
Why Meta Needs Its Own AI Chip?
Right now, Meta buys most of its AI chips from Nvidia and AMD. These GPUs are powerful but expensive and hard to get in large numbers. The memo says fitting the newest Nvidia GPUs into Meta's giant data centers has been a heavy lift and has cost the company time.
Meta wants to fix this by building chips suited to its own needs. Iris will help run AI systems behind Facebook, Instagram and WhatsApp. It will not fully replace Nvidia or AMD chips. Meta will still buy GPUs from both companies. But Iris will reduce how much Meta depends on outside suppliers and could lower its massive AI spending over time.
Also Read Meta’s Paid Social Era Begins: Instagram, WhatsApp and Facebook Get Plus Plans
The Numbers Behind Meta AI Chip Mass Production
Meta's AI ambitions come with a huge price tag. The company expects to spend between 125 billion and 145 billion dollars this year on AI infrastructure alone. Its computing capacity is set to grow from 7 gigawatts in 2026 to 14 gigawatts by 2027, based on the internal memo.
Meta first showed Iris and three other AI chips to the public in March. It now plans to release a new chip roughly every six months through 2027. Most chip companies take a year or longer between major releases. This faster pace shows how urgent the AI race has become for Meta.
The company has also lined up its supply chain early. It is sourcing memory from Samsung, storage from Sandisk and optical networking gear from Sumitomo Electric, according to Reuters. Broadcom is helping design the chip, and TSMC will manufacture it in Taiwan.
How Other Tech Giants Fared With Their Own Chips
Meta is not the first to try this. Google and Amazon started years earlier, and their results show why Meta is willing to wait. Google began building TPUs in 2015 and now runs almost all its AI on them. Its latest TPU cuts AI job costs by over 40 %.
Anthropic trains heavily on Google TPUs, and its revenue jumped from 9 billion to over 30 billion dollars in months. Amazon built Trainium the same way. It now runs over one million Trainium chips, saving customers 30 to 50 % over Nvidia GPUs. Both took close to a decade to scale. Meta started only in 2023, so Iris is still catching up.
Also Read Meta Overtakes Google in Digital Ads 2026: What Changed
What Mainstream Media Missed?
Most reports have focused only on Meta's new AI chip. But the bigger story is how the company's AI spending is affecting the global technology industry. Just a week before this announcement, reports said Meta was planning a cloud business that could rent out its extra AI computing power. The news pushed down the shares of Nvidia, AMD, Intel and Micron. It also affected Samsung and SK Hynix in South Korea.
Another important point is India's growing role in Meta's AI plans. Meta has already partnered with Reliance Industries to lease India's first dedicated AI data centre. As the company builds more of its own AI chips and expands its computing capacity, India is becoming an important part of Meta's global AI infrastructure, not just a market for its apps.
The announcement has also raised a bigger question for investors. JPMorgan analysts have said the huge profits of AI chip makers and the massive spending by companies like Meta may not remain balanced for long. This means the discussion is no longer only about a new chip. It is also about whether the AI industry's spending can continue at the current pace.
Also Read Anyone Can Use Your Public Instagram Photos With Meta's New AI—Here's How to Stop It
Why It Matters for Bharat
India is a major market for Meta, with hundreds of millions of Facebook, Instagram and WhatsApp users. Faster and cheaper AI chips can mean quicker improvements to features Indian users rely on daily, from content recommendations to safety tools.
For India's own semiconductor ambitions, Meta's move is a signal. Big tech firms are moving away from buying every chip from outside vendors and building their own instead. India's push under its semiconductor mission could learn from this shift, especially in chip design and testing partnerships.
For the government, Meta's data center plans with Reliance also mean more foreign investment and jobs in India's data center sector. For citizens, it means the apps they use every day are being built on cheaper and more efficient hardware, which could translate into better performance over time.
Also Read Can India Rival Taiwan in Semiconductors? Bharat's ₹1.64 Lakh Crore Chip Bet Explained
News4Bharat Verdict
Meta's new AI chip is more than a hardware update. It shows that big tech companies are done depending fully on Nvidia and AMD, and are ready to build their own path instead. The six week testing success proves Meta has finally fixed the problems that slowed its chip program for years. This move will not replace Nvidia overnight, but it marks the start of a shift that every AI company will now watch closely. For India, this story goes beyond Meta.
It connects directly to the Reliance data center deal and shows how India is becoming a real part of the global AI supply chain, not just a market for apps and users. The coming months will decide if Iris lives up to its promise, but one thing is already clear. The AI chip race has entered a new and more competitive phase.
