Nvidia (NVDA) Stock Drops as Meta Set to Make Its Own AI Chip in September

TLDR

• Nvidia stock was down 0.3% at $202.14 in premarket trading on Friday
• Meta plans to start manufacturing its in-house AI chip, codenamed “Iris,” from September
• Meta’s chip is designed to augment — not replace — GPU purchases from Nvidia and AMD
• Google and Amazon are also preparing to sell their own AI chips to external customers
• Analysts say Nvidia’s absolute dollar spending from hyperscalers will likely still grow, even as relative market share dips

---

Nvidia stock was down 0.3% at $202.14 in premarket trading on Friday, edging lower as news broke that Meta Platforms plans to begin manufacturing its own in-house AI chip as early as September.

NVIDIA Corporation, NVDA

Meta’s chip, codenamed “Iris,” is part of a four-generation project under its Meta Training and Inference Accelerator (MTIA) program. The chip is being designed with help from Broadcom and will be manufactured by Taiwan Semiconductor Manufacturing Co (TSMC).

The internal memo reviewed by Reuters showed that bug testing took just six weeks and turned up no major issues — a faster-than-expected timeline for a program that has had a rocky start since launching more than five years ago.

Meta plans to release a new chip roughly every six months through 2027, a notably faster cadence than the industry norm of once a year or more.

The goal is straightforward: lower computing costs and reduce reliance on external chip suppliers like Nvidia and AMD.

That said, “Iris” is not designed to replace Nvidia GPUs entirely. The chip is aimed at augmenting the large volumes of GPUs Meta already buys. Meta’s own internal memo acknowledged that adopting the latest GPUs at its scale “has been a heavy lift, and it has cost us time.”

What Analysts Are Saying

Benchmark Research analyst Cody Acree pushed back on the idea that this spells trouble for Nvidia. He noted that while Nvidia may lose relative market share within Meta’s growing build plans, overall hyperscaler spending is expected to more than double, meaning Nvidia’s absolute revenue from these customers could still grow.

Meta expects to spend up to $145 billion on AI infrastructure this year alone, as part of a broader Big Tech push projected to exceed $700 billion.

The Bigger Picture for Nvidia

The Meta news is one piece of a larger story. Google and Amazon are both preparing to offer their own AI chips — Google’s Tensor Processing Units and Amazon’s Trainium processors — to external customers.

Both are capable of running the latest AI models and may offer more cost-efficient performance in certain workloads, putting additional pressure on Nvidia’s dominance.

Nvidia’s stock has lagged the broader chip sector recently, with AMD up 5.67% and Meta up 4.70% on the day, while Nvidia sat in the red.

Futures tied to the Nasdaq 100 were down 0.2%, meaning Nvidia’s move was broadly in line with the wider market, though the company-specific headlines added an extra layer of pressure.

Meta plans to deploy seven gigawatts of computing infrastructure in 2026 and double that figure in 2027, with long-term supply agreements already locked in with Samsung, Sandisk, and Sumitomo Electric to support that expansion.