Nvidia enters correction territory as stock falls 10% from all-time highs

Nvidia entered “correction territory,” after shares briefly fell 10% from their most recent all-time closing high.

Shares had recovered by Wednesday’s close when they were only about 8% off the high.

The company, which makes graphics processing units — or GPUs — has been a key beneficiary of the artificial intelligence boom, which boosted demand for its chips.

Nvidia GPUs are commonly used for compute-intensive AI applications, such as OpenAI’s ChatGPT AI chatbot. Its server chips are also a key component of data centers.

The company’s financial performance has been torn in the past year. It reported a 486% jump in non-GAAP earnings per diluted share in the December quarter, citing huge chip demand, thanks to the popularity of generative AI models.

The stock has come under pressure for the past two weeks, however. On Tuesday morning, shares were 10% from their last all-time closing high of $950 apiece, which they hit on March 25. The stock closed at $853.54 on Tuesday, down 2% for the session.

Nvidia’s shares closed up 1.97% on Wednesday.

Definitions of what constitutes a market correction vary, but it is generally considered to be a sustained drop of 10% or more from all-time highs.

Nvidia declined to comment on this story.

What’s the reason for the decline?

The exact reason for the downward move hasn’t been immediately clear. Investors could be taking profit on the stock, after a wild gain of more than 200% for the shares in the last 12 months. And on Tuesday, rival chipmaker Intel unveiled a new AI chip called Gaudi 3, aimed at powering large language models — the cornerstone technology behind generative AI tools like OpenAI’s ChatGPT.

Intel said the new chip is over twice as power-efficient as Nvidia’s H100 GPU — the U.S. chip giant’s most advanced graphics card — and can run AI models 1½ times faster than Nvidia’s GPU.

Analysts at D.A. Davidson said in a research note that they expect a “shrinking” of the size of AI models, including alternatives like Mistral’s Large model and Meta’s LLaMA system, to drive down demand for Nvidia’s stock over time.

“Although NVDA (Neutral-rated) should deliver a spectacular 2024 (and perhaps into 2025), we continue to believe recent trends set up a significant cyclical downturn by 2026,” D.A. Davidson analysts said in the note Tuesday.

“A combination of shrinking models, more steady growth in demand, maturing hyper scaler investments, and increased reliance by the largest customers on their chips do not bode well for NVDA’s out years.”