Nvidia shares remain mired in negative territory for the year, and are down more than $500 billion from their early-January peak, but investors are starting to rework near-term sales forecasts for the AI-chip giant after its biggest customers announced big spending plans.
Nvidia (NVDA) maintains a commanding grip on the market for AI-powering chips, including its newest line of Blackwell processors, which boost computing efficiency while using less energy and has seen enormous revenue gains since the mainstream adoption of the new technology in early 2023.
💵💰Don’t miss the move: Subscribe to TheStreet’s free daily newsletter 💰💵
The group’s fiscal 2024 sales more than doubled, to just under $70 billion, and analysts are forecasting a tally of around $129 billion for its fiscal year 2025, which ended in January.
Expectations for another 50% growth rate for the coming fiscal year were challenged late last month, however, by the emergence of DeepSeek, the China-based AI startup that claims to have built, trained and developed a large-langue-model system and chatbot for a fraction of the cost borne by its U.S. rivals.
Nvidia CEO Jensen Huang has called the group’s new Blackwell GPUs “the engine of AI”.
PATRICK T. FALLON/Getty Images
News of the DeepSeek AI agent, and reports suggesting it was trained for less than $6 million, wiped nearly $600 billion from Nvidia’s share price in a single day, the largest session market-capitalization decline on record.
DeepSeek raises AI capital-spending questions
But skepticism has arisen about DeepSeek’s claims, and reports say that it used high-end Nvidia chips that were barred from export to China under rules put in place by the Biden administration, which have tempered the market’s concern about Nvidia’s near-term forecast, which the group tacitly affirmed in a statement last month.
Related: Analysts rework Alphabet stock price targets after earnings shock
“DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling,” the company said. “DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely available models and compute that is fully export control compliant. Inference requires significant numbers of Nvidia GPUs and high-performance networking.”
That was certainly echoed last week by a host of Nvidia’s biggest customers, the so-called hyperscalers, the providers of cloud and data center infrastructure at a massive global scale, through their combined capital spending plans.
The major providers of cloud services and infrastructure plan major boosts in capital spending.
Microsoft (MSFT) , Meta Platforms (META) , Google parent Alphabet (GOOGL) and Amazon (AMZN) are set to spend a collective total of $325 billion this year alone. That’s more than the combined market value of Citigroup and Pfizer.
Their three-year capex run rate, starting in 2023, is slated to increase nearly three-fold to around $690 billion, more than the market value of Visa and just shy of JP Morgan Chase.
“Chips are a key ingredient in the compute that drives training and inference,” Amazon CEO Andy Jassy told investors Thursday. “Most AI compute has been driven by Nvidia chips, and we obviously have a deep partnership with Nvidia and will for as long as we can see into the future.”
Nvidia sees ‘staggering’ Blackwell demand
Not all of that will be spent on Nvidia products, of course, and each of the four biggest hyperscalers has spoken about reducing its reliance on a single supplier by ramping investments in chips and processors that produced in-house.
But even with that subtle shift in focus, the demand for Nvidia GPUs, which finance chief Colette Kress described as “staggering” in a call with investors in late November, isn’t likely to stumble.
“We currently do not believe a major capex trajectory change is afoot as we tend to lean skeptical on the DeepSeek
debate and believe compute expands with lower pricing should be offset with higher volumes,” said Raymond James analyst Josh Beck.
Related: Analysts rework Amazon stock price targets after Q4 earnings surprise
Cantor Fitzgerald analyst C.J. Muse had a similar view, arguing in a late January note that DeepSeek’s advances puts the market closer to the entry of artificial general intelligence, or AGI, a form of AI with human-like cognitive abilities.
“Work will continue on pre-training, post-training, and time-based inference/reasoning, and future investments in large-scale clusters will only accelerate,” he said.
“All of this is bullish for AI [and] we see this progress as positive in the need for more and more compute over time, not less.”
Nvidia ‘creating its own market’
Nvidia, which reports its fourth quarter results after the close of trading on Feb. 26, will be certain to address both DeepSeek’s impact on the broader AI landscape, its expectations for a share of the $325 billion committed by the biggest hyperscalers and the growing growing presence of rivals like Broadcom (AVGO) and Marvell (MRVL) and Advanced Micro Devices (AMD) .
“Nvidia is still our top holding regardless of all that has transpired the last few weeks, even with the DeepSeek news and restrictions on exports to China,” said Ken Mahoney, CEO of Mahoney Asset Management.
More AI Stocks:
Analysts overhaul Apple stock price targets after record Q1 earningsVeteran fund manager reveals startling AI stocks forecast for 2025These agentic AI stocks could soar in 2025
“If [capex estimates] deaccelerate, we will consider changing the stance, but for now this is still a tailwind for Nvidia and the tech sector in general,” he added. “Nvidia has never been looking to get in on some market share; they are creating markets themselves.”
Nvidia shares closed at $129.84 each Friday, after rising 0.9% on the session to trim their year-to-date decline to around 6.1%.
Related: Veteran fund manager issues dire S&P 500 warning for 2025
