New York, February 2, 2026, 17:39 EST — After-hours
- Nvidia shares slipped 2.9%, closing at $185.61 in Monday’s late trading session
- A report indicates OpenAI is investigating alternatives for certain AI “inference” workloads
- Attention turns to OpenAI funding discussions and Nvidia’s earnings report on Feb. 25
Nvidia (NVDA.O) shares dropped 2.9% to $185.61 in after-hours trading Monday, as investors weighed new concerns about demand from a major AI computing customer. The stock fluctuated between $183.82 and $190.28, with roughly 164 million shares traded.
This matters because the next battleground is “inference” — when a trained AI model answers a user’s prompt — and speed per query is where customers push hardest on costs. Any hint that major buyers are testing substitutes can shake a trade built on tight supply and long order books.
That sensitivity cuts right through OpenAI. Over the weekend, Jensen Huang announced plans for a “huge” investment in OpenAI, dismissing talk of any friction as “nonsense.” He clarified the deal wouldn’t approach $100 billion, countering a Wall Street Journal report that the earlier plan had stalled. Huang also said he “really love(s) working with” Sam Altman. (Reuters)
Eight sources familiar with the situation say OpenAI is searching for alternatives to some of Nvidia’s newest chips for specific inference tasks like coding, citing concerns over response speeds. The company is reportedly weighing options from Advanced Micro Devices and startups Cerebras and Groq, focusing on designs with more on-chip SRAM to reduce data retrieval times. Nvidia maintains that customers choose its GPUs for inference because they offer the best performance and total cost of ownership at scale. An OpenAI spokesperson added that Nvidia still powers the vast majority of its inference fleet. (Reuters)
Supply constraints remain a concern. In Taipei, Huang quipped that Taiwan Semiconductor Manufacturing Co “needs to work very hard” since he requires “a lot of wafers.” He also flagged a looming production squeeze in memory chips critical for AI workloads. (Reuters)
The broader market ignored Nvidia’s jitters. U.S. stocks climbed Monday, with chipmakers generally pushing higher. AI-related chip components helped boost both the S&P 500 and Nasdaq. (Reuters)
Investors see this as more than a dispute over a single client. Should OpenAI—and eventually others—move a significant portion of inference tasks to different chips, it could undercut pricing power, even if demand for training remains strong.
Switching isn’t simple either. Big deployments take their time, and OpenAI itself pointed out that Nvidia still handles the bulk of inference capacity these days.
Nvidia is set to announce its fourth-quarter and full-year fiscal 2026 earnings on February 25, followed by a conference call at 2 p.m. Pacific, according to the company’s investor calendar. (NVIDIA Investor Relations)
Traders are waiting to see if “testing alternatives” turns into real orders and if more clients jump on board beyond just one marquee name. Tuesday’s session will offer the first clear glimpse of how long this concern might last.
