Nvidia Stock Is Going to Surge After Feb. 26


Since the beginning of 2023, Nvidia (NASDAQ: NVDA) has added a whopping $2.8 trillion to its market capitalization on the back of soaring demand for its data center chips, which are the gold standard for developing artificial intelligence (AI). At the same time, Nvidia stock is trading down 11% from its record high set in early January 2025 following a sharp sell-off over the past month.

The sell-off was sparked by news that China-based research lab, DeepSeek, has found a way to train competitive AI models with a fraction of the computing power (and financial resources) of its American peers. Investors feared this would trigger a collapse in the demand for data center chips, thus crushing Nvidia’s core business.

On Feb. 26, Nvidia will report its latest financial results for its fiscal 2025 fourth quarter (ended Jan. 31), and I think the information it contains will squash some of the recent investor concerns. Here’s how I predict the stock will react once the results hit the wires.

Nvidia's headquarters with a black Nvidia sign in the foreground.
Image source: Nvidia.

Nvidia’s H100 graphics processor (GPU) was the hottest AI data center chip in the world during 2023, helping the company capture an incredible 98% market share. It remains a top seller but it was superseded by the H200, and then an entirely new generation of GPUs based on Nvidia’s Blackwell architecture.

The Blackwell-based GB200 NVL72 system can perform AI inference at 30 times the pace of the equivalent H100 system, paving the way for developers to deploy the most advanced AI models to date. Nvidia CEO Jensen Huang told investors demand was “insane” shortly after Blackwell’s broad release at the end of 2024, and from what we know so far, sales are living up to expectations.

However, the DeepSeek saga rocked Wall Street’s confidence in January. The Chinese start-up revealed it spent just $5.6 million to train its V3 AI model, yet it matches the performance of some of the best models from American start-ups like OpenAI, which have invested tens of billions of dollars to reach this point. Moreover, DeepSeek used older generations of Nvidia’s GPUs, which left investors wondering whether AI developers really need the latest and greatest Blackwell chips.

But some of those concerns have since been put to bed by Nvidia’s largest customers. Meta Platforms CEO Mark Zuckerberg thinks a drop in training workloads will be offset by inference workloads, which are now consuming an increasing amount of computing power because newer AI models spend more time “thinking” (which is known as test-time scaling). Meta expects to spend up to $65 billion on AI data center infrastructure during 2025, up from $39.2 billion last year, so it certainly isn’t pulling back.


Leave a Comment