nvidia (NVDA) 1.13%)) The stock closed a new $150 shy record in January, but has since dropped 16%. The drop was sparked by news that a Chinese artificial intelligence (AI) startup called Deepseek has found a way to train advanced models with less costly costs than their American peers, and less computing power.
As Nvidia offers the world’s most powerful data center graphics processing units (GPUs) for AI development, investors were concerned that Deepseek’s methodology would drive a collapse in demand for those chips. However, Nvidia CEO Jensen Huang made a series of comments last week suggesting that some of the latest AI models from DeepSeek and other developers could actually lead to a significant increase in GPU demand.
Here’s what he said: And why am I predicting comments and predicting that Nvidia stock could skyrocket this year.
Image source: nvidia.
Nvidia is away from the best financial year ever
Before diving into Huang’s comments, let’s review Nvidia’s financial results for 2025 (ends January 26th) released last Wednesday. The company had recently launched its new Blackwell GB200 GPU, which has become the gold standard for AI development, so it was anticipating a great result.
Nvidia generated a record $1300.5 billion in total revenue for fiscal year 2025. This was a 114% increase from the previous year, comfortably surpassing management forecasts by $128.6 billion. The data center segment totaled $115.1 billion. This is an increase of 142% from the previous year.
Nvidia began shipping commercial volumes of Blackwell GPUs for the first time in the fourth quarter of 2025. They generated $11 billion in sales, surpassing management expectations and became the fastest product ramp-up in the company’s history.
The Blackwell GB200 GPU is a game changer. This is because in some configurations it can perform AI inference at 30 times faster than Nvidia’s previous flagship data center chip, H100. Inference is the process by which an AI model uses live data and prompts to make predictions or shapes. Higher inferences can speed up responses in chatbot applications, but models can also process more data and provide better answers.
Nvidia says that Blackwell production needs to continue scaling due to high demand. Some of the company’s top customers have already revealed how much they plan to spend on AI data centers and chips this year.
Not all of that money will specifically flow to Nvidia, but it is clear that Deepseek’s innovations aren’t reducing their appetite for more computing power.
Jensen Huang has provided incredible news to investors
Openai is one of America’s top AI startups. Since its founding in 2015, it has spent over $20 billion building data center infrastructure, building talented teams and training AI models. So it was a huge surprise when Deepseek revealed he had trained the V3 model. This works on par with Openai’s GPT-4O model on several benchmarks – runs for just $5.6 million.
This does not include $500 million in infrastructure spending (according to Semianalysis estimates), but it has alerted up and down Wall Street.
Deepseek used a set of clever techniques on the software side to offset the lack of computing power, as it is unable to access Nvidia’s latest chips due to its ban on exports to China. One of them is called distillation. This involves training small models that accelerate progress quickly using the best proven models (such as the GPT-4O). This reduces traditional training workloads, which results in much less resources.
If all AI developers rely on distillation, chip demand for training workloads will likely collapse. However, startups like Openai know that supplying an infinite amount of data to AI models no longer produces the desired outcomes, so they are now focusing on building “inference” models, spending more of their “thinking” to create the best response.
This means they do not rely on the entire traditional training method, shifting their computing resources to inference workloads instead. It brings me the latest comments from Huang. During a conference call with investors last Wednesday, he said the inference model could consume 100 times more calculations than its predecessor. In the future, he believes that some models can consume thousands, or even millions of times more calculations, producing extremely complex simulations and other outputs.
Deepseek R1, Xai’s Grok 3, Anthopic’s Claude 3.7 Sonnet, and Openai’s GPT4-O1 to GPT-4O3 to GPT-4O3 are examples of inference models available today. Most of the top developers have clearly moved away from their pre-training methods and instead headed in this direction. This means that Nvidia could be in a cusp in a new demand phase for chips.
Nvidia’s inventory could be a bargain
Based on NVIDIA’s earnings per share (EPS) of $2.99 for fiscal year 2025, the shares trade at a price of 42.5 (P/E) ratio, with a 28% discount of an average of 59.3 for 10 years. Additionally, Wall Street consensus estimates (provided by Yahoo!) suggest that the company will be able to generate $4.49 in EPS in 2026, with a forward P/E ratio of just 27.7.
NVDA PE ratio data by YCHARTS.
In other words, Nvidia shares should rise 53% over the next 12 months to maintain their current P/E ratio.
Deepseek Saga has been a key reason for the recent stock decline, so recent comments on a potential 100-fold increase in computational requirements for inference workloads could be a catalyst for folding investors, particularly if the company’s financial results support its forecasts this year.
As a result, I think Nvidia’s inventory could soar for the next 12 months (and possibly beyond that).
John Mackey, former CEO of Amazon subsidiary Whole Foods Market, is a member of Motley Fool’s board of directors. Randi Zuckerberg, a former director of market development, Facebook spokeswoman and sister to Metaplatform CEO Mark Zuckerberg, is a member of Motley Fool’s board of directors. Suzanne Frey, an executive at Alphabet, is a member of the board of directors of Motley Fool. Anthony di Pizio does not occupy any of the stocks mentioned. Motley Fool has positions for Alphabet, Amazon, Meta Platforms, Microsoft, and Nvidia, and is recommended. Motley Fool recommends the following options: A $395 phone at Microsoft for January 2026 length and a $405 phone to Microsoft for January 2026 short term. Motley Fools have a disclosure policy.
