AI demand remains strong but can't boost the stock price! Nvidia has risen only 1% so far in the fourth quarter, as market sentiment turns cautious.
Despite the continued expansion of capital expenditure in the field of artificial intelligence, Nvidia’s stock performance has cooled off. This AI chip giant has risen only about 1% since the fourth quarter, with its current price-earnings ratio around 24 times, roughly equal to the Nasdaq 100 Index, indicating the market is reassessing its valuation premium.
Changes in the competitive landscape have become a core driver of wait-and-see sentiment. This month, Nvidia CEO Jensen Huang acquired inference hardware startup Groq’s technology license for about $20 billion and recruited most of its chip team—a move that itself attests to the competitiveness of other companies in certain areas. At the same time, Cerebras signed a $10 billion rapid inference chip supply agreement with OpenAI, and Anthropic has also reached cooperation deals with multiple non-Nvidia chip suppliers.
These transactions are reshaping the market’s perception of the AI chip landscape. Several startups report seeing a marked increase in potential investor interest since the Groq deal. SambaNova has even abandoned discussions to sell the company at much lower than its previous valuation, turning toward a new funding round instead.
For investors, this series of signals means: although Nvidia remains the undisputed leader in AI chips, its monopoly may no longer be as unbreakable as before. The market is shifting from “betting on the single leader” to “repricing competitive risk.”
Inference chip market becomes the focal point of competition
In the arena of AI chips, more and more startups and investors are focusing on “inference”—the key stage of running models and generating answers after training is complete. This niche field is being viewed as a breakthrough point to challenge Nvidia’s dominance.
Earlier this month, trading company Jump co-led a $230 million funding round for inference chip startup Positron and has become its client. The company’s CTO Alex Davies stated:
“Almost everyone uses Nvidia for both training and inference, but we’re seeing changes in the industry, and this won’t last forever. We don’t believe there will be only one winner.”
Leveraging high-bandwidth memory chips, Nvidia dominates in large-scale parallel training computation. However, a group of startups is attempting to achieve faster response speeds in inference scenarios by exploring different types of memory architectures. Meanwhile, as inference AI models conduct real-time judgments upon queries—rather than solely relying on pre-trained results—the boundary between training and inference is blurring, creating opportunities for new chip architectures.
Sid Sheth, CEO of AI chip company D-Matrix (backed by Microsoft), points out that since DeepSeek debuted early last year, there’s been a significant increase in market interest for fast inference chips. The company completed a $275 million funding round in November last year.
Tech giants accelerate development of proprietary AI chips
Large tech companies are racing to develop their own AI chips to reduce their reliance on Nvidia. OpenAI released its first model running on Cerebras chips on Thursday; Anthropic has signed usage agreements with Amazon Trainium and Google TPU; last month Microsoft rolled out the second generation of its own AI chip Maia and received rights to use OpenAI’s chip IP.
Startups are also actively deploying. Inference chip company Etched raised about $500 million last month, targeting Nvidia’s dominance; AI model startup Simile came out of stealth, raising $100 million led by Index Ventures, and aims to help enterprises predict human behavior.
Nevertheless, even as giants accelerate their in-house development, Amazon, Google, Microsoft, OpenAI, and others are still purchasing large volumes of Nvidia GPUs to support their AI products and cloud services. This reality highlights that Nvidia’s position as market leader remains solid, even as the competitive landscape quietly shifts.
Nvidia’s defense and market prospects
Nvidia has proven to be a powerful market leader. The company boasts multiple product lines and commits to a complete redesign of its chips every year. The deal with Groq offers Nvidia further expansion opportunities. When asked whether the agreement will lead to the launch of new chips specially for inference, Jensen Huang declined to promise, only stating “perhaps somewhere we may create something unique.”
Sheth expects Nvidia will announce some measures at its flagship March conference to address the need for faster inference chips. According to Bloomberg, at various points both startups and established companies have claimed they could compete with Nvidia, but in most cases they could not—at least not at scale or comprehensively. Yet, cracks are beginning to appear in the market.
Davies said:
“If you look at the growth rate of this industry, you see dedicated hardware. That’s how it’s always been in engineering history. You start with something general, then it grows like crazy, and then someone realizes you simply can’t have just one thing.”
Risk Warning and DisclaimerThe market has risks, investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article fit their particular circumstances. Investing accordingly is at your own risk.