The recent financial results of NVIDIA have exceeded the most optimistic expectations, however the reception of investors remained lukewarm. A reaction that obviously follows the Deepseek phenomenon, which was the equivalent of a gigantic pavement in the great fed of AI (and Wall Street at the same time). Ultra-effective open-source models that have slightly sowed doubts abouthas Nvidia ability to maintain its domination in the sector.
Jensen Huang wanted to clarify the situation at the last conference on the results of his firm and the message clearly passed: Nvidia will remain nvidia Despite these small opposite winds.
The insatiable hunger of advanced algorithms
Far from constituting a threat to Nvidia, Deepseek’s advances would represent an accelerator of the request for calculation power, according to Huang. The CEO was categorical at the conference call : ” Reasoning models can consume 100 times more computing power. Future reasoning systems will be able to consume much more ».
In other words: Deepseek, competitors? Certainly not! The more the AI develops, the more Nvidia will sell its overpowered chips, a simple equation. No fear to have concerning a potential slowdown in demand For the famous GPUs of the Santa Clara firm that the Giants of the AI tears away.
Like Sam Altman of Openai, Huang did not fail to greet the contribution of Deepseekqualifying the work of the Chinese company as a “” Excellent innovation ». He also took the opportunity to emphasize the beautiful place in the sun that his flagship model was made, Deepseek R1: ” They made everyone a model of peak reasoning. Result: almost all of the developers in AI teach R1, or are inspired by its reasoning techniques in stages and learning by strengthening, to boost the performance of their own creations ».
Inference, a new battlefield
The AI sector is far from maturity and evolves, in particular At the level of the distribution of calculation charges. “” The vast majority of our computing power today is actually devoted to inference, and Blackwell carries all this at a higher level “Said Huang in reference to the new generation of company chips.
Unlike the initial training of the models, inference represents the operational phase where AI treats real -time requests and generates specific responses. Today, this segment becomes the real economic engine of the market, transforming algorithms into marketable services. Two years ago, it was an almost theoretical concept, Nowadays, it’s a lucrative reality.
Even if the Huang Society always remains very largely first In this race, ” Competition begins to weigh on Nvidia’s position, although it is not very significant at this stage », Observes Lucas Keh, analyst at Third Bridge. Indeed, the competitive landscape was still densified in just a few months.
New companies, specialized in the design of neural networks acceleration chips (dedicated to inference) begin to emerge slowly. One could appoint Tenstorrent, a company led by Jim Keller a renowned engineer who worked at Apple and AMD, who announced nearly $ 700 million in fundraising. Etched, another American company that benefited from $ 120 million from investors last year.
Another concern for NVIDIA’s investitator: the development of PUCES IA personalized by cloud giants like Google and Amazon, likely to erode the dominant position of the company. “” We heard that their market share [NDLR : celle de NVIDIA] for inference could fall to 50 % with changes in the sector “Said Keh to Business Insider.
For the moment everything is fine, Nvidia remains very wisely seated on her thronebut it is not said that the sector does not know some stir in the medium/long term. If there is one thing that it has taught us since the Chatgpt explosion: the AI market is very unpredictable and The giants of yesterday will not necessarily be those of tomorrow.
- Despite exceptional financial results from Nvidia, investors remain cautious about the rise of Deepseek and other AI players.
- Jensen Huang claims that the boom in advanced models does not threaten Nvidia, but on the contrary reinforces the demand for computing power.
- Inference becomes the new engine of the market, attracting new competitors and raising questions about the future domination of Nvidia.