Last year, the annual GTC conference in NVIDIA – which was exposed to “Woodstock of Ai” – established a crowd from 18,000 to a packed square worthy of rock legends such as The Rolling Stones. On the stage, CEO JenSen Huang, who wears a bright black leather jacket, delivered his main delivery to the annual AI Chip Beheemoth conference with addresses of addresses.
Today, a year later, Huang was on stage again, as he launched a series of shirt guns and wears this time in a black -class motorcycle jacket that deserves a width at the end of the first half. This time, NVIDIA observers reported the “Super Bowl Of Ai” borrowing like football. Nafidia He was not ashamed to compare the pork skin, as he presented a major event “before the match” and the live broadcast that had guest commentators such as the CEO of Dell Michael Dell Calling on how NVIDIA continues to rule the world of artificial intelligence.
When Huang took the theater in front of the stadium size of the NVIDIA headquarters-it is certain that the 3D Gaussian Sliging “Gaussian Sliging” is a clear “Gaussian Sliging” of his fans in the field of technology-his message was clear, even if it was not announced: the best defense in NVIDIA is a strong crime. With modern thinking models from the start of the Chinese Deepseek, the IQ is vibrated, followed by other companies including Openai, Baidu and GoogleNVIDIA wants its businessmen to know that they need more graphics processing units and programs than ever.
That is because the R1 Deepseek model, which first appeared in January, created some doubts about the NVIDIA momentum. The new model, as its maker claimed, claimed a small part of the cost and computing in the American models. As a result, NVIDIA shares are being beaten by investors who are concerned that companies will not need to buy the largest number of NVIDIA chips.
Thinking models require more computing power
But Huang believes that those who sell made a big mistake. He said that thinking models require more Computing power, not less. A a lot Moreover, in fact, thanks to their most detailed answers, or in the language of the people of artificial intelligence, “inference”. The Chatgpt Revolution revolves around Chatbot sparkling answers to inquiries – but you should “think” of today’s models, which require more “symbols”, or the basic text models of the basic units – whether it is a word in a phrase or only part of a word.
The higher the symbols used, the more efficient customers, the more efficiently require computing energy models. So, make sure that NVIDIA customers can process more symbols, faster, is a non-secret NVIDIA-and Huang did not need to mention Deepseek up to one hour in the keyword to get this point.
All NVIDIA GTC ads have been taken into account. Securities monitors may have wanted to see an accelerated schedule for the new AI chip from NVIDIA, which is Vera Robin, which will be released at the end of 2026, or more details about the company’s short -term road map. But Huang focused on the fact that, although Amnesty International insisted over the past year that the pace of artificial intelligence with just a slowing improvement, NVIDIA believes that the improvement of Amnesty International improvements in the “range” is getting faster than ever. Of course, this will be in favor of NVIDIA in revenue. “The amount of the account we need as a result of the artificial intelligence agent, as a result of thinking, is easily more than 100 times, which we thought we needed this time last year,” Huang said.
Will NVIDIA efforts to push growth will be enough to win?
The NVIDIA ads that followed all of this were related to ensuring that customers understand that they would get everything they need to keep pace with a world where the extreme speed in providing detailed answers and the best thinking is the difference between the company’s success and its failure in the company. Blackwell GPUS, the latest NVIDIA chips, the top of the AI chips, is in full production – with 3.6 million of them already used. Developed version, Blackwell Ultra, features 3x performance. The new Vera Robin chip and the road infrastructure. The “smallest AI’s super computer in the world” in the world is ready. A program for artificial intelligence agents is quickly used in the material world, including self -driving cars, robots and manufacturing.
But will NVIDIA efforts to advance growth will be sufficient to maintain the investment of institutions companies in NVIDIA products? Will the purchase of the AI chips from NVIDIA- which can cost between $ 30,000 to $ 40,000 each, and expensive, due to artificial intelligence investments that are still unlimited? Ultimately, the excellent NVIDIA options and scoops require enough customers ready to keep drilling.
Huang is sure that there is enough-and that Super Bowl’s victory in Nvidia is not just a victory for the 31-year-old company. “Everyone wins,” insisted.
Perhaps, but there is no doubt that while Nvidia seeks to create a strain in the era of artificial intelligence, expectations remain higher than ever. Huang seems to be, for his part, is not interested even as it continues to develop at a high speed. It always reaches the copper ring, apparently – or in this case, the Super Bowl episode.
This story was originally shown on Fortune.com
