Nvidia led San José by assault this year, with a record of 25,000 attendees going mass to the San José Convention Center and the surrounding buildings of the center. Many workshops, talks and panels were so full that people had to lean against the walls or sit on the floor, and suffer the anger of the organizers shouting commands to align correctly.
Nvidia is currently at the top of the AI world, with record financeHigh margins of gain, and there are still no serious competitors. But the coming months also have an unprecedented risk for the company, since it faces US tariffs, Veteranand changing priorities of the best customers of AI.
In GTC 2025The CEO of Nvidia, Jensen Huang, tried to project trust, presenting powerful new chips, Personal “supercomputers And, of course, Really nice robots. It was an exhaustive sales argument: one aimed at investors staggering Nvidia’s Pico stock.
“The more you buy, the more you save,” Huang said at a time during a key note on Tuesday. “It’s even better than that. Now, the more you buy, the more you will.”
Inference boom
More than anything, Nvidia in this year’s GTC sought to assure attendees, and the rest of the world observing, that the demand for their chips will not be reduced in the short term.
During its opening, Huang He said that almost the “whole world was wrong” in Traditional scaling that falls out of fashion. Lab of the Chinese Deepseek, which earlier this year launched a highly efficient “reasoning” model called R1It caused fears among investors that Nvidia monstrous chips may no longer be necessary to train the competitive AI.
But Huang has repeatedly insisted that hungry energy models, in fact, will generate more demand from the company’s chips, no less. That is why in GTC, Huang showed The next line of Vera Rubin GPU of Nvidia, claiming that they will make inference (that is, execute models of AI) at approximately twice the rate of the best Nvidia current Blackwell chip.
The threat to the Nvidia business, Huang, spent less time going to the principles of brains, Groq and other low -cost inference hardware suppliers and clouds. Almost all hyperscaler are developing a personalized chip for inference, if not training, too. AWS has graviton and inferentia (that, according to reports, is aggressively discounting), Google has TPUS and Microsoft has Cobalt 100.
Along the same line, technological giants that currently depend on NVIDIA chips, including OpenAi and Meta, seek to reduce those ties through internal hardware efforts. If they, and the other rivals mentioned above, are successful, will almost surely weaken the Nvidia domain in the AI chip market.
That’s why maybe The price of Nvidia shares is based on Around 4% after the key note of Huang. Investors could have maintained the hope of “one last thing”, or perhaps an accelerated launch window. In the end, they did not realize.
Tariff tensions
Nvidia also sought to relieve concerns about tariffs in GTC 2025.
The United States has not imposed tariffs on Taiwan (where Nvidia gets most of its chips) and Huang reclaimed Tariffs would not “significant damage” in the short term. However, he stopped in which Nvidia would be protected from long -term economic impacts, any form they finally take.
Nvidia has clearly received the “America First” message from the Trump administration, with Huang promising GTC To spend hundreds of billions of dollars in manufacturing in the United States, while that would help the company diversify its supply chains, it is also a massive cost for NVIDIA, whose multimillion -dollar assessment depends on healthy gains margins.
New business
As you seek to sow and grow businesses other than its line of central chips, NVIDIA in GTC caught attention to its new investments in Quantum, an industry that the company has historically neglected. On the first quantum day of GTC, Huang apologized to the CEO of the main quantum companies for causing an accident of minor actions in January 2025 after suggesting that technology would not be very useful during the next 15 to 30 years.
Tuesday, Nvidia announced which would open a new center in Boston, NVAQC, to advance quantum computing in collaboration with “leaders” hardware and software markers. The center, of course, will be equipped with NVIDIA chips, which according to the company will allow researchers to simulate quantum systems and the necessary models for the correction of quantum errors.
In the most immediate future, Nvidia sees what it is Call “IA personal supercomputers” as a potential new income manufacturer.
In GTC, the company launched DGX Spark (previously called Project digits) and the DGX station, which are designed to allow users to prototyp, adjust and execute AI models in a range of sizes on the edge. Neither of them is exactly economic, they are sold for thousands of dollars, but Huang proclaimed boldly representing the future of personal PC.
“This is the AU era,” Huang said during his key note. “This is how computers should be, and this is what computers will execute in the future.”
We will see if customers agree.