NVIDIA drove San Jose this year into a storm, and 25,000 attendees flocked to the San Jose Convention Center and the surrounding city buildings. Many workshops, dialogue and panels were packed so much that people had to lean on the wall or sit on the floor, and the organizers shouted the command and lined up and ordered them.
NVIDIA is currently at the top of the world’s world and is at the top of the world’s world, with record financial, high interest margin, and no serious competitors. But for the next few months, the company faces the US tariffs and takes an unprecedented risk when changing priorities from the top AI customers.
NVIDIA CEO Jensen Huang attempted to announce confidence by releasing strong new chips, personal “supercomputers”, of course, in GTC 2025. It was a thorough selling pitch that targeted investors from NVIDIA’s imitation stocks.
Huang said during a keynote on Tuesday, “The more you buy, the more you can save.” “It’s much better than that. The more you buy, the more you buy it.”
Reasoning boom
Above all, GTC’s NVIDIA of this year tried to ensure that the attendees and others in the world would soon slow down.
During his keynote, Huang argued that the world was “wrong” about traditional AI scaling. Earlier this year, China’s AI Lab DeepSeek, which released a very efficient “reasoning” model called R1, feared among investors that NVIDIA’s monster chips could no longer need to train a competitive AI anymore.
However, Huang repeatedly insisted that the inferred model of inferiority would actually be more demand for the company’s chips. That’s why Huang introduced the Vera Rubin GPU, the next line of NVIDIA, and insisted that it would infer almost twice the ratio of NVIDIA’s best Blackwell chip (ie, run the AI model).
NVIDIA’s threat to business Huang was a new company like the brain, groq and other inexpensive reasoning hardware and cloud providers. Almost all hyperscalers are developing customized chips for reason if they are not training. AWS has Graviton and Ferentia (is known to be active), Google has a TPU and Microsoft has cobalt 100.
Following the same context, the technology giants are currently very dependent on NVIDIA chips, including Openai and META. I want to reduce this relationship through my in -house hardware effort. If the other competitors mentioned above are successful, they will almost weaken NVIDIA’s strokes in the AI chip market.
Perhaps NVIDIA’s share price fell about 4% after Huang’s keynote. Investors would have had hope for the last one or accelerated launch period. After all, no one got.
Tariffs
NVIDIA also suffered from worrying about the tariffs of the GTC 2025.
The United States did not impose tariffs on Taiwan (where NVIDIA gets most chips), Huang argued that tariffs would not cause “serious damage” in the short term. He did not promise that NVIDIA would be protected from the long -term economic impact.
NVIDIA has clearly received the Trump administration’s “American First” message, and Huang will help the company diversify the supply network, promising to spend hundreds of millions of dollars in US manufacturing, and the massive cost of NVIDIA depends on healthy profit margins.
New business
GTC’s NVIDIA has noted a new investment in Quantum, a historically ignored industry, in order to seed and grow business other than the core chip line. In GTC’s first Quantum Day, Huang apologized to the CEOs of major bilateral companies who suggested that technology is not useful for the next 15 to 30 years in January 2025.
On Tuesday, NVIDIA announced that it will open a new center in NVAQC Boston to develop quantum computing in cooperation with “main” hardware and software markers. Of course, the center will be equipped with NVIDIA chips, and the company says the company can simulate the models necessary for modifying quantum systems and quantum errors, the company says.
In the more immediate future, NVIDIA sees it as a potential new profit producer to call it “personal AI SuperComputers”.
In the GTC, the company has launched DGX SPARK and DGX Station, and both users are designed to prototype, fine -tune and execute the AI model in various sizes at the edge. They are not exactly cheap. They are now retailed for thousands of dollars, but Huang boldly declared that it represents the future of a private PC.
Huang said in a keynote speech, “This is a computer of the AI era.” This is how the computer will look, and this will run in the future. “
If the customer agrees, you will soon know.
