GTC felt more bullish than ever, but Nvidia’s challenges are piling up

Nvidia, the 25,000 participants around the San Jose Convention Center and surrounding buildings, the record was stormed by San Jose with 25,000 participants. Many workshops, negotiations and panels were packed so that people should be leaned on the walls or sit on the ground and the organizers were angry with the orders that they are in the correct order.

NVIDIA is currently sitting at the beginning of the AI ​​world, with Record breaking financeThe sky-high profit is the limit and yet there are serious competitors. However, the next month also keeps an unprecedented risk for the company for facing US tariffs, DepthAnd change the priorities from the best AI customers.

Horse GTC 2025NVIDIA CEO Jensen Huang, confidence, tried to open Strong new chips, Private “supercomputers” and, of course, Really lovely robots. This was a full sales pitch – someone aimed at investors who reboot from Nvidia roar.

“How much do you save, how much do you save,” Huang said at once during an opening on Tuesday. “It’s better than that. Now, how much you get it.”

Incontinence boom

More than everything, Nvidia wanted to provide this year’s GTC participants and the rest of the world – this will not slow down the demand for its chips.

Huang during Keynote Claimed nearly “that all the world’s wrong” side As a traditional AI scale. The Chinese AI Lab DeepSeek, which is earlier this year Left a highly efficient “grounding” model named R1NVIDIA’s monster chips have already asked for fears among investors whose competitive AI is not necessary.

However, Huang, power-acting models, in fact, insisted many times that the company will need more demand. Therefore in GTC, Huang showed The next line of NVIDIA claims that NVIDI will result in the approximately twice ratio of the best Blackwell Chip of Blackwell Chip (ie AI models).

Nvidia’s business Huang has risen to less time seeking less time, cerebras, GRQ and other low-precious inferences and cloud providers. If you do not practice almost every hyperscaler, but also develop a special chip to surrender. AWS has Graviton and Inferentia (It is reported to be a discount as an aggressor), Google has TPUS and Microsoft Cobalt 100.

Technological giants, which are very confident in NVIDIA plants, including Openai and Meta, want to reduce these ties at home hardware effort. If they are successful, and the above opponents are successful, almost the NVIDIA’s stranger in the AI ​​chips market.

Perhaps that’s what NVIDIA’s stock price was immersed Around 4% after the main part of Huang. Investors could have hoped for a “one last thing” or perhaps an accelerated launch window. In the end, what they didn’t.

Tariff tension

Nvidia also tried to look for concerns about tariffs on GTC 2025.

The United States does not apply any tariffs in Taiwan (which Nvidia has received the most chips) and Huang claimed Tariffs would not “significant damage” in a short time. For a period of time promising that NVIDIA protects against long-term economic effects, no matter what.

Nvidia obviously obtained the “America’s first” message of the Trump Administration with Huang Collateral in GTC To spend hundreds of dollars in the United States to spend hundreds of dollars, it will help the company supply chains, as well as a large number of valuable evaluation, healthy profit.

New work

As NVIDIA’s nucleus of NVIDIA’s core and growth, in GTC, NVIDIA noted the company’s new investments in an industry that historically neglected. On the first quantum day of GTC, Huang apologized After proposing the General Diriors of large quantum companies, which caused a small stock accident in January 2025, it would not be very useful for the next 15-30 years of technology.

Tuesday, Nvidia declared Boston will open a new center in the NVAQC to ensure quantum calculation in cooperation with “leading” hardware and software markers. The center will, of course, will be equipped with NVIDIA chips in the company, which will allow researchers to simulate the models necessary to correct quantum systems and quantum errors.

Nvidia sees what is in the near future Calls “Private AI supercomputers” as a potential new income manufacturer.

Company in GTC has initiated DGX spark (previously called) Project figures) and both are designed to use the prototype, subtle regulation and AI models in a series of sizes. What’s completely cheap – they sell for thousands of dollars – but Huang boldly said they represent the future of the individual computer.

“This is the computer’s computer,” Huang said during his words. “It should look like computers and that’s what computers work in the future.”

We will soon see that customers agree.