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

Nvidia conquered San Jose this year with storms, with a record -breaking 25,000 attendees who flowed to the San Jose Convention Center and surrounding buildings in the city center. Many workshops, conversations and panels were so full that people had to lean against walls or sit on the floor – and the wrath of organizers suffered those commands to get them well in line.

Nvidia is currently at the top of the AI ​​world, with record-breaking financial data, sky-high profit margins and not yet serious competitors. But in the coming months there are also an unprecedented risk for the company because it is confronted with American rates, deep key and shifting priorities of top AI customers.

On GTC 2025, Jensen Huang, CEO of Nvidia, tried to project trust, to reveal powerful new chips, personal “supercomputers”, and of course really cute robots. It was an exhausting sales talk – a focused on investors who are faltering from the diving stock of Nvidia.

“The more you buy, the more you save,” Huang said at a certain point during a keynote on Tuesday. “It’s even better than that. Now, the more you buy, the more you make.”

Inference

More than whatever, Nvidia on this year’s GTC tried to insure those present – and the rest of the world – that demand for his chips will not slow down.

During his keynote, Huang claimed that almost the “whole world was wrong” on traditional AI scaling that fell from Vogue. Chinese AI Lab Deepseek, which released a very efficient “reasoning model” R1 earlier this year, led to fears among investors that Nvidia’s sample chips might no longer be necessary for training competitive AI.

But Huang has repeatedly demonstrated that power -hungry reasoning models will in fact stimulate more demand for the company’s chips, no less. That is why Huang showed up the next line of Nvidia by Vera Rubin GPUs at GTC, claiming that they will perform inference (ie AI models) with about double the current Best Blackwell chip from Nvidia.

The threat for Nvidia’s business Huang spent less time tackling on starts such as cerebras, groq and other cheap inference hardware and cloud providers. Almost every hyperscaler develops an adapted chip for inference, if not training, too. AWS has graviton and inferentia (who is reportedly agreed aggressively), Google has TPUs and Microsoft has Cobalt 100.

Along the same spirit, Tech giants currently want extremely dependent on Nvidia chips, including OpenAi and Meta, which reduce tires through internal hardware efforts. If they – and the other rivals mentioned above – are successful, this will almost certainly weaken the stranglehold of Nvidia on the AI ​​-chips market.

That is perhaps the reason why Nvidia’s share price fell around 4% after the Keynote of Huang. Investors may have kept hope for “one last thing” – or perhaps an accelerated launch window. In the end they have neither.

Rate tensions

Nvidia also tried to worry about rates on GTC 2025.

The US has not imposed rates on Taiwan (where Nvidia gets the most chips), and Huang claimed that rates would not cause “considerable damage” in the short term. He did not stop promising that Nvidia would, however, be protected from the economic effects in the long term, which ultimately also accept them.

Nvidia clearly received the “America First” message from the Trump administration, with Huang promise on GTC To spend hundreds of billions of dollars on production in the US, while this would help the company to diversify its supply chains, it is also a huge costs for NVIDIA, whose multitrillion-dollar rating depends on healthy profit margins.

New case

Because it wants to sow and grow Core Chips line other than its Core Chips, Nvidia draws attention to its new investments in Quantum at GTC, an industry that the company has historically neglected. On the first quantum day of GTC, Huang apologized to the CEOs of large quantum companies for causing a small stock crash in January 2025 after he suggested that the technology would not be very useful for the next 15 to 30 years.

On Tuesday, Nvidia announced That it would open a new center in Boston, NVAQC, to promote Kwantum Computing in collaboration with “leading” hardware and software markings. The center will of course be equipped with NVIDIA chips, the company says that researchers can simulate quantum systems and the models needed for quantum error correction.

In the more direct future, Nvidia sees what the ‘personal AI-SUPERCOMPUTERS’ calls as a potential new income maker.

At GTC, the company launched DGX Spark (previously called Digits Project) and DGX station, both of which were designed to enable users to make prototype, to refine and perform AI models in different sizes on the edge. Neither of them is exactly cheap – they sell thousands of dollars – but Huang stated courageously that they represent the future of the personal PC.

“This is the computer of the Ai era,” said Huang during his keynote. “This is what computers should look like, and this is what computers will be performed in the future.”

We will soon see whether customers agree.