Nvidia took San Jose by storm this 12 months, with a record-breaking 25,000 attendees flocking to the San Jose Conference Middle and surrounding downtown buildings. Many workshops, talks, and panels have been so packed that folks needed to lean towards partitions or sit on the ground — and endure the wrath of organizers shouting instructions to get them to line up correctly.
Nvidia presently sits on the prime of the AI world, with record-breaking financials, sky-high revenue margins, and no severe rivals but. However the coming months additionally maintain unprecedented danger for the corporate because it faces U.S. tariffs, DeepSeek, and shifting priorities from prime AI prospects.
At GTC 2025, Nvidia CEO Jensen Huang tried to undertaking confidence, unveiling highly effective new chips, private “supercomputers,” and, after all, actually cute robots. It was an exhaustive gross sales pitch – one geared toward buyers reeling from Nvidia’s nosediving inventory.
“The more you buy, the more you save,” Huang stated at one level throughout a keynote on Tuesday. “It’s even better than that. Now, the more you buy, the more you make.”
Inference increase
Greater than something, Nvidia at this 12 months’s GTC sought to guarantee attendees – and the remainder of the world watching – that demand for its chips received’t decelerate anytime quickly.
Throughout his keynote, Huang claimed that almost the “entire world got it wrong” on conventional AI scaling falling out of vogue. Chinese language AI lab DeepSeek, which earlier this 12 months launched a extremely environment friendly “reasoning” mannequin referred to as R1, prompted fears amongst buyers that Nvidia’s monster chips might not be crucial for coaching aggressive AI.
However Huang has repeatedly insisted that power-hungry reasoning fashions will, in truth, drive extra demand for the corporate’s chips, not much less. That’s why at GTC, Huang confirmed off Nvidia’s subsequent line of Vera Rubin GPUs, claiming they’ll carry out inference (that’s, run AI fashions) at roughly double the speed of Nvidia’s present finest Blackwell chip.
The risk to Nvidia’s enterprise Huang spent much less time addressing was upstarts like Cerebras, Groq, and different low-cost inference {hardware} and cloud suppliers. Practically each hyperscaler is creating a customized chip for inference, if not coaching, as properly. AWS has Graviton and Inferentia (which it’s reportedly aggressively discounting), Google has TPUs, and Microsoft has Cobalt 100.
Alongside the identical vein, tech giants presently extraordinarily reliant on Nvidia chips, together with OpenAI and Meta, want to cut back these ties by way of in-house {hardware} efforts. In the event that they – and the aforementioned different rivals – are profitable, it’ll nearly assuredly weaken Nvidia’s stranglehold on the AI chips market.
That’s maybe why Nvidia’s share worth dipped round 4% following Huang’s keynote. Buyers would possibly’ve been holding out hope for “one last thing” — or maybe an accelerated launch window. In the long run, they received neither.
Tariff tensions
Nvidia additionally sought to allay worries about tariffs at GTC 2025.
The U.S. hasn’t imposed any tariffs on Taiwan (the place Nvidia will get most of its chips), and Huang claimed tariffs wouldn’t do “significant damage” within the quick run. He stopped in need of promising that Nvidia can be shielded from the long-term financial impacts, nevertheless — no matter kind they in the end take.
Nvidia has clearly obtained the Trump Administration’s “America First” message, with Huang pledging at GTC to spend a whole lot of billions of {dollars} on manufacturing within the U.S. Whereas that will assist the corporate diversify its provide chains, it’s additionally an enormous value for Nvidia, whose multitrillion-dollar valuation is determined by wholesome revenue margins.
New enterprise
Because it appears to seed and develop companies apart from its core chips line, Nvidia at GTC drew consideration to its new investments in quantum, an business that the corporate has traditionally uncared for. At GTC’s first Quantum Day, Huang apologized to the CEOs of main quantum corporations for inflicting a minor inventory crash in January 2025 after he recommended that the tech wouldn’t be very helpful for the following 15 to 30 years.
On Tuesday, Nvidia introduced that it could open a brand new heart in Boston, NVAQC, to advance quantum computing in collaboration with “leading” {hardware} and software program markers. The middle will, after all, be outfitted with Nvidia chips, which the corporate says will allow researchers to simulate quantum programs and the fashions crucial for quantum error correction.
Within the extra quick future, Nvidia sees what it’s calling “personal AI supercomputers” as a possible new revenue-maker.
At GTC, the corporate launched DGX Spark (beforehand referred to as Challenge Digits) and DGX Station, each of that are designed to permit customers to prototype, fine-tune, and run AI fashions in a variety of sizes on the edge. Neither is strictly cheap – they retail for hundreds of {dollars} – however Huang boldly proclaimed that they characterize the way forward for the private PC.
“This is the computer of the age of AI,” Huang stated throughout his keynote. “This is what computers should look like, and this is what computers will run in the future.”
We’ll quickly see if prospects agree.
