Jensen Huang never misses an opportunity to hype the crowd, and this time he did it over Korean fried chicken. No joke.
Following a dinner with NVIDIA and SK Hynix engineers at a restaurant in Santa Clara, California, the NVIDIA CEO sat down with South Korea’s The Korea Economic Daily and dropped a statement that immediately broke the internet: “At next month’s GTC 2026, we’ll unveil a chip that will surprise the world.”
And that’s it. No name, no specs, no hints. Just Jensen being Jensen, building the suspense like a season finale cliffhanger.
GTC 2026 is scheduled for March 16 to 19 in San Jose, with Huang’s keynote set for March 16. So yes, we have a date. Now we just need to figure out what’s actually hiding behind that leather jacket.
What could this mystery chip actually be?
The tech community went into full speculation mode the moment those quotes hit the web, and honestly, there are a few solid candidates on the table.
The most logical guess points toward something tied to the Vera Rubin platform. At CES 2026, NVIDIA showcased its Vera Rubin AI lineup, revealing it was in full production and included six newly designed chips, including the Vera CPUs and Rubin GPUs. So a derivative or a deeper Rubin reveal wouldn’t be shocking, but Huang’s choice of words suggests something bigger.
Then there’s the wild card: Feynman. Feynman is the platform slated to follow Rubin around 2028 and was first added to the company’s roadmap at GTC in 2025.
It is rumored to tackle the “Moore’s Law limit” by introducing technologies like silicon photonics, which uses light for data transfer instead of electricity, and would be built on TSMC’s A16 manufacturing process. If Huang pulls a Feynman prototype out of his sleeve two years early, that would 100% qualify as a world-class surprise.
Oh, and don’t sleep on the AI SSD angle either. Several reports indicate that NVIDIA has partnered with SSD makers such as SK Hynix and Kioxia to develop “AI SSD” hardware, reportedly targeting 100 million IOPS of performance, accelerating AI workloads constrained by the speed of traditional SSDs.
Could NVIDIA’s first AI storage chip be what’s coming? It’s not as flashy as a GPU reveal, but it would still be a major move.
The bigger picture behind the hype
What’s interesting here isn’t just the chip teaser, it’s the context. Huang’s promise comes amid intense competition from rivals like AMD and the maturing custom silicon from Google (TPU), Amazon (Trainium), and Microsoft (Maia), which have started closing the performance gap in specific workloads, giving customers real alternatives.
NVIDIA knows the stakes. And when Jensen Huang promises to “surprise the world,” he’s not doing it casually. The man has a track record.
During that same dinner conversation, Huang dismissed AI bubble talk entirely, saying: “There is no bubble. We’re at the start of the largest infrastructure project in human history, tens of trillions of dollars.”
Bold words. March 16 can’t come fast enough. What do you think Jensen is hiding, a Rubin variant, an early Feynman prototype, or something none of us saw coming? Drop your best guess in the comments!