Jensen Huang Just Landed in Taiwan With a Secret Product Nobody Has Seen — Computex 2026 Could Change Everything

Computex 2026 Jensen Huang is about to deliver what may be the most consequential keynote in the trade show’s 45-year history. The Nvidia CEO touched down in Taipei alongside AMD’s Lisa Su, and he brought three things with him: the Vera Rubin AI platform that promises to obliterate every performance record in AI computing, a “surprise new product that we haven’t told anyone about yet,” and plans for a $100 billion Taiwan spending commitment that dwarfs any foreign technology investment the island has ever seen.

His keynote is scheduled for Monday, June 1, 2026, at 11:00 AM local time at the Taipei Music Center. But the real show has already started.

Computex 2026: Jensen Huang Arrives in Taipei

Hours after landing, Huang made an unannounced stop at “Meet-a-Claw,” a developer event where engineers are getting hands-on with autonomous AI agents built on Nvidia’s platforms. The visit — captured in Nvidia’s own Instagram post with the caption “Consider GTC week officially started” — signals that Computex 2026 Jensen Huang’s agenda goes far beyond traditional hardware announcements.

This year’s Computex, running from June 1-5, is positioned as the most important hardware show of the year. While CES in January focuses on consumer electronics and Mobile World Congress covers telecommunications, Computex has become ground zero for the AI infrastructure that powers everything else. And Nvidia is the undisputed headliner.

The setting is appropriate: Taiwan manufactures the vast majority of the world’s most advanced semiconductors, and Nvidia’s relationship with TSMC — the company that fabricates Nvidia’s chips — is one of the most consequential partnerships in technology. Huang flew directly to TSMC before the conference, underscoring the depth of that relationship.

Vera Rubin: The Chip That Makes Blackwell Look Slow

The centerpiece of Nvidia’s Computex 2026 presence is the Vera Rubin platform, the company’s next-generation AI computing architecture. Named after the astronomer who proved the existence of dark matter, Vera Rubin combines the Vera CPU with the Rubin GPU into a six-chip system designed for data center-scale AI workloads.

The performance claims are staggering. According to Morgan Stanley’s analysis, Vera Rubin delivers approximately 3.5 times the training performance and 5 times the inference performance of its predecessor, the Blackwell platform. Inference costs drop to one-seventh of current levels. For AI companies spending billions on compute — like the frontier labs racing to train ever-larger models — these improvements translate directly into either massive cost savings or the ability to train fundamentally more capable models on the same budget.

The six-chip system architecture represents a shift in how Nvidia designs its highest-end products. Rather than relying on a single monolithic chip, Vera Rubin uses a chiplet design where multiple specialized dies work together, connected by high-speed interconnects. This approach mirrors what AMD has done with its EPYC server processors and allows Nvidia to push performance beyond what’s possible with a single piece of silicon.

The Surprise Product Nobody Has Seen

The statement that has the tech world on edge came directly from Huang himself: “The second half of this year is going to be very, very busy with Grace Blackwell, Vera Rubin, and we have a surprise new product that we haven’t told anyone about yet.”

Speculation about the mystery product has been intense. The leading theories include the rumored Nvidia N1X — an ARM-based personal computer processor that would pit Nvidia directly against Apple’s M-series and Qualcomm’s Snapdragon X Elite. Tom’s Guide and other outlets have reported on the N1X possibility, noting that Nvidia has the GPU expertise and ARM architecture license to build a competitive PC chip.

Other possibilities include a new category of AI accelerator designed specifically for inference rather than training, a networking product that extends Nvidia’s reach deeper into the data center stack, or an entirely new hardware category that doesn’t fit existing classifications. Huang’s history of dramatic product reveals — remember the leather jacket on stage at GTC? — suggests he’s saving something genuinely significant for the Computex keynote.

Whatever it is, the fact that Nvidia is keeping it secret this close to the announcement suggests it’s either highly competitive or represents a new market entry that Nvidia doesn’t want rivals to pre-empt.

Nvidia’s $100 Billion Taiwan Bet

Perhaps the most jaw-dropping number to emerge from pre-Computex briefings is Nvidia’s Taiwan spending. Huang revealed that Nvidia’s annual spending in Taiwan has climbed from $10-15 billion five years ago to $100 billion today. That’s not a typo — Nvidia is spending $100 billion per year in a single country, primarily on chip fabrication at TSMC.

To put this in context, Taiwan’s entire GDP is approximately $800 billion. Nvidia’s spending represents roughly 12.5% of the country’s economic output. No single foreign company has ever had this level of economic influence over Taiwan, and it underscores both the extraordinary scale of AI chip demand and the concentration risk that comes with relying on a single island for the world’s most critical technology.

The spending also reflects TSMC’s pricing power. As the only foundry capable of manufacturing chips at the 3nm and 2nm nodes that Nvidia requires, TSMC can charge premium prices. Nvidia’s willingness to pay is itself a signal about how much profit margin exists in the AI chip business — you don’t spend $100 billion on manufacturing unless you’re making significantly more than that in revenue.

The Constellation Campus: Nvidia’s New Overseas HQ

Huang announced that Nvidia is building the “Constellation” campus in Taiwan, designed to serve as the company’s new overseas headquarters. While details are still emerging, the campus represents a deepening of Nvidia’s commitment to Taiwan that goes beyond just buying chips.

The Constellation campus will likely house R&D facilities, engineering teams, and potentially manufacturing support operations. For Taiwan, this is a significant economic and strategic win — having Nvidia’s overseas headquarters on the island strengthens the case for Taiwan’s continued importance in the global technology supply chain.

For Nvidia, the campus serves both practical and political purposes. Practically, being closer to TSMC accelerates the chip development cycle. Politically, it demonstrates commitment to Taiwan at a time when geopolitical tensions continue to shape technology supply chain decisions.

What Computex 2026 Means for the AI Chip Race

Computex 2026 Jensen Huang’s keynote comes at a pivotal moment for the AI chip industry. Nvidia’s dominance has been challenged from multiple directions: AMD’s MI-series accelerators are gaining enterprise traction, Google’s TPUs are improving rapidly, and custom silicon from companies like Amazon (Trainium) and Microsoft (Maia) is entering production.

International competitors are also closing in. Chinese chip designers have made significant progress despite U.S. export restrictions, and European initiatives to build domestic AI chip capabilities are gaining funding. Nvidia’s response has been to accelerate its roadmap, moving from annual to near-continuous product cycles.

The Vera Rubin announcement, combined with the mystery product, is designed to demonstrate that Nvidia isn’t just ahead — it’s accelerating away from the competition. If the 3.5x training and 5x inference performance improvements over Blackwell hold up in real-world deployments, competitors will need to make extraordinary leaps just to reach parity with where Nvidia was last year.

The Competition Waiting in Taipei

Jensen Huang isn’t the only tech CEO in Taipei this week. AMD’s Lisa Su has also arrived, and AMD is expected to announce its own next-generation AI accelerator lineup. Intel is showcasing its Gaudi and Falcon Shores AI chips. Qualcomm, MediaTek, and ARM are all presenting.

The presence of so many major chipmakers at a single event creates an atmosphere of competitive intensity that’s unusual even by tech industry standards. But all eyes will be on Huang’s Monday keynote. In the AI chip race, Nvidia sets the pace and everyone else reacts.

What to Watch on June 1

When Huang takes the stage at the Taipei Music Center on June 1, watch for these key announcements. Vera Rubin detailed specifications and availability timeline — when can customers actually buy these chips? The surprise product reveal — what has Nvidia been hiding? Partnership announcements with cloud providers, enterprise customers, and sovereign AI initiatives. And any commentary on pricing — Nvidia’s ability to raise prices while maintaining demand is the single most important factor in its financial outlook.

Computex 2026 Jensen Huang’s keynote will almost certainly be the most-watched tech presentation of the year. When the CEO of the world’s most valuable company takes the stage in the country where his chips are made, with a secret product nobody has seen, in an industry that’s reshaping the global economy — you pay attention.

June 1. Taipei Music Center. 11:00 AM. The AI chip race restarts.