Inside Huang's Taiwan Gambit: What NVIDIA's CEO Really Said (And Why It Matters)

A closed-door meeting in Taipei just became Silicon Valley’s biggest intel leak. On November 5, NVIDIA CEO Jensen Huang gathered 12 execs from TSMC, Foxconn, Quanta, Wistron, plus two major U.S. VC firms at the Grand Hyatt. No phones. No recordings. No notes allowed.

Within hours, everything hit Financial Times anyway.

The Core Arguments

Huang’s core thesis: China’s AI advantage is structural, not cyclical.

The numbers: Over 1 million people working on AI in China vs. ~20,000 in Silicon Valley. “That’s not competition—that’s asymmetry,” one attendee noted.

The timeline: By 2027, China will have more AI computing power than the rest of the world combined. Huawei’s Ascend 910C chips? Already at 200k units/month, only 8-12% slower than H100.

The message on sanctions: Huang reportedly called U.S. export controls “the stupidest policy ever”—not because they don’t work, but because they mobilize. “The more you sanction them, the stronger they get.”

The Real Question

Here’s what everyone’s actually debating: Did this leak happen to Huang, or for Huang?

Theory A: He was genuinely betrayed by attendees desperate to signal-boost their own importance.

Theory B: This is calculated messaging—Huang softens NVIDIA’s position on China while maintaining plausible deniability, pressures Washington without directly confronting Trump/regulators, and quietly admits what every AI executive already knows.

Either way, the message is clear: U.S. policy is losing a long game where China’s only playing one move—scale. Whether that comes from leaked speeches or quarterly earnings calls, it’s the same story.

What Changes

Nothing, probably. Sanctions stay. China keeps iterating. NVIDIA keeps navigating. But now everyone knows the CEO thinks the current strategy is broken.

That’s the real news.