No TSMC, no Nvidia—and Huang just proved he meant it.
Jensen Huang's journey to Taiwan was less a business trip than a declaration — that the appetite for artificial intelligence infrastructure has not crested, and that Nvidia intends to meet it before the wave arrives. By personally negotiating expanded wafer production with TSMC, the world's most valuable company is placing a calculated bet that the cost of scarcity outweighs the cost of commitment. In the long arc of technological transformation, this moment may be remembered as the point when AI hardware demand shifted from aspiration to industrial imperative.
- Nvidia's CEO flew to Taiwan with a specific and urgent ask: a 50% surge in TSMC's most advanced 3-nanometer chip production, signaling that current supply is already straining against future demand.
- Even as next-generation Rubin GPUs enter the production line, current Blackwell chips remain in fierce demand from the world's largest tech companies, leaving Nvidia stretched across two generational frontiers simultaneously.
- A November 19th earnings report looms as the market's moment of reckoning — analysts project 54% earnings-per-share growth, but guidance for Q4 will determine whether the AI spending supercycle is real or overstated.
- Wall Street has answered with near-unanimity: 40 of 47 analysts rate Nvidia a strong buy, with price targets stretching to $300, as UBS raises its global AI capital expenditure forecast to $571 billion by 2026.
Jensen Huang flew to Taiwan this past weekend with a single, urgent purpose: securing more chips. Meeting with executives at Taiwan Semiconductor Manufacturing Company — the foundry responsible for nearly all of Nvidia's processors — he requested a dramatic ramp in production of TSMC's most advanced 3-nanometer wafers. Reports from Taiwanese media suggest the ask could push monthly output at TSMC's Southern Taiwan facility from 100,000 wafers to 160,000, with a substantial share flowing directly to Nvidia. The precision of that figure signals something important: Nvidia has already done the math, and it has decided the danger of falling short outweighs the risk of overcommitting.
This was Huang's fourth visit to Taiwan this year. Speaking to reporters, he described Nvidia's business as growing "month by month, stronger and stronger," and offered a candid acknowledgment of the relationship at the center of it all: "No TSMC, no Nvidia." TSMC's CEO confirmed the request and told employees he expects record sales year after year. The Rubin GPU architecture — Nvidia's next generation — has already entered production, built on TSMC's N3P process and incorporating HBM4 memory for what the company calls a substantial leap in compute performance.
Meanwhile, current-generation Blackwell GPUs continue to sell as fast as they can be made, with demand extending across Nvidia's broader ecosystem of CPUs, networking gear, and switches. The company's market capitalization has reached $4.69 trillion, making it the world's most valuable corporation, with shares up 37.5% year-to-date despite recent volatility.
The next inflection point arrives November 19th, when Nvidia reports third-quarter earnings. Analysts project 54% earnings-per-share growth and revenues of $54.83 billion, with Q4 guidance expected near $61.3 billion. UBS has raised its global AI capital expenditure forecast to $423 billion for 2025 and $571 billion for 2026. Of 47 analysts covering the stock, 40 rate it a strong buy. Bank of America calls it "one of the most compelling ways to invest in the AI revolution." Citi notes that supply — not demand — has become the binding constraint. Huang's Taiwan visit, in this light, was not diplomacy. It was a supply chain power move by a CEO betting that the future he is building for will arrive exactly on schedule.
Jensen Huang flew to Taiwan over the weekend with a single, urgent request: more chips. The Nvidia CEO met with executives at Taiwan Semiconductor Manufacturing Company, the foundry that builds nearly all of Nvidia's processors, and asked them to ramp up production of their most advanced wafers. It was a straightforward business conversation, but it carried an unmistakable signal. When the leader of the world's most valuable company personally travels across the Pacific to secure additional manufacturing capacity, he is telling the market something: the artificial intelligence boom is not plateauing. It is accelerating, and Nvidia intends to own as much of it as possible.
Wafers are the silicon foundations upon which advanced chips are built. Nvidia's next-generation Rubin GPU lineup will depend heavily on TSMC's 3-nanometer process—the most cutting-edge manufacturing technology available. According to reports from Taiwanese media, Huang's request could trigger a dramatic expansion: TSMC's 3-nanometer capacity at its Southern Taiwan facility could increase from 100,000 wafers per month to 160,000, a jump of more than 50 percent. A significant portion of that new capacity would flow directly to Nvidia. The specificity of the number matters. It suggests Nvidia has already modeled its demand, calculated what it will need, and decided the risk of overcommitting is smaller than the risk of running short.
During his fourth visit to Taiwan this year, Huang told reporters that Nvidia's business was "very strong, and it's growing month by month, stronger and stronger." He acknowledged the obvious: "No TSMC, no Nvidia." TSMC's chief executive, C.C. Wei, confirmed that Huang had requested increased wafer production and told employees he expects the company to achieve record sales year after year. The Rubin architecture, Huang noted, has already entered the production line. It will feature TSMC's advanced N3P process and incorporate HBM4 memory technology, delivering what the company describes as a substantial compute leap over current generations.
Nvidia's current-generation Blackwell and Blackwell Ultra GPUs are still in furious demand. Huang emphasized that this demand extends beyond processors alone—Nvidia manufactures CPUs, networking equipment, and switches that all feed into the Blackwell ecosystem. The company's market capitalization has reached $4.69 trillion, making it the world's most valuable corporation. Year-to-date, Nvidia stock has climbed 37.5 percent, though it has weathered recent volatility tied to broader tech sector concerns and a $5.83 billion stake sale by Japan's SoftBank Group.
The real test comes Wednesday, November 19th, when Nvidia reports third-quarter earnings after the market closes. Analysts project earnings per share will grow 54 percent year-over-year to $1.25, with revenue climbing 56 percent to $54.83 billion, driven primarily by data center sales. Fourth-quarter guidance is expected to reach approximately $61.3 billion in revenue. UBS has raised its forecast for global artificial intelligence capital expenditure to $423 billion for 2025 and $571 billion for 2026, up from earlier estimates of $375 billion and $500 billion. That spending is flowing directly into Nvidia's coffers.
Wall Street remains decidedly bullish. Of 47 analysts covering the stock, 40 rate it a "Strong Buy," with an average price target of $234.12, implying 27 percent upside from current levels. Bank of America reaffirmed a "Buy" rating and $275 price target, calling Nvidia "one of the most compelling ways to invest in the AI revolution." Citi lifted its target to $220, noting that supply—not demand—has become the constraint. Melius Research set a $300 target, citing strong hyperscaler buying and positive commentary on demand trends. The consensus is clear: Huang's Taiwan visit was not a courtesy call. It was a supply chain power move, executed by a CEO confident enough in future demand to lock in production capacity now. What happens next depends on whether that confidence proves justified.
Citações Notáveis
The business is very strong, and it's growing month by month, stronger and stronger.— Jensen Huang, Nvidia CEO
Supply, not demand, remains the key bottleneck in AI chips.— Citi analyst Atif Malik
A Conversa do Hearth Outra perspectiva sobre a história
Why does it matter that Huang personally went to Taiwan? Couldn't he have just called?
Because showing up in person signals desperation—or rather, confidence so strong it looks like desperation. You don't fly across the world unless the stakes are existential. It tells TSMC, "We need this," and it tells the market, "We know what's coming."
But TSMC already makes chips for Nvidia. Why would they need convincing?
They don't need convincing to make chips. They need convincing to expand capacity. That's capital, that's risk, that's betting on Nvidia's forecast being right. Huang is essentially saying, "Trust me enough to build factories."
What if the AI boom slows down? Doesn't Nvidia risk having too much supply?
Absolutely. But Huang is betting that the downside of being caught short is worse than the downside of having excess capacity. If demand explodes and Nvidia can't deliver, competitors move in. If demand softens and Nvidia has extra wafers, they absorb the cost. One kills the business; the other is just expensive.
The earnings report is coming in a few days. Is that why the stock has been volatile?
Partly. But the real volatility is about whether Huang's optimism is real or just salesmanship. The earnings will show whether AI spending is actually accelerating or just holding steady. If it's accelerating, his Taiwan trip looks prescient. If it's flat, it looks like he overcommitted.
What does "supply, not demand" being the bottleneck actually mean?
It means every Nvidia chip that exists is already sold. The constraint isn't convincing people to buy—it's physically making enough units. That's why Huang is in Taiwan asking for more wafers. He has orders he can't fill.
So the next-generation Rubin chips—are those already spoken for too?
Huang said they've entered production. That usually means major customers have already committed. He wouldn't ramp production on a new architecture unless he had visibility into demand. The 50 percent capacity increase suggests he has a lot of visibility.