NVIDIA’s myth: it was vetoed for the first time by its own downstream partner.
Kyber-rack delivery has been pushed back to 2028, and Asian PCB stocks took a sharp dive on the spot—Yifangdian, Kingboard Laminate, and Samsung Electro-Mechanics all fell hard. Every trending headline was filled with cries of “the AI hardware bubble has burst,” but nobody asked one simple question: did the bubble burst in the stock price, or in the narrative?
The truly explosive details are hidden in the third paragraph of the news. NVIDIA came up with a backup plan: pair two old rack units back-to-back to cobble together enough compute. But cloud providers collectively rejected it, citing reasons like “too ugly, too expensive, and impossible to maintain.” Translated plainly: even the “prime contractor dads” at buyers like Amazon, Microsoft, and Google—who spend every day shouting about “betting on AI”—have started to嫌弃 NVIDIA’s engineering ability. A company that has survived on an annual cadence of “performance doubles” for years was, for the first time, overruled by downstream stakeholders with their feet.
Heat dissipation can’t be handled, power consumption goes through the roof, and PCB precision can’t be manufactured—these are physics problems that can’t be solved by drawing a roadmap PPT. Over the past two years, NVIDIA has fed Wall Street’s imagination with a “arms race”-style release schedule. But this time, imagination collided with a wall beyond Moore’s Law.
The most ironic part is that the one getting smashed may actually be the wrong stocks. What was really delayed is the Ultra rack that was not supposed to appear until 2027; this year, in the autumn, the current-generation Rubin—intended for shipment to the big eight cloud providers—will be delivered as scheduled, with no cash-flow shortfall at all. This time, the market didn’t even figure out which story went wrong—it went ahead and ground the entire supply chain into the ground.
If even NVIDIA’s execution can be questioned, shouldn’t the global liquidity that’s been drained over the past two years by AI capital expenditures also start to loosen? #英伟达AI服务器延期亚洲PCB股下挫
#英伟达
Kyber-rack delivery has been pushed back to 2028, and Asian PCB stocks took a sharp dive on the spot—Yifangdian, Kingboard Laminate, and Samsung Electro-Mechanics all fell hard. Every trending headline was filled with cries of “the AI hardware bubble has burst,” but nobody asked one simple question: did the bubble burst in the stock price, or in the narrative?
The truly explosive details are hidden in the third paragraph of the news. NVIDIA came up with a backup plan: pair two old rack units back-to-back to cobble together enough compute. But cloud providers collectively rejected it, citing reasons like “too ugly, too expensive, and impossible to maintain.” Translated plainly: even the “prime contractor dads” at buyers like Amazon, Microsoft, and Google—who spend every day shouting about “betting on AI”—have started to嫌弃 NVIDIA’s engineering ability. A company that has survived on an annual cadence of “performance doubles” for years was, for the first time, overruled by downstream stakeholders with their feet.
Heat dissipation can’t be handled, power consumption goes through the roof, and PCB precision can’t be manufactured—these are physics problems that can’t be solved by drawing a roadmap PPT. Over the past two years, NVIDIA has fed Wall Street’s imagination with a “arms race”-style release schedule. But this time, imagination collided with a wall beyond Moore’s Law.
The most ironic part is that the one getting smashed may actually be the wrong stocks. What was really delayed is the Ultra rack that was not supposed to appear until 2027; this year, in the autumn, the current-generation Rubin—intended for shipment to the big eight cloud providers—will be delivered as scheduled, with no cash-flow shortfall at all. This time, the market didn’t even figure out which story went wrong—it went ahead and ground the entire supply chain into the ground.
If even NVIDIA’s execution can be questioned, shouldn’t the global liquidity that’s been drained over the past two years by AI capital expenditures also start to loosen? #英伟达AI服务器延期亚洲PCB股下挫
#英伟达