Post

Jensen Huang personally steps in to "debunk rumors": Is the $500 billion not a circular financing? Do you believe it? Last night, Nvidia did something big. Together with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR—six Wall Street giants—they plan to create a $500 billion AI computing power financing platform. Jensen Huang personally appeared on CNBC to announce it, with top executives from the six institutions rarely sharing the stage. "Not a single one refused," he said. Then Nvidia's stock price dropped 2.86%, wiping out about $13 billion in market value. A $500 billion "super positive" news triggered an intraday plunge, and the Philadelphia Semiconductor Index closed down 2.94%. The bigger the good news, the harder the fall. Even stranger was another event— Nvidia's 5-year credit default swap (CDS) surged nearly 6 basis points on Monday, rising to 77.215 basis points, marking the largest single-day increase in two weeks. Since late May, the cost of default insurance on Nvidia's debt has doubled, from 41.6 basis points to 77.5 basis points. A company announces it will help clients raise $500 billion, yet the market starts worrying it can't repay its debts. Think about that logic carefully. Jensen Huang couldn't sit still. On August 10, he personally responded to doubts on social platform X. The positive side, Huang said this— First, the money is not mine. "This $500 billion represents third-party capital, neither Nvidia's revenue nor a commitment from a single fund or to a single client." Translation: I'm just the connector, not putting up the money. Second, AI factories are already "investable assets." Huang said the AI industry has moved from the era of "enterprises purchasing chips project by project" to a new stage where "AI factories can be financed as productive infrastructure." "This is indeed the first time technology chips have become an investable asset class." Third, hardware is not fast-moving consumer goods; it is a long-term asset. He cited the A100 as an example: continuous CUDA software upgrades keep the hardware commercially viable for six years; computing power assets are not quickly depreciating electronic waste. "An AI factory can serve multiple clients and workloads simultaneously; when one client's demand changes, another client can use it. This broad ecosystem helps protect residual value." In plain language: my chips are not disposable; they're egg-laying hens. But the market is unconvinced. The opposing view is louder— Doubt one: CDS doesn't lie. Nvidia's CDS price rose from about 42 basis points at the end of June to 57 basis points in mid-July, then hit a historic high of 83.7 basis points at the end of July. After the $500 billion "good news," CDS rose another 5.3 basis points. The market is betting real money on Nvidia's credit risk rising. It's like someone announces they will help a friend borrow $500 billion, but their own credit card limit gets cut by the bank. Doubt two: "Circular financing" is not baseless. At the end of July, Nvidia just announced over $500 billion cooperation with SK Group. It is also discussing with OpenAI, possibly providing up to $250 billion in guarantees. Left hand finances SK, right hand guarantees OpenAI, with Nvidia chips in the middle. Critics' core concern: companies Nvidia invests in or holds shares in are usually the main buyers of its chips. Sell chips to you → lend you money to buy → you buy my chips → my revenue grows → I lend money to others to buy. Doubt three: Chanos' "Congressional hearing" prophecy. "Big short" Jim Chanos sarcastically posted: "If these people sit together next time to explain AI financing, it might be at a Congressional hearing in 2031..." Chanos has long warned: current AI data centers and similar projects generally package returns with low capitalization rates of 5%-7% plus high leverage; once interest rates rise, this business model built on cheap capital will collapse instantly. He also said something more cutting: "If you buy chips from Nvidia, rent someone else's data center, then sublease computing power to Microsoft or Google, you're an equipment leasing company, not a tech company, but a financial company." To be honest— Jensen Huang's calculation is actually very clever. He brought in private equity to provide bridge loans to second-tier manufacturers with tight cash flow, betting that AI applications will explode in 2-3 years and they can break even. At the same time, he nurtures a batch of "loyal troops"—these new elites supported by private equity lack chip R&D capabilities and must cling to Nvidia forever. Even if Microsoft or Google rebel, Huang still has people long-term using his cards. But what is the biggest risk? If AI applications are still unprofitable in 2-3 years? Then this $500 billion is not bridge funding but a "debt black hole." When that happens, the fall won't be just Nvidia alone. Jensen Huang says: "The demand is real." Chanos says: "This is a replay of 2008." But one thing is certain— CDS doesn't lie. When the market votes with its feet, the voice is loudest. Nvidia is tied to half of Wall Street's lifeblood, tied to U.S. tech hegemony, tied to global pensions. When the fire burns through everything, who will bail it out? Isn't this the same script as Fannie Mae back then— Win big and make huge profits, lose and everyone pays the price. $NVDA $META $SNDK #英伟达推动5000亿美元AI基建融资