To The Moon

On August 10, 2026, a significant event unfolded on Wall Street and in Silicon Valley. Nvidia officially announced that it has signed a memorandum of understanding with six of Wall Street's top-tier private capital and financial giants—Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR. The agreement establishes an independent computing financing platform aimed at unlocking over $500 billion (equivalent to more than 3.6 trillion RMB) in third-party capital for global AI infrastructure construction.

This move comes at a time when there are signs of a turning point in global AI cloud computing capital expenditures, and the computing sector is experiencing massive volatility. Nvidia's initiative is clearly intended to inject a calming effect into the semiconductor market. However, upon the announcement, Nvidia's stock price experienced a "flash crash" during trading, ultimately closing down over 2%. Wall Street opinions quickly split into two sharply contrasting camps.

Two Factions at Odds

The two viewpoints are as follows: One faction praises Jensen Huang for completing the ultimate transformation from a chip vendor to the chief financial architect of AI infrastructure, packaging computing power into investable asset packages akin to commercial real estate and toll roads. The other faction issues a cold warning, questioning whether this is a reincarnation of the "vendor financing" and "circular lending" seen during the dot-com bubble, now in the AI era.

As an investor, before rushing to take sides, it is crucial to understand why Jensen Huang is doing this. It stems from an extremely harsh industrial reality: the AI computing power race has completely shifted from a bottleneck in chip production capacity to a physical bottleneck in capital scale and cash flow.

By 2026, the evolution of large AI models has not stopped due to diminishing marginal effects. On the contrary, as frontier laboratories push towards 10-gigawatt-scale supercomputing clusters, the physical cost of building a modern AI factory is skyrocketing exponentially. Constructing a data center equipped with tens of thousands of Blackwell or Rubin architecture chips, along with supporting utilities, power transformers, and liquid cooling facilities, now requires sunk costs of hundreds of billions of dollars. Even tech giants as strong as Microsoft, Meta, Google, and Amazon all have capital expenditures exceeding $100 billion in 2026. For OpenAI, Anthropic, various mid-sized AI cloud providers, and national sovereign AI projects, the shortage is no longer the desire to buy chips, but the cash on hand to pay for them.

Jensen Huang is certainly aware of this. A more pragmatic calculation is that if Nvidia relies solely on its customers' balance sheets to pay for GPUs, its performance growth will inevitably be constrained by the limited cash flow of those enterprises. To break this funding bottleneck, Nvidia has made a highly disruptive business upgrade. It does not burn its own balance sheet but instead builds a platform, directly linking Wall Street's trillions of dollars in private credit, insurance funds, and pension capital—all seeking returns—with its own computing ecosystem.

In this model, known as the computing financial flywheel, the six Wall Street giants provide the capital to establish an independent financing platform, offering low-cost, long-term debt financing. The computing demand side (AI labs, cloud providers) receives loans to purchase Nvidia GPUs and infrastructure. Once the data center is built and online, customers generate cash flow by selling computing services and large model APIs to repay Wall Street's debt. Nvidia contributes some ecosystem support, locking in forward orders and cash prematurely.

Jensen Huang summarized the move with a profound statement: "We started out making chips; today, we are creating a new, investable asset class—the AI factory." His message is that he is no longer just a seller of picks and shovels; he has packaged the entire water channel, reservoir, and even the financial tools for discounting future water bills into one offering.

Opportunities and Risks

Any massive capital operation, while amplifying returns, inevitably comes with hidden risks. Analyzing this $500 billion financing deal reveals two distinct poles.

On the opportunity side: First, the securitization of AI infrastructure. The most significant aspect of this deal is that Wall Street's highest-level think tanks—Blackstone, BlackRock, and Apollo—have officially labeled computing infrastructure as an investment-grade asset. This means computing power is transitioning from a short-term consumable to a long-term income-generating asset similar to telecom towers, high-voltage power grids, and water conservancy projects. Second, the strengthening of Nvidia's moat. For Nvidia, this is not just about selling chips; it locks in customers through the capital chain. When an AI startup or cloud service provider borrows low-interest loans from this joint financing platform, they are bound to the full CUDA ecosystem and Nvidia hardware architecture. AMD or other ASIC chipmakers seeking to gain market share will face not only a technical barrier but also a customer's question: "Others come with $500 billion in subsidized financing. What do you offer?" Third, a resurgence in the AI, power, and real estate chain. The $500 billion will not be spent entirely on chips; part will go to building factories, requiring transformers, cooling towers, green power stations (nuclear/solar), and industrial real estate. Companies with access to high-voltage grids, liquid cooling leaders, and energy giants will benefit from this wave of construction funded by Wall Street credit.

On the risk side: First, the shadow of "circular financing." This is the reason some savvy short-sellers chose to sell after the announcement. Leading up to the dot-com bubble burst in 2000, telecom equipment giants like Cisco and Lucent lent heavily to startup telecom companies to buy their own routers. The risk is that if downstream applications fail and startups cannot generate sufficient cash flow from computing power, this debt chain will default, triggering a cascading collapse. Therefore, lending money to customers to buy your own products is a double-edged sword. Second, the balance between free cash flow and capital expenditure. Although this $500 billion is third-party capital, it raises the overall leverage ratio of the tech industry. If the revenue growth rate from AI applications cannot keep pace with interest expenses and computing depreciation, Wall Street's anxiety over tech giants burning cash will resurface, causing a rapid contraction in the valuation of the entire tech sector. Third, concentration risk. Wall Street's top-tier private credit institutions are placing a significant portion of their bets on a single ecosystem—Nvidia. If a major technological shift occurs, other chip manufacturers or in-house chips from big tech succeed, or a core frontier lab like OpenAI suffers a commercial setback, this financial alliance could turn into a systemic asset restructuring storm.

Market Skepticism

Returning to the capital markets, the reaction to risk seems more pronounced. The Philadelphia Semiconductor Index fell 2.94%. The reason is simple: the capital expenditures of cloud computing providers—especially the five major ones: Microsoft, Google, Meta, Amazon, and Oracle—amounting to over $700 billion this year, have propped up the extraordinary profits of semiconductor companies like Nvidia, SK Hynix, and Samsung Electronics. However, in last month's earnings reports, Microsoft reduced its capital expenditure (though partly an accounting adjustment), Meta only raised the lower bound of its capex guidance, and while Google and Amazon increased theirs, this represents a pause or even a retreat compared to the previous surge. If this year marks the peak of capital expenditure for these giants, then the peak of profit growth for semiconductor companies will also occur this year.

The core issue is that this year's base is exceptionally large, and valuations have been driven up accordingly. While many analysts suggest the semiconductor sector has been largely cleared out and valuations are low, significant divergence remains in the capital markets. The bearish view is clear: the current low valuations are anchored to a base of over $700 billion in annual capital expenditure. The problem is that these expenditures are being widely questioned, and the stock prices of cloud computing providers have been languishing for over a year as a result. In business, if this money cannot be converted into actual performance, the capital market will vote with its feet. A compelling story can attract capital, but investor patience is finite, as fund managers have monthly, quarterly, and annual KPIs. Even long-term capital, such as BlackRock, Vanguard, sovereign wealth funds, and national pension funds, have liabilities to match, actuarial return assumptions, Sharpe ratios, and maximum drawdown limits to consider. They cannot wait indefinitely.

So, upon deeper analysis, you find that the coefficients in valuation models are not fixed, and there are many assumptions. If the annual capital expenditure of cloud providers decreases, the orders for AI semiconductor companies will shrink, causing EPS to decline. Would valuations still appear low then? The colder reality lies in market sentiment. After significant rallies and crashes, which have captured so many investors, can their emotions recover quickly? A more immediate question is: do they still have the capital to buy the dip? These questions remain unanswered. The answers to these questions will determine whether AI semiconductors can rebound and how high that rebound might be.

Concluding Thoughts

During the downturn in the AI semiconductor sector, Nvidia's move to lead capital giants in providing this solution can be seen as a timely rain. It might be the simplest and most feasible plan, faster than IPOs, secondary offerings, or bond issuance. In the long term, this plan has its merits and can even be considered quite forward-thinking, as no one can deny the importance of computing power for the AI industry. Theoretically, once the peak of capital expenditure is passed, the subsequent phase is akin to a utility business—collecting steady, predictable, and sustainable cash flows, which is precisely what Wall Street giants desire. On the surface, these giants are staffed by smart people who have likely put deep thought, countless calculations, and simulations into committing such a large sum. Otherwise, it would become an epic Wall Street joke.

But the problem lies precisely here. The capital market can interpret this as "fresh capital has arrived, there is hope," or as "this confirms the cash crunch at those big companies." In the short term, this divergence will persist. In the long term, the value of computing power is difficult to falsify. This brings us back to the well-worn logic: short-term pressure, long-term value. Given this, the question of value no longer needs debate; the key is about timing. After all, on the investment journey, no one wants to become a martyr.