In today's era, marked by the explosive growth of generative AI, computing power has emerged as a highly valuable resource, surpassing even oil in its significance. Yet, with the initial cost of a cluster of H100 GPUs often soaring into the hundreds of millions, the vast majority of entrepreneurs find themselves priced out. As the industry grapples with the dual challenges of 'computing power scarcity' and 'capital constraints,' NVIDIA steps forward—not merely as a conventional hardware vendor, but as a trailblazing financial powerhouse. It introduces a groundbreaking model known as 'computing power loans,' which facilitates the exchange of GPU usage rights for a share of revenue, fundamentally transforming the financial landscape of the AI sector. Leading the charge in embracing this transformative approach are a new breed of enterprises known as 'Token factories.'

01 From 'Selling Shovels' to 'Establishing a Bank'

For an extended period, NVIDIA has been perceived as a 'shovel seller,' supplying high-demand GPUs to cloud service providers and tech behemoths. However, as the competition in developing large-scale models intensifies, the delivery time for chips like the H100 and B200 has stretched beyond six months, with prices doubling in the secondary market. For AI companies eager to commercialize their models, the upfront cost of acquiring computing power represents not only a substantial financial burden but also a formidable barrier to entry.  

Traditional financing avenues—equity financing or bank loans—prove highly inefficient when dealing with asset-light, high-risk AI startups. Consequently, NVIDIA has drawn inspiration from the SaaS industry's well-established 'revenue-sharing financing' model, introducing 'computing power loans.' The core logic is compelling: rather than allowing GPUs to sit idle in the distribution channel awaiting buyers, NVIDIA proactively treats them as credit assets, directly deploying them to the most promising AI application endpoints.

This transition signifies a profound transformation in NVIDIA's role. It transcends its identity as a mere hardware supplier, evolving into a 'computing power bank' that not only controls core production resources but also possesses risk assessment and capital recovery capabilities. This strategic shift elevates NVIDIA's business model from one-time transactions to leveraging long-term cash flow across the entire AI ecosystem, with its valuation now anchored in the total lifecycle value of the ecosystem rather than hardware shipments alone.

02 The Cash Flow Enchantment of Token Factories

The operational model of 'computing power loans' operates like a cash flow enchantment for 'Token factory'-style enterprises. These companies specialize in providing large-model inference API services, charging customers based on the volume of tokens generated. Their business model is straightforward: input computing power, output tokens, and profit from the margin between token consumption and computing power costs.

Under the new computing power loan scheme, a Token factory is exempt from pledging any fixed assets. Instead, it simply needs to demonstrate to NVIDIA or its designated financial partners its API call volume, customer contracts, and projected growth trajectory. Upon approval, the factory gains immediate access to a substantial GPU cluster computing power quota. In return, the factory commits to regularly paying NVIDIA a predetermined percentage (e.g., 5% to 15%) of its token sales revenue over a specified period, until the cumulative payout reaches a preset cap (e.g., 1.3 to 1.8 times the initial hardware value).

This design ingeniously resolves the timing mismatch issue. The 'Drainage period' (user acquisition phase), during which startups consume the most computing power, coincides with their most cash-strapped period. The revenue-sharing model directly converts their largest fixed cost into a variable cost that fluctuates with revenue. When the factory's business booms and token generation per second soars, its shared revenue increases proportionally; conversely, if business slows, its cost pressure naturally diminishes. This flexibility significantly enhances the survival rate and trial-and-error space for AI startups, transforming heavy computing power liabilities into elastic 'computing power subscriptions + profit-sharing.'

In fact, the trend towards the financialization of computing power has been quietly building momentum. NVIDIA-backed GPU cloud service provider CoreWeave pioneered the use of its stockpiled H100 chips as collateral to secure hundreds of millions of dollars in debt financing from Wall Street. This demonstrated that top-tier GPUs possess creditworthiness comparable to real estate. 'Computing power loans' take this concept a step further, deepening the credit basis from 'owning GPU ownership' to 'possessing the ability to generate cash flow with GPUs.'

This directly reshapes the AI industry's valuation system. Token factories are no longer just tech companies; they have evolved into financial entities with predictable, auditable cash flows. Their core assets, in addition to their algorithm teams, now include stable token production capabilities secured by NVIDIA's computing power loans. When future investment firms evaluate such companies, they may place less emphasis on user numbers and focus more on 'computing power sharing coverage' and 'token gross margin.' A financial revolution centered on computing power credit has already commenced, with computing power becoming true capital and token output data serving as the most reliable credit voucher.

04 NVIDIA's Vision to Become a Computing Power Bank

NVIDIA's pursuit of 'computing power loans' is part of a broader ecological strategy. Through financial means, it ensures that its GPUs are directed towards the most growth-oriented application layers, optimizing the utilization of its high-end chip capacity while deeply binding the next generation of potential unicorns. Under this model, even if future competitors match its hardware performance, they will find it challenging to dislodge customer loyalty built on financial contracts and long-term sharing agreements.

However, this model also carries significant risks. By assuming the role of a bank, NVIDIA exposes itself to credit default risks. If the AI application bubble bursts and many Token factories underperform in revenue, default risks could accumulate and negatively impact NVIDIA's financial stability. More profoundly, this ultra-low-barrier computing power supply may further intensify market dependence on NVIDIA's ecosystem, creating a closed system absolutely dominated by it, with startups' fates deeply intertwined with its own.   

The emergence of 'computing power loans' signifies the AI industry's official entry into a new era of deep financial capital integration. Computing power, the lifeblood of the digital world, is undergoing a paradigm shift from a commodity to a capital asset. For new-generation players like Token factories, computing power loans represent a bold gamble to trade future potential for present opportunities and an excellent springboard to leverage limited resources for substantial gains. What Jensen Huang is offering is no longer just chips but the financial operating system of the AI era.