Summary of Key Points
The core message of this news is that the AI industry is transitioning from a phase where there is a shortage of GPU chips to one where there is a lack of capital for large-scale asset expansion. NVIDIA, in collaboration with global financial institutions, has launched an AI computing power financing platform aimed at leveraging $500 billion in third-party capital. The essence of this initiative is to package GPUs, data centers, electricity, and long-term computing power contracts into financiable assets, making computing power accessible to more customers who cannot afford large-scale projects on their own. This marks the entry of AI into a “financialized” phase, where computing power is no longer just a technical issue but has become an infrastructure that requires capital support. This transition also brings new challenges, such as rapid technological obsolescence, unstable demand, and risk dissemination.
Detailed Analysis
1. The Financial Challenges in the AI Industry: From a Shortage of GPUs to a Lack of Funds
Over the past three years, AI companies have faced the problem of not being able to obtain enough GPUs; now, the challenge has shifted to not having sufficient funds to purchase them. Why? Because AI projects have become much larger. What used to require just a few servers now requires the construction of entire data centers, costing billions of dollars (including land, electricity, cooling systems, and GPUs). Many model companies and small-to-medium-sized cloud service providers do not have the financial resources to fund such large-scale projects. For example, NVIDIA’s data center business revenue has increased by 92% this year, indicating ongoing demand, but customers no longer have the funds necessary. Therefore, the constraint on AI expansion is no longer chip production capacity but rather the ability to raise funding.
2. NVIDIA’s $50 Billion is Not from Its Own Pocket: It Acts as a Capital Matchmaker
NVIDIA claims it will mobilize $50 billion in capital, but this money does not come from its own resources; instead, it is third-party capital. Its role is to organize these funds by packaging AI computing power projects into assets that financial institutions can invest in. For instance, Goldman Sachs is negotiating with U.S. financial institutions to divide the projects into different risk levels of products (such as “senior” and “junior” debt) for investors with varying risk tolerances. NVIDIA itself will contribute only a small portion of the funds (e.g., up to $2.3 billion in potential losses), using its technical credibility and industry connections to attract more external capital—essentially helping customers secure funding while selling more GPUs.
3. Why Do Wall Street Institutions Invest in Computing Power Assets?
Financial institutions are not naive; what drives their investment in AI computing power is the potential for long-term cash flows. For example, power plants can be financed because they have long-term electricity purchase agreements that ensure stable revenue, and airplanes can be financed because they can still be rented out even if the airline goes bankrupt. For AI computing power to be considered viable as an investment, three conditions must be met:
- Long-term demand: AI will not disappear suddenly, so there is a continuous need for computing power.
- Transferability of assets: If one customer defaults, the GPUs and data centers can be rented out to another.
- Cash flow to cover debts: Long-term contracts ensure that projects generate enough revenue to repay loans.
Financial institutions are interested in the stable income stream generated by the combination of GPUs, data centers, electricity, and long-term contracts—this is more akin to financing a power plant construction project rather than a routine computer purchase.
4. Where Do the Risks Lie?
This model carries three major risks:
- Rapid technological obsolescence: GPUs are updated frequently (new chips often perform significantly better than older ones), and if financial models miscalculate the residual value of old GPUs, it could lead to loan losses.
- Unstable demand: AI projects are typically planned for a 10-year cycle, but customer needs can change within months. If an AI company’s commercialization efforts fail, the long-term contracts may not be renewed, resulting in financial losses.
- Risk dissemination: In the past, AI companies relied on venture capital (VC) for funding; if these projects failed, the risks were borne by the VC firms. Now, by turning to banks and insurance companies, the risks are spread throughout the financial system. For example, a decline in AI demand could lead to low data center utilization rates and unpaid loans, affecting these financial institutions.
5. NVIDIA’s Strategy: Financing Ability = Sales Ability; Capital Cost Becomes a New Battleground
For NVIDIA, helping customers with financing is not about charity but about selling more GPUs. It’s similar to how aircraft companies help airlines obtain loans—only when customers have the funds can sales increase. NVIDIA’s current position is that of a “technology + capital organizer.” By leveraging its CUDA ecosystem (software platform) to improve the versatility of GPUs (e.g., allowing old GPUs to be repurposed), it increases the value of these assets. Additionally, by providing financing through its platform, it reduces the cost of financing for customers and expands its customer base.
The new competitive factor in the AI industry will be “capital cost.” Companies that can obtain funds at lower interest rates and for longer terms will be able to build more data centers and operate them at lower costs, gaining a competitive advantage. For instance, the competition between OpenAI and CoreWeave involves not only the quality of their models but also who can acquire GPUs, electricity, and capital more efficiently.
Final Conclusion
The financialization of GPUs signifies that AI has evolved from a technological revolution in laboratories to an infrastructure revolution that requires capital support. While this can accelerate the widespread adoption of AI, it also spreads risks from technology companies to the entire financial system. The future will depend on how capital is priced and how risks are managed.
(End of translation)