Summary of Key Points
The recent decline in Broadcom's stock price is not due to performance issues, but rather concerns among investors about the potential credit risks associated with the "seller guarantee" model of its AI financing platform. This model represents a mainstream financial innovation in the AI sector: it involves securitizing computing power assets, allowing AI companies, chip suppliers, and financial institutions to collaborate and divide the workload. While this approach has contributed to the development of computing power, it also carries risks similar to those associated with the subprime mortgage crisis (CDOs). However, there are fundamental differences between the two. The market needs to acknowledge these risks rather than deny the value of the innovation.
1. Broadcom's Stock Price Falls by 5.9%: High Guarantee Ratio Raises Concerns
Broadcom, in collaboration with Blackstone and Apollo, has established an AI XPV financing platform that essentially helps AI companies (such as Anthropic) purchase computing power. The computing power purchased by these companies is placed in a special "special purpose vehicle" (SPV), for which Broadcom guarantees 85% of the bonds issued. This means that if the AI company is unable to pay the rent, Broadcom would have to bear 85% of the loss.
Why are investors concerned? The 85% guarantee ratio is considered too high. If the AI company's business fails and it cannot afford the rent, Broadcom would have to foot the financial shortfall, which could directly impact its own profits. As a result, market concerns led to a decline in Broadcom's stock price.
2. A Popular Financing Model in the AI Sector: Everyone Benefits
This model is not unique to Broadcom; it is a common practice in the AI industry where all parties can gain benefits:
- AI Companies (e.g., Anthropic): They do not need to invest heavily in purchasing computing equipment and can simply pay rent monthly, effectively operating with a low-cost, high-capacity infrastructure.
- Chip Suppliers (Broadcom, Nvidia): They can not only sell their chips consistently but also earn additional profits by participating in the revenue sharing of the computing centers through the guarantee.
- Financial Institutions (Blackstone, Apollo): The guarantee provided by Broadcom reduces their risk; they can earn stable bond interest while also benefiting from any additional profits generated by the AI companies' growth (e.g., higher rents).
- Operators (Fluidstack): They manage the computing clusters and charge service fees.
The essence of this model is to securitize computing power assets, allowing different stakeholders to leverage their respective strengths to create a more efficient and profitable ecosystem.
3. Why Is It Compared to Subprime CDOs?
Some compare this model to the CDOs from the 2008 subprime mortgage crisis, and there are indeed similarities, but key differences exist:
- Similarities: Both rely on the continuous growth of a certain asset (housing prices for CDOs and computing power demand for this model). If demand drops suddenly, it could lead to default risks.
- Differences:
1. Source of Cash Flow: CDOs depend on rising housing prices (without actual operating income), while this model relies on rent payments from AI companies (real operating income, received monthly), making it more reliable.
2. Risk Capacity: During the subprime crisis, insurance companies' guarantee amounts far exceeded their capital, leading to collapses; in contrast, Broadcom's guarantees are backed by its own capital, providing greater risk resilience.
4. What Are the Risks of This Model?
Although safer than CDOs, this model still poses several risks:
- Credit Risk: If AI company demand falls short of expectations and they cannot afford the rent, Broadcom's guarantee could result in actual losses.
- Collaboration Risks: Issues with the cooperation among parties (e.g., poor management by operators or hardware failures) could affect rental income.
- Liquidity Risk: If the market loses confidence in AI, no one may be willing to buy these asset packages, making it difficult to liquidate them.
- Excessive Guarantee Exposure: Broadcom's 85% guarantee ratio is quite high, and any problems could have a significant impact on the company.
5. Facing Risks: Don't Abandon Innovation Due to Concerns; Manage Leverage
This model is a crucial driver of the AI computing power revolution, and we should not give up on it due to potential risks. The key is to:
- Control the Guarantee Ratio: Broadcom could reduce this ratio to spread the risk among more parties.
- Ensure Stable Cash Flow: Rent contracts with AI companies should be carefully crafted to prevent defaults.
- Regulatory Oversight: Implement measures to prevent excessive leverage (similar to what happened during the subprime crisis) and keep risks within manageable limits.
In summary, this financial innovation is an essential part of AI development, with both benefits and challenges. Investors' concerns are understandable, but there's no need for panic. By properly managing leverage and risks, this model can continue to drive the advancement of AI technology.