The author also raises questions about the role of banks in this trend, such as whether they are financing the purchase of GPUs, buying GPUs for themselves, or working directly with NVIDIA or other tech companies. The article concludes by advising readers to watch for the growth of specialty "GPU debt funds" and the potential risks if AI progress starts to outpace Moore’s Law, reducing hardware life cycles.
Key takeaways:
- GPUs, or Graphics Processing Units, are being used as collateral for large loans due to their high demand and short supply. Companies like CoreWeave are securing billion-dollar debt facilities using GPUs as collateral.
- There is a growing trend of financing the purchase of GPUs, especially among AI startups and BigTech cloud providers. These companies are using GPUs to train their own AI models or rent them out to other developers and clients.
- The high cost and demand for GPUs are creating a capital-intensive race. However, the author warns that the size of these financings and the potential for supply shortages to level off or advancements in AI to shorten the lifespan of GPUs could pose risks.
- Banks are beginning to participate in this trend, either by financing the purchase of GPUs or by partnering with NVIDIA or other cloud providers. The author suggests watching for which banks launch B2B businesses leveraging their new AI compute capabilities and whether specialty "GPU debt funds" grow significantly.