NVIDIA's AI Bubble: Debt, Depreciation, and Unused GPUs | Better Offline

The AI Infrastructure Spending Spree

• 🚀 Big tech companies like Microsoft, Amazon, Google, and Meta are projected to spend over $776 billion on AI infrastructure between 2023 and 2025.
• 💰 An estimated $2 trillion in new AI revenue is needed by 2030 to justify this massive capital expenditure.

Depreciation and Accounting Concerns

• 💡 GPUs depreciate over their useful life, a cost spread over time through accounting, which can make current profits appear healthier.
• ⚠️ Companies are depreciating GPUs over 5-6 years, despite their likely useful lifespan being only 1-3 years due to rapid obsolescence and wear, creating an "accounting puzzle."
• 📈 Spreading costs over a longer period allows hyperscalers to reduce their net income and present a better financial picture to the market.

The Mystery of Unused GPUs

• ❓ Despite Nvidia's claims of shipping millions of Blackwell GPUs, there is little evidence they are in active service, suggesting they may be warehoused.
• 🔌 A significant challenge is the lack of sufficient power and data center capacity to operate these newly acquired GPUs, with buildings lacking adequate power infrastructure.
• 📉 Nvidia's success appears dependent on customers' credit ratings and financial backers, as many AI companies and hyperscalers are relying heavily on debt and venture capital.

Profitability and the AI Ecosystem

• 📉 Leading AI companies like OpenAI and Anthropic are burning billions of dollars and are heavily subsidized by venture capital and debt, not generating significant profits.
• 📊 Even the champions of the AI boom are struggling, with revenues not covering the immense costs of compute and infrastructure.
• 💸 The entire AI ecosystem, from startups to hyperscalers, appears to be propped up by debt and venture capital, with Nvidia being the primary beneficiary.

The Unsustainable AI Growth Model

• ⚠️ The current model relies on a continuous flow of venture capital and debt, which is unsustainable and risks breaking when these funding sources dry up.
• 📉 Nvidia's revenue growth is tied to selling assets that often result in losses for its customers, raising questions about the long-term viability of this model.
• ❓ The lack of transparency regarding GPU usage and the actual revenue generated from AI services makes it difficult to justify the continued massive spending and pre-ordering of expensive hardware.