Everyone Is Waiting for the AI Bubble to Pop (NVIDIA Earnings)
Speakers
Summary
- NVIDIA reported $68.1B in Q4 revenue (up 70% YoY) and guided $78B for Q1, beating Wall Street estimates of $72B.
- The "AI Bubble" thesis is being challenged by physical constraints: the grid cannot supply enough electricity ("watts") nor can foundries supply enough chips ("wafers") to create an oversupply glut yet.
- Anthropic's new coding agents are disrupting legacy moats, specifically targeting IBM's dominance in COBOL maintenance, allegedly wiping out significant market cap.
- Meta has diversified its hardware supply chain by signing a $6B partnership with AMD, signaling strong demand for second-tier GPU providers.
Trade Ideas
NVIDIA beat every target (EPS and Revenue), guiding $78B for Q1 vs $72B estimated. Demand for even 5-year-old GPUs is so high they rent for 1.5x their original cost. The "Bubble" argument relies on overbuilding and underutilization. However, hyperscalers are fully utilizing every chip they buy. Furthermore, the physical scarcity of energy prevents companies from overbuilding fast enough to crash the market. LONG. The "watts and wafers" constraint extends the cycle duration, keeping pricing power high. Regulatory intervention or a sudden, unexpected solve for energy constraints leading to rapid oversupply.
Anthropic launched a plugin that autonomously fixes COBOL code (a legacy business language). The speaker notes IBM stock reacted negatively (claiming a 15% drop) because they hold a monopoly on COBOL consulting. IBM's high-margin consulting business relies on the complexity and scarcity of human COBOL experts. If an AI agent can update and fix this dinosaur code for the cost of a prompt, IBM's moat evaporates. SHORT. AI is deflationary for legacy IT services and manual code maintenance. IBM successfully pivoting to integrate these tools faster than clients can adopt them directly.
Meta signed a $6B compute partnership with AMD, causing the stock to jump significantly. Hyperscalers (like Meta) are desperate to diversify away from total reliance on NVIDIA. This "circular economy" ensures that second-place hardware providers receive massive spillover capital simply to ensure supply chain redundancy. LONG. AMD is the primary beneficiary of the "Anyone But NVIDIA" diversification spend. NVIDIA solving supply constraints rapidly, rendering inferior chips less desirable.
Josh Kale
Co-Host, Limitless Podcast (Bankless)
"We're constrained by watts and wafers, meaning we don't have enough electricity... and as a result, that becomes the limiting factor." If the AI supercycle is bottlenecked by the electrical grid rather than software demand, the value accrues to power producers (VST/CEG) and grid infrastructure providers (ETN). They hold the scarce resource required for AI expansion. LONG. These are the "pick and shovel" plays for the energy crisis described in the transcript. Regulatory caps on power pricing or hyperscalers building off-grid nuclear solutions that bypass public utilities.
