Investment theme
AI Compute Build-out
While AI workloads grow faster than the efficiency of the hardware that runs them, demand for accelerators, memory and the capacity to make them keeps running ahead of supply.
- Evidence
- Losing support
- Cycles
- Pulling both ways
- Valuation
- Richer than its own past
- Companies
- 9 tracked · 2 potential opportunities
Companies riding it
9 tracked| Company | Zone | Valuation vs own history | Warnings | Themes |
|---|---|---|---|---|
| VRT Vertiv Holdings LLC Potential opportunity | Discount | 12th percentile | None | 2 |
| NVDA NVIDIA Corporation Potential opportunity | Discount | 15th percentile | None | 1 |
| MU Micron Technology Inc. | Reference | 39th percentile | None | 1 |
| AVGO Broadcom Inc. | Above reference | 64th percentile | None | 1 |
| AMD Advanced Micro Devices Inc. | Above reference | 66th percentile | None | 1 |
| LRCX Lam Research Corporation | Above reference | 72nd percentile | None | 1 |
| KLAC KLA Corporation | Above reference | 90th percentile | None | 1 |
| AMAT Applied Materials Inc. | Above reference | 92nd percentile | None | 1 |
| MRVL Marvell Technology Inc. | Above reference | 96th percentile | None | 1 |
Also in the theme, not tracked here: TSM, ANET, ASML.
Zones and valuation are measurements against each company’s own history, not buy or sell instructions. See them on the stocks list →
Narratives that test it
Losing support-
Advanced memory is becoming an AI supply bottleneck
Supports the theme if it holds · Too little evidence yet
The timeline for relief from advanced memory constraints is extending further into the future. Industry executives now project that RAM shortages will persist through 2028, pushing memory prices for 2027 significantly higher than current levels. This prolonged supply-demand imbalance reinforces the status of high-bandwidth memory as a durable bottleneck in the AI hardware supply chain.
Latest development: Industry expectations for the duration of the memory shortage have extended, with executives now forecasting constraints to last through 2028 and driving 2027 prices significantly higher.
-
AI spending is outrunning its returns
Challenges the theme if it holds · Recent evidence challenges the theme
Evidence increasingly supports the claim as infrastructure outlays impair cash flow. Meta's shift into negative free cash flow and record credit default swap prices, alongside funding limits at specialized clouds, shows strain reaching major operators and compounding earlier debt stress at CoreWeave. Counterevidence includes strong revenue gains at model makers like Anthropic. The critical question is whether enterprise monetization can accelerate quickly enough to justify sustained capital outlays.
Latest development: On October 22, 2026, Meta's shares dropped nearly 7% as capital expenditure on AI pushed its free cash flow into negative territory, sending credit default swaps to record levels. On the same date, Nvidia chief executive Jensen Huang stated that capital scarcity is holding back growth for specialized cloud providers.
-
Export controls are fragmenting the technology supply chain
Challenges the theme if it holds · Evidence on both sides
The enforcement of US export controls remains a complex challenge, as evidenced by the ongoing illicit flow of restricted hardware. The arrest of a tech CEO accused of smuggling $300 million worth of Nvidia chips into China highlights the massive financial incentives driving black-market supply chains. This underscores the difficulty of fully isolating the Chinese AI ecosystem through trade restrictions alone.
Latest development: A tech CEO was arrested for allegedly smuggling $300 million in Nvidia chips into China, demonstrating the scale of black-market efforts to bypass US export controls.