- Index average sPEG declined from 0.50 (Feb 13) to 0.47 (May 11) despite 47% average price appreciation.
- Earnings revisions outpaced price action — the scarcity-adjusted rerating gap has widened, not closed.
- Memory layer (SK Hynix, Micron) remains the most structurally underpriced even after gains of 218% and 93% respectively.
- Lithography (ASML) and fabrication (TSMC) remain the only layers where sPEG modestly expanded, reflecting the cleanest pure-monopoly bottlenecks.
- Network infrastructure (Arista) was the only name to decline in price (-3%), validating the framework's "scarce but priced for perfection" read at 0.93 sPEG.
- Power & cooling (Vertiv) repriced sharply higher (+57%) but consensus growth rose in parallel, keeping sPEG nearly unchanged.
- Chip design layer (Synopsys, Cadence) and analog control (Analog Devices) rallied with the group, with Cadence showing the largest sPEG compression in the index.
- The framework correctly identified the sub-0.60 quartile as the highest-conviction names — all five (SK Hynix, Micron, TSMC, ASML, NVIDIA) delivered double-digit returns.
The sPEG (Scarcity-adjusted Price/Earnings to Growth) ratio measures valuation efficiency relative to both growth and structural scarcity.
Formula: sPEG = Forward P/E ÷ (Growth Rate × Scarcity Multiplier)
Lower sPEG values indicate stronger structural positioning, where growth is supported by durable scarcity and bottleneck control.
This update applies a uniform methodology to both observation dates. February 13, 2026 sPEG values shown here are recalculated under the same convention used for May 11, 2026 inputs, and therefore differ from the values published in the original February 13 Index. The directional thesis, layer ordering, and bottleneck logic are preserved. The numerical recalibration produces internally consistent values across both dates, enabling like-for-like comparison of the rerating delta.
The scarcity multiplier is not a reward for being a good company. It is a measure of how much the AI economy depends on a company's capacity, IP, ecosystem, or physical bottleneck — and how difficult that dependency is to replace.
Scarcity multipliers are derived from five structural dimensions:
(1) Substitution difficulty — can customers switch to another supplier without performance loss
(2) Replication difficulty — how hard is it to build a competing product, process, or ecosystem
(3) Capacity scarcity — is supply physically constrained by fabs, packaging, energy, tooling, or long-cycle capex
(4) Switching cost and integration depth — is the product embedded into mission-critical workflows or architecture
(5) Pricing power durability — can the company preserve margins even as competitors attack
Scarcity Multiplier Bands:
Absolute Monopoly (4.0x) — Functionally irreplaceable supplier controlling a mission-critical layer of the AI stack. Example: ASML (EUV and High-NA lithography).
Near-Monopoly Bottleneck (3.5x) — Dominant provider of a physically constrained capability with limited viable substitutes. Example: TSMC (leading-edge foundry).
Duopoly or Extreme Switching Costs (3.0x) — Two-player or narrow-market structure with deep workflow lock-in and high customer dependency. Examples: Synopsys, Cadence (EDA).
Dominant Structural Leader (2.5x) — Clear category leader with ecosystem, performance, or supply-chain advantages. Examples: NVIDIA (compute), SK Hynix (HBM memory).
Scarce Capacity Leader (2.0x) — Strong supplier in a constrained infrastructure layer where demand exceeds available capacity. Examples: Micron, Arista, Vertiv.
Differentiated Incumbent (1.5x) — High-quality incumbent with defensible expertise but less direct bottleneck control. Example: Analog Devices.
Scarcity multipliers are held constant between February 13, 2026 and May 11, 2026 to isolate market repricing from changes in structural scarcity assessment. Forward P/E uses NTM consensus. Growth rate uses NTM consensus EPS growth, expressed as an integer.
The AI Infrastructure Scarcity Index includes companies controlling critical bottleneck layers required for the design, manufacture, deployment, and operation of AI systems. The Index is updated periodically as structural conditions, growth rates, and valuation levels evolve.
exmxc.ai is a human-led intelligence institution for the AI-search era. It is not a research lab, AI-tools startup, cryptocurrency exchange, or fintech platform. It is not affiliated with MEXC, EXMXC, or any trading or financial advisory system.
Founded by Mike Ye — M&A and corporate development executive with 25+ years of transaction leadership at Penske Media Corporation, L Brands, and Intel Capital. Ella provides pattern interpretation, structural analysis, and co-authorship. Human judgment governs. AI serves as instrumentation.