sPEG vs. Traditional PEG: Valuing the AI Infrastructure Bottleneck

sPEG vs. Traditional PEG: Valuing the AI Infrastructure Bottleneck

Traditional valuation metrics such as PEG assume growth is the primary driver of enterprise value. That assumption holds in demand-driven markets and breaks in constraint-bound ones, where supply scarcity — not demand — determines who captures the economics. The Scarcity-Adjusted PEG (sPEG) doctrine corrects the distortion by pricing the scarcity that produces growth, not just the growth itself. This piece runs the two lenses side by side and applies them to the AI infrastructure stack.

Scarcity Is the New Growth

For decades, institutional capital has used growth as the primary signal of value creation. P/E and PEG were built on a simple premise: faster growth deserves a higher multiple.

That premise works when demand is the binding constraint.

It fails when capacity is.

In constraint-bound systems, demand is effectively unlimited and capacity is the limiter. The companies that control scarce capacity do not merely participate in growth — they regulate it. This is the distinction the orthodox metric cannot see:

Participants benefit from expansion. Controllers determine its pace. Controllers capture the majority of long-term value.

Growth does not create that value. Scarcity captures it.

How Traditional PEG Works — and What It Assumes

The PEG ratio normalizes valuation by dividing forward P/E by expected growth:

PEG = Forward P/E ÷ Growth Rate

A PEG near 1.0 is read as fair; below 1.0, cheap relative to growth; above 2.0, expensive. As a screen it is genuinely useful — it stops investors overpaying for slow growth and flags fast growers that look expensive on P/E alone.

But PEG carries one silent assumption: that a unit of growth is a unit of growth, regardless of where it comes from. Every percentage point of expected EPS growth is treated as equally valuable and equally durable.

In a demand-driven market, that is a reasonable simplification. In a constraint-bound system, it is the precise point of failure.

Where Traditional PEG Breaks

When capacity is constrained, reported growth reflects capacity availability, not the size of the opportunity. A bottleneck controller running flat-out can post "moderate" growth while holding disproportionate structural leverage over the entire system's expansion.

Worse, PEG cannot distinguish between two growth profiles that look identical on the page:

A cyclical trough that will revert when the cycle turns, and a structural bottleneck whose constraint compounds as demand scales.

Both can present as "high growth, low multiple." PEG scores them the same. The market, knowing one is historically cyclical, refuses to pay up — and the apparent cheapness persists for years. PEG offers no mechanism to resolve which case you are looking at. That unresolved ambiguity is where structural mispricing lives.

The Correction: Scarcity-Adjusted PEG

sPEG inserts scarcity as a primary multiplier on growth:

sPEG = Forward P/E ÷ (Growth Rate × Scarcity Multiplier)

The Scarcity Multiplier is a proprietary coefficient derived from five factors:

• Irreplaceability — how essential and non-substitutable the asset or layer is
• Replication timeline — the capital and years required for a competitor to replicate it
• Supply concentration — how few players control the capacity
• Throughput control — the degree to which the asset governs system-wide throughput
• Demand durability — whether the demand is structural and multi-year, not a transient pull

A high multiplier lowers sPEG below PEG, surfacing controllers that orthodox screens read as merely fair. A low multiplier raises it, exposing participants whose growth is real but unprotected. The metric shifts valuation from a demand-centric model to a constraint-aware one — it identifies assets that regulate capacity rather than consume it.

‍

A Worked Example: Memory in the HBM Era

Consider a leading high-bandwidth memory (HBM) producer in early-to-mid 2026.

Under traditional PEG: A forward P/E in the low teens against very high near-term EPS growth produces a PEG in the neighborhood of 0.2x — among the lowest readings for a company of its scale. The orthodox conclusion is "statistically cheap." But the market has historically priced memory as a commodity cyclical, so the discount persists. PEG flags the cheapness and cannot explain why it endures. The investor is left with an unresolved question: structural bargain, or value trap before the next down-cycle?

Under sPEG: The five factors resolve it. Irreplaceability is high — leading-edge HBM is process-bound and supply is committed to the most advanced AI accelerator platforms. Replication timeline is severe — a new fab runs roughly $15–20B and two-to-three years. Supply concentration is extreme — three producers control the overwhelming majority of DRAM. Throughput control is direct — allocation decisions gate how fast AI compute can scale. Demand durability is structural — multi-year hyperscaler capex, not a consumer refresh cycle. A scarcity multiplier in that regime drives sPEG well below the headline PEG (in the exmxc book, this position scores near 0.07).

The reframe is the point: this is not a cyclical memory name that happens to look cheap. It is a structural bottleneck controller temporarily priced as a cyclical participant. Traditional PEG sees the price. sPEG sees the position.

(Illustrative figures as of early-to-mid 2026; see citations. The point is the method, not a price target.)

Application Across the Four Forces of AI Power

AI is not constrained by demand for compute, memory, or deployment — that demand is effectively unlimited. It is constrained by capacity, and capacity is controlled at each layer of the Four Forces:

• Compute — GPU supply, concentrated in NVIDIA
• Fabrication — advanced-node manufacturing, concentrated in TSMC
• Memory — HBM production, concentrated in SK Hynix and Micron
• Energy — power generation and grid interconnection

Each of these regulates throughput. None merely participates in AI expansion; each determines how fast expansion can occur. Traditional PEG misses the leverage. sPEG is built to surface it.

The Cross-Domain Proof: Authority Is Also Scarce

This is not only a semiconductor argument. Scarcity capture is universal, and my own clearest evidence came from media, not chips.

In media, scarcity is not manufacturing capacity — it is authority capacity, which cannot be manufactured on demand and must be accumulated over decades. The most valuable assets I worked on in media M&A were never the fastest-growing on a trailing basis. They were the ones that controlled irreplaceable authority in a category.

Billboard was not a bet on its growth rate; it was a recognition of its structural position as the definitive authority in music charts — it regulates recognition. SXSW occupies a singular, unreplicable intersection of technology, culture, and innovation, the product of decades of institutional accumulation. New Year's Rockin' Eve controls a fixed position in time and ritual that cannot be expanded or duplicated.

Just as a memory producer regulates computational visibility, Billboard regulates cultural visibility. Just as a foundry regulates technological convergence, SXSW regulates cultural convergence. In both domains, scarcity — not growth — governs value. sPEG is the instrument that prices it.

Implications for Institutional Allocation

Growth measures participation. Scarcity measures control.

Assets that control scarce capacity — fabrication, HBM, GPU compute, grid power, or cultural authority — capture the majority of long-term enterprise value. The sPEG doctrine gives allocators a way to identify those controllers before traditional growth metrics fully reflect their advantage, and to avoid paying control-level multiples for participants who merely ride the same wave.

Doctrine Statement

Scarcity, not growth, is the primary determinant of valuation in constraint-bound systems. Capacity constraint is universal — it governs semiconductors, energy, infrastructure throughput, and cultural authority alike. Traditional growth-based metrics systematically understate the structural advantage of the assets that regulate capacity. Scarcity-Adjusted PEG corrects this by pricing scarcity directly. Allocators who adopt it will identify durable winners earlier and allocate more accurately across the AI infrastructure stack and every other supply-constrained system.

This doctrine is an analytical framework for institutional decision-making, not investment advice or a recommendation to buy or sell any security. Figures cited are illustrative and time-stamped.

Read Related Topics:

→ sPEG (Scarcity-Adjusted PEG) — Lexicon→ sPEG Index Series→ Four Forces of AI Power→ Judgment Is the Last Scarce Resource→ sPEG Framework on GitHub

‍

‍

‍