Grid Standards v1.0

CC BY 4.0

Compute Supply Chain Intelligence Vocabulary

Published: May 2026 · Version: 1.0 · License: CC BY 4.0

Purpose

Grid Standards is an open, versioned vocabulary for compute supply chain intelligence. It defines the terms, classifications, and methodologies The Grid uses to observe and report on AI compute market conditions.

These definitions exist so that compute buyers, analysts, and policymakers can discuss operational market dynamics in language that is stable over time. Grid Standards are public, free, and open for reuse with attribution.

Grid Standards v1.0, The Grid, May 2026. thegridco.ai/standards

Pillar 1: Market Condition Descriptors

Six independent descriptors classify the state of AI compute markets. Each has exactly three values. Assessed daily at 05:30 AEST.

Descriptors are independent. Supply can be Tight while Demand is Light (a production bottleneck, not demand-driven). Price can be Firming while Forward is Easing (current pressure with upstream relief visible).

1.1

Supply

Tight | Adequate | Abundant

Availability of GPU compute capacity across tracked providers and rental markets, measured relative to the 30-day observed baseline.

Tight. Capacity utilisation elevated. Fewer rental offers than baseline. Single-quote pricing across multiple GPU models indicates allocation constraints.
Adequate. Capacity meeting demand at current price levels. Multiple competing offers per SKU. No unusual availability constraints.
Abundant. Surplus capacity. Increasing offer count, widening provider coverage, declining utilisation signals.

1.2

Demand

Strong | Moderate | Light

Downstream pull on compute capacity, measured through rate-limit complaint frequency, provider queue degradation, API latency drift, and social signal volume relative to the 7-day adaptive baseline.

Strong. Elevated rate-limit complaints. Provider incident volume above baseline. Social signal volume indicating user-visible capacity constraints.
Moderate. Demand signals within normal range. No unusual complaint volume or queue pressure.
Light. Below-baseline demand indicators. Declining complaint volume. Provider capacity underutilised.

1.3

Price

Firming | Steady | Easing

Direction of GPU rental pricing and token pricing across tracked providers, measured relative to the 14-day rolling average.

Firming. Median GPU rental pricing trending above 14-day average. Token pricing unchanged or increasing. Pricing power shifting toward providers.
Steady. Pricing within normal variance of 14-day average. No directional trend.
Easing. Median GPU rental pricing trending below 14-day average. New provider pricing competitive with or below established rates.

1.4

Trend

Accelerating | Stable | Decelerating

Rate of change in the Compute Pressure Index (CPI velocity, not CPI level) over the trailing 7-day window.

Accelerating. CPI rising at an increasing rate. Market pressure compounding. Conditions deteriorating for buyers faster than the previous period.
Stable. CPI velocity near zero. Conditions holding steady regardless of absolute level. A Stable trend at CPI 75 means conditions are bad but not getting worse.
Decelerating. CPI rising more slowly, flat, or falling. Pressure easing. Conditions improving for buyers or deterioration slowing.

1.5

Availability

Constrained | Available | Surplus

Provider incident frequency and severity relative to the 30-day rolling baseline. Measures whether buyers can obtain capacity today, regardless of price.

Constrained. Provider incidents elevated. Service degradation affecting model availability. Buyers facing allocation decisions rather than price optimisation.
Available. Normal incident frequency. All tracked providers operating within historical baseline. Capacity obtainable at market rates.
Surplus. Below-baseline incident rates. Providers stable. New capacity online or demand easing producing excess availability.

1.6

Forward

Tightening | Unchanged | Easing

Direction of upstream supply chain signals that predict buyer-side conditions 1-3 months ahead. Sources include TSMC monthly revenue, HBM allocation status, interconnection queue growth, and hyperscaler capex trajectory.

Tightening. Upstream signals indicating future supply pressure: foundry revenue acceleration, memory allocation sold out, interconnection queue delays, capex lagging demand.
Unchanged. Upstream signals within historical range. No change in forward outlook.
Easing. Upstream signals indicating future supply relief: new capacity announcements, memory production expansion, interconnection queue clearing, capex acceleration.

1.7 Color Convention

Descriptors use buyer-side color inversion. Green is favorable for the compute buyer. Red is unfavorable. This is the opposite of equity market convention.

Color	Meaning	Example
Green	Favorable for buyer	Supply: Abundant, Price: Easing
Amber	Neutral	Supply: Adequate, Price: Steady
Red	Unfavorable for buyer	Supply: Tight, Price: Firming

This color convention is permanent and will not change in future versions.

Pillar 2: Compute Pressure Index (CPI)

A 0–100 composite score measuring aggregate pressure on AI compute markets from the buyer's perspective. Higher values indicate tighter conditions. Lower values indicate buyer-favorable conditions. Calculated daily at 05:30 AEST.

2.1 Sub-Indicators

CPI is composed of six weighted sub-indicators, each independently normalised to 0–100 using min/max scaling against its 30-day rolling history.

Sub-Indicator	Weight	Measurement
GPU Rental Pricing Trend	25%	7-day rolling average of H100 SXM daily median spot pricing, normalised against 30-day range
Memory Pressure Score	20%	7-day rolling average of DRAM spot pricing (DDR5 + GDDR6 8Gb), normalised against 30-day range
Provider Incident Rate	15%	48-hour severity-weighted incident count. Severity: critical ×4, major ×3, minor ×1, maintenance ×0.5
OCPI Spot Index	15%	Latest H100 spot value from the Ornn Compute Price Index, normalised against 30-day range
Supply Chain Forward Signal	15%	Composite: TSMC monthly revenue MoM%, interconnection queue capacity, SEC-filed capex trajectory
Social Signal Volume	10%	Daily post count across 11 compute-related X/Twitter keywords, normalised against 14-day range

2.2 Normalisation

normalised = ((value - min) / (max - min)) × 100

Clamped to [0, 100]. When min equals max: defaults to 50 (neutral). When no data exists: contributes 50 at its weight.

2.3 Composite

CPI = Σ (normalised_sub_indicator × weight)

Weighted sum, rounded to two decimal places, clamped to [0, 100].

2.4 Bands

Range	Band	Buyer Interpretation
0–25	Abundant	Buyer’s market. Favorable pricing, ample availability. Spot purchasing conditions.
25–45	Balanced	Equilibrium. Supply meeting demand at stable prices. Standard procurement posture.
45–65	Tight	Emerging pressure. Pricing firming, availability narrowing. Consider locking reservations.
65–80	Constrained	Seller’s market. Premium pricing, limited availability. Secure capacity commitments.
80–100	Critical	Structural shortage. Pricing spikes, widespread degradation. Emergency procurement posture.

Band boundaries are permanent. They will not change in future versions.

2.5 Comparable Indices

CPI follows the precedent of published composite indices: VIX (equity volatility, 0–100), CNN Fear & Greed (market sentiment), PMI (manufacturing activity), BDI (shipping demand). No equivalent composite index exists for AI compute markets. CPI is the first.

Pillar 3: Compute Supply Chain Layer Taxonomy

A 12-layer system from raw materials to financial instruments. Each layer has characteristic lead times determining how long upstream changes take to affect buyer-side conditions.

3.1 Layers

Layer	Name	Lead Time
0	Raw Materials & Wafer Supply	18–24 months
1	Fabrication (Foundry)	4–6 months
2	Advanced Packaging	3–6 months
3	Memory (HBM/DRAM)	3–6 months
4	Chip Design & Architecture	2–4 months
5	Networking & Interconnects	3–6 months
6	Systems Integration	1–3 months
7	Data Center & Power Infrastructure	12–36 months
8	Cloud Deployment & GPU Markets	1–4 weeks
9	Token Economics & API Layer	Real-time to 1 week
10	Buyer & Consumer Demand	Real-time
11	Financial & Forward Pricing	1–6 months

3.2 Causal Dependency Chains

Chain 1 — The HBM Bottleneck (dominant constraint, 2025–2027)

HBM allocation sold out→GPU production capped (3–6mo)→Cloud GPU availability constrained (1–3mo)→GPU rental pricing firms (real-time)→Token pricing floor established (1–3mo)

Chain 2 — The Packaging Constraint

CoWoS packaging at limit→GPU lead times extend (3–6mo)→Rental pricing spikes (1–3mo)→Provider margin compression or price pass-through

Chain 3 — Power Infrastructure (structural, 3–6 year resolution)

Interconnection queue saturated→Data center capacity delayed (12–36mo)→Structural compute scarcity (6–12mo)→Elevated pricing persists

3.3 Supply Chain Zones

Zone	Layers	Description
Upstream	0–3	Raw materials through memory. Where physical constraints originate
Midstream	4–7	Chip design through data center infrastructure. Where constraints are resolved
Downstream	8–11	Cloud markets through financial instruments. Where buyers experience effects

Pillar 4: Observation Classification

Every data point in The Grid's editorial output is classified by its epistemic status. The classification is always visible to the reader.

OBSObserved

Directly scraped from a public source with machine-readable provenance.

"H100 SXM median $2.94/hr [OBS] (GetDeploying, n=58, 2026-05-10)"

ASTAssessed

Editorial synthesis with stated confidence and reasoning basis.

"Memory constraint likely to firm rental pricing Q3 [AST] (confidence: medium, basis: HBM sold-out + 3–6 month lag)"

ESTEstimated

Model-derived value where observed data is unavailable. Methodology stated.

"CoWoS monthly output approximately 35K wafers [EST] (basis: TSMC commentary + sell-side consensus)"

Pillar 5: Temporal Horizon Format

Format communicates confidence. As the time horizon extends, the format shifts from precise to probabilistic to directional. This prevents false precision.

Horizon	Format	Confidence
Today (0–24h)	Specific numbers with deltas	Very high
This Week (1–7d)	Ranges with directional indicator	High
This Month (1–4w)	Directional probabilities	Medium
This Quarter (1–3m)	Above/below baseline with probability weights	Low–medium
Structural (3+m)	Narrative with causal attribution	Low on timing, high on direction

Governance

Public Source Priority. No proprietary data, no insider information, no NDA-protected relationships. Value comes from structure, methodology, and daily discipline — not access.

No Investment Positions. The Grid does not hold investment positions in the companies it covers. Structural commitment.

Prediction Accountability. Every forecast is falsifiable, time-bound, and tracked to resolution. Calibration statistics — including failures — are published.

Methodology Transparency. Weights, formulas, band definitions, and normalisation methods are published in this document. Changes are versioned. Historical values labelled with methodology version.

Permanence. Band boundaries and color conventions in v1.0 are permanent. Historical CPI values maintain their original meaning indefinitely.

Changelog

v1.0 — May 2026
Initial publication. Five pillars: Market Condition Descriptors, Compute Pressure Index, Compute Supply Chain Layer Taxonomy, Observation Classification, Temporal Horizon Format.

Grid Standards v1.0. Published by The Grid (thegridco.ai).

Licensed under CC BY 4.0 — free for reuse with attribution.

Grid Standards v1.0, The Grid, May 2026. thegridco.ai/standards