Scry Fund

AI Infra Framework

AI Infrastructure Bubble Tracker

Utilization, leverage and shortage pricing signals for the post-Nvidia-capex cycle.

24 May 2026 · YK Research

The Mispricing

AI demand is real. The bubble risk is not “AI is fake.” The risk is that investors extrapolate peak scarcity into permanent economics. In every capacity bubble, the market first pays for access, then realizes the scarce asset was financed at the wrong price, with the wrong leverage, against utilization that only existed during shortage.

Main edge
scarcity fade
Best signal
utilization
Kill switch
debt
Model
fiber 2.0

The tracker asks one question: is AI capex producing durable ROIC and useful GPU-hours, or are lower-quality capacity wrappers being bid up because everything with “AI infrastructure” in the label temporarily looks sold out?

Framework based on Gavin Baker’s AI-infrastructure bubble comments, prior telecom/shale/solar cycles, and YK Research interpretation. Static first version; live data collectors should be treated as next phase.

Real Pricing Snapshot

This chart is actual public pricing scraped from RunPod and Lambda pricing pages, not a made-up bubble score. It tracks posted on-demand GPU-hour prices; it does not claim true utilization.

Rows
14
Sources
2
H100 range
$1.99–$4.29/hr
As of
2026-05-24

Normalized Memory Cost

Same source rows normalized by advertised VRAM. Useful for spotting scarcity premium, but still imperfect because instance bundles include CPU, RAM, storage and networking.

Public on-demand GPU pricing snapshot collected from provider pricing pages. Prices are listed per GPU-hour before tax unless otherwise noted. This is pricing/availability data, not utilization data.

What We Can Track Right Now

GPU rental $/H100-equivalent-hour

trackable

Source: cloud list prices, marketplaces, spot feeds

Read: falling price while capacity expands = shortage easing

Provider availability / lead time

trackable

Source: AWS/GCP/Azure/OCI pages, neocloud quote screens

Read: more regions available = weaker scarcity premium

Cloud capex productivity

trackable

Source: AMZN/MSFT/GOOGL/META filings

Read: capex up + ROIC stable = healthy; capex up + ROIC down = air pocket

Operating margin under AI load

trackable

Source: AWS/Azure/GCP segment disclosures

Read: margin compression says suppliers/power capture value

Leverage in infra wrappers

trackable

Source: public debt, private-credit reports, lease disclosures

Read: debt-funded GPUs are fragile if rental rates fall

Quality minus junk spread

trackable

Source: public equity baskets

Read: junk outperforming quality = shortage beta / mania

HBM/DRAM pricing and supply

trackable

Source: Micron/SK Hynix/Samsung commentary, contract-price reports

Read: memory still tight = bottleneck real; supply surge = margin risk

Social / options mania

trackable

Source: call volume, short interest, X/Reddit mentions

Read: retail flows moving low-float AI labels = bubble behavior

Proxy-only / hard data

  • True global GPU utilization outside owned clusters
  • Model FLOPs utilization by frontier lab
  • Private neocloud occupancy, cancellation rights and customer concentration
  • GPU loan-to-value, covenants and refinancing terms in private credit
  • Cost per useful token by model after uptime, batching, software efficiency and power

Source Rows

Raw rows behind the charts. Every value comes from the committed public-pricing snapshot; no placeholder GPU coverage scores remain.

RunPod · H200
141GB VRAM · Secure Cloud Pods · listed
$3.59/hr
$0.0255/GB-hr
RunPod · B200
180GB VRAM · Secure Cloud Pods · listed
$5.98/hr
$0.0332/GB-hr
RunPod · H100 NVL
94GB VRAM · Secure Cloud Pods · listed
$2.59/hr
$0.0276/GB-hr
RunPod · H100 PCIe
80GB VRAM · Secure Cloud Pods · listed
$1.99/hr
$0.0249/GB-hr
RunPod · H100 SXM
80GB VRAM · Secure Cloud Pods · listed
$2.69/hr
$0.0336/GB-hr
RunPod · A100 PCIe
80GB VRAM · Secure Cloud Pods · listed
$1.19/hr
$0.0149/GB-hr
RunPod · A100 SXM
80GB VRAM · Secure Cloud Pods · listed
$1.39/hr
$0.0174/GB-hr
RunPod · L40S
48GB VRAM · Secure Cloud Pods · listed
$0.79/hr
$0.0165/GB-hr
Lambda · B200 SXM6
180GB VRAM · 1x GPU instance · listed
$6.99/hr
$0.0388/GB-hr
Lambda · GH200
96GB VRAM · 1x GPU instance · listed
$2.29/hr
$0.0239/GB-hr
Lambda · H100 SXM
80GB VRAM · 1x GPU instance · listed
$4.29/hr
$0.0536/GB-hr
Lambda · H100 PCIe
80GB VRAM · 1x GPU instance · listed
$3.29/hr
$0.0411/GB-hr
Lambda · A100 SXM
40GB VRAM · 1x GPU instance · listed
$1.99/hr
$0.0498/GB-hr
Lambda · A100 PCIe
40GB VRAM · 1x GPU instance · listed
$1.99/hr
$0.0498/GB-hr

Provider Universe

First live collector should expand from the current RunPod/Lambda snapshot into public cloud quotes and marketplace listings, then layer availability status and contract term. Do not clone proprietary SemiAnalysis data; build independent snapshots from public/provider data.

AWS P4/P5/P6, Trn1/Trn2/Trn3GCP A3/A4/TPUAzure ND/GB-seriesOCI GPU bare metalCoreWeaveLambdaCrusoeRunPodVast.aiNebiusTogether / Modal / Replicate

Price metric

$/GPU-hour and $/GB VRAM-hour are live in this static snapshot.

Availability metric

Current snapshot records “listed”; next collector should record available / limited / waitlist / sold out.

Utilization proxy

Only build after repeated price + availability snapshots exist. No made-up utilization scores.

Owned-cluster metric

DCGM / nvidia-smi / ROCm SMI gives real utilization only if we operate the machines.
Sources: RunPod (https://www.runpod.io/gpu-instance/pricing); Lambda (https://lambda.ai/pricing)

Leverage: Where Bubbles Break

Hyperscalers

Track capex/operating cash flow, depreciation/revenue, cloud operating margin and ROIC. Healthy if AI capex keeps ROIC stable or rising.

Neoclouds

Track debt/contracted revenue, interest/gross profit, customer concentration, GPU useful life assumptions and refinancing windows.

Power / data center

Track debt/MW, committed vs energized MW, interconnection queue, turbine/transformer lead times and revenue per MW.

The telecom-fiber lesson is brutal: demand can be real and investors can still lose money if the asset is financed at shortage prices and bandwidth prices collapse before the debt amortizes. GPU wrappers have the same shape if H100/H200 rents normalize faster than lenders expect.

Prior Bubble Analogues

No numeric overlap score here. These are qualitative analogues only; turning them into numbers without a source would be fake precision.

Telecom fiber

Closest analogue: real demand, debt-funded overbuild, unit-price collapse.

Shale

High-cost capacity looks brilliant during shortage, then breaks when marginal price normalizes.

Solar / EV

Real adoption plus too many entrants; leaders survive while weak balance sheets die.

Dot-com

Real usage growth with bad monetization and new metrics hiding weak unit economics.

Analogue Reads

  • Dot-com: real user adoption, bad monetization, new metrics hiding weak unit economics.
  • Telecom fiber: real bandwidth demand, debt-funded overbuild, price per unit collapses.
  • Shale: high-cost producers outperform during shortage, then die when marginal price normalizes.
  • Solar / EV: real adoption curve, too many entrants, capex boom, margin collapse, leaders survive.

Live Tracker Roadmap

1. Pricing index

Normalize all quotes to $/GPU-hour, $/GB VRAM-hour and $/FP8 PFLOP-hour across Nvidia and AMD chips.

2. Availability index

Capture available, limited, quote-only, sold-out, unknown by provider, region, SKU and contract term.

3. Effective-cost index

Adjust list price by uptime penalty, utilization penalty, memory size, interconnect quality and software maturity.

4. Financial overlay

Join capex, depreciation, operating margin, ROIC, debt and backlog for hyperscalers, neoclouds, power and memory names.

5. Bubble score

Score whether shortage pricing is being validated by utilization and ROIC or just financed by leverage and narrative.

Data Audit Notes

  • Pricing charts are backed by committed JSON from public RunPod and Lambda pricing pages, refreshed by scripts/fetch-ai-infra-gpu-pricing.py.
  • The current tracker uses posted on-demand GPU-hour prices and advertised VRAM. It does not infer utilization from invented scores.
  • Global utilization, private neocloud occupancy, cancellation rights and private-credit LTV remain explicit data gaps until source feeds exist.
  • Provider prices are bundled instance economics; CPU/RAM/storage/network quality and software maturity can make effective token cost differ materially.

How To Read It

Not bubble

High capex, high utilization, rising ROIC, durable cloud margins and stable/rising revenue per GPU/MW.

Early warning

Capex up, utilization flat, GPU rentals falling, capacity easier to source, depreciation/revenue rising.

Bubble

Low-quality names outperform, leverage rises, retail/call-volume mania appears, and baskets replace company-level analysis.

Opportunity

Misclassified names in the wrong bucket: e.g. switch/interconnect exposure treated as copper-loser exposure.

Gavin-style bottom line: do not ask whether AI is real. Ask who owns scarce supply, who financed it, what utilization is assumed, and what happens when scarcity ends.