A growing library of deeply researched subjects and themes. Each brief builds a single subject up from the primary documents, surfaces the facts that carry the story, and leaves the conclusions to you.
Two full reports on CoreWeave: Atlas examines the financials and valuation scenarios. AIR investigates the business, its financing and the debates behind the stock.
Read the brief →Two full reports on CoreWeave: Atlas examines the financials and valuation scenarios. AIR investigates the business, its financing and the debates behind the stock.
Three neoclouds buy the same Nvidia silicon and sell it to the same handful of buyers. What durably separates them is the price of the money used to buy the fleet: 9% at CoreWeave, under 2% at IREN and Nebius.
Every enterprise rents the same models at the same collapsing prices. What separates working AI from theater is not where it sits on the org chart but whether anyone holds the authority to kill a use case.
Shadow AI is a governance gap, not a malice problem. Competent employees close a capability gap under time pressure with tools that leave no artefact, and data that enters a training run cannot be deleted.
The money in pharma used to come from owning a patent and pricing it freely. GLP-1 is where that breaks: the durable rent has moved into manufacturing cost per delivered milligram and the discovery royalty.
States are not hoarding metal; they are taking positions in the plants that process it. Metal by metal, which ones are becoming strategic reserves and who captures the rent.
Two-thirds of a US ride-hail fare goes to the driver. Remove the driver and the wage pool does not vanish — it gets fought over by platforms, AV makers, fleet owners and insurers.
Durable rent in the two-trillion-dollar private-credit market sits in two layers only: liabilities that cannot be recalled, and origination broad enough to leave direct lending when it stops paying.
Why AI pilots that work still create no value — the four gates where the money leaks, told through documented programs from MD Anderson to Klarna.
AI-era rack power density — roughly 135 kW for an NVIDIA GB300 NVL72 cabinet against a typical rack near 10 kW, with about 90% of the heat rejected to liquid — is the mechanism…
The problems that actually cost enterprises money and expose them to risk are structural and unglamorous — unpredictable unit economics, data-governance regimes built for a…
The durable profit has left the GPU die for the rack-scale fabric that turns 72 chips into one machine — the thing NVIDIA's rivals can't yet replicate. The memory and networking rents stacked around it are real, and their owners have already published the date they end.
Colorado's water scarcity is real and worsening — but the state's legal architecture routes the rent away from anyone who could book it as profit. The shortage is real; the "water play" mostly isn't.
None of the hyped SMR pure-plays yet clears the survivor bar — and what the developers actually own that's hard to replicate is the capital they raised, not the physics they designed.
An AI data center can only be built as fast as its slowest input, and in 2026 two inputs are slow at once: memory chips, whose makers cut capacity in the 2023 slump and need two…
Of the three segments SpaceX reports, durable rent sits in exactly one — Connectivity — and the lock-up overhang looks mispriced.
The region's crises aren't ideology — they're one recurring engine: deficit-financed redistribution behind an administratively defended exchange rate.
The profit that lasts sits in proprietary engineering estates — the controller, its tool, its protocol — not the robot arm.
Structurally-defended profit hides in three places the market hasn't paid for: enrichment, the already-built fleet, and contracted firm power.
Where durable value collects across new-home construction, land development and the entitlement stack — mapped for operators, not headlines.
The densest controlled energy source we have — how its physics and hard constraints explain both its stalled past and its present revival.