Independent research on AI economics and infrastructure.
AI is attracting enormous investment while changing the economics of the businesses expected to pay for it. Threadbaire investigates where the value is actually being created, who gets to keep it, and who ends up paying for it.
By Lida Liberopoulou · All research CC BY-SA 4.0
The companies building AI are increasingly also financing, supplying, and buying from one another. These relationships make it difficult to distinguish independent demand from activity financed within the same ecosystem, or to see where financial risks ultimately accumulate.
Threadbaire examines how these dependencies formed, what sustains them, and where the costs and risks might land when those relationships come under pressure. The research draws on earlier infrastructure industries and the evolution of the software economy to identify patterns that company-by-company analysis can miss.
The foundational argument
Why AI is undermining the economics of the software industry, and why attempts to preserve existing business models may deepen the problem.
The structural and historical framework
How AI's industrial costs collide with eroding revenue, how computing once escaped the historical infrastructure pattern, and how the conditions of that escape were gradually reversed.
Part 1: AI's Gilded Age · Part 2: Tech's Great Escape · Part 3: The Return of the Bind
The applied analysis
AI financing relationships read through corporate disclosures, with claims recorded before the earnings and then checked against the filings.
AI did not create the structure producing its concentration. It inherited it. Part 3 traces how the separations that made the PC, independent software, and the open internet possible were gradually reversed through cloud infrastructure, subscriptions, externalised technical capability, gated distribution, joined identity, and the institutional conditions that allowed those dependencies to accumulate.
4 September 2026
Six structural claims about hyperscaler earnings disclosure, filed before the event with pass/fail conditions, all confirmed. The score matters less than what it reveals: a recurring pattern in which investment effects, commercial demand, and infrastructure costs are each visible, but the connections between them are not.
31 July 2026
Nvidia is guaranteeing its own customer's debt. Google is booking unrealised gains on companies that buy its cloud compute. Amazon's Anthropic position contributed $12.3 billion to quarterly income from a price adjustment. On Wednesday and Thursday, Microsoft and Apple report from opposite sides of this pattern. Six falsifiable predictions about what the filings will disclose and what they will leave impossible to calculate.
27 July 2026
I take a small number of commissions. Each is an independent, written analysis of a specific AI exposure, thesis or decision: what the evidence shows, what it depends on, and what to watch.
Threadbaire combines economic and historical analysis with direct experimentation using the systems it examines. Working across AI models reveals practical dependencies and limitations that aren't always visible from outside. The research uses an open methodology that keeps context portable and decisions traceable across tools.
The method and templates are freely available.
