AI Bubble or a Financing Stress Test?

This working paper examines whether the October 2026 selloff in AI-related equities signals the beginning of an AI investment cycle collapse or a repricing of financing and competitive risks. The analysis focuses on two developments: the sharp increase in secondary-market yields on CoreWeave's debt, highlighting the financial vulnerabilities of capital-intensive AI infrastructure expansion; and ChangXin Memory Technologies' (CXMT) announced progress in 4F² DRAM architecture, raising questions about future memory supply and competitive dynamics. Drawing on credit-market signals, industry forecasts, corporate disclosures, and institutional research, the paper distinguishes between underlying demand for AI computing capacity and the financial sustainability of the companies building that infrastructure. The central argument is that current evidence does not yet establish a fundamental collapse in AI computing demand, despite genuine financing pressures and growing concerns about an AI investment bubble. The paper also considers the possibility of near-term market stabilization while explicitly acknowledging the uncertainty of short-term price movements. To support future evaluation, it identifies observable conditions that could confirm or challenge its conclusions, including credit-market developments, hyperscaler capital expenditure, memory supply-demand dynamics, and corporate earnings expectations. This independent working paper is intended to establish a time-stamped, testable research perspective rather than provide investment recommendations. Information cutoff: October 19, 2026, 1:30 a.m. EDT (Toronto). Series: EraBridge Institute Research Note No. 003 | Version 1.0