The current phase of artificial intelligence development (2024–2026) is characterized by a definitive shift from a "weightless" software narrative to an industrial reality. Progress is now bound by physical throughput: high-end GPUs, massive data-center shells, reliable power contracts, and complex networking. This environment has birthed a specific financial architecture that may be both a growth catalyst and a source of systemic fragility.The Circular AI EconomyA central feature of this boom is circular financing, or "round-tripping," where a small group of companies act simultaneously as investors, suppliers, and customers. In these loops, capital-rich providers (like Nvidia or Microsoft) fund AI developers (like OpenAI), who then spend large portions of that capital back on the funder's cloud services or hardware. The most prominent example is Nvidia's planned $100 billion investment in OpenAI, which is expected to be spent largely on Nvidia’s next-generation "Vera Rubin" infrastructure. Critics warn this creates "revenue of lower quality" than organic demand, potentially masking the true strength of market pull.Structural FlashpointsThe industry is currently navigating several major risks:• The Revenue Gap: Estimates suggest the AI sector will need approximately $2 trillion in annual revenue by 2030 to justify current and planned investments in infrastructure. For comparison, OpenAI’s annualized revenue reached $13 billion in August 2025—a fraction of the required breakeven.• Utilization Risk: Companies are signing long-dated "take-or-pay" commitments, paying for reserved compute capacity regardless of actual usage. If product adoption slips, these fixed obligations could become a "millstone" for buyers.• Physical Bottlenecks: Power is now the "quiet governor" of AI expansion. AI data centers could consume up to 12% of all US electricity by 2030, forcing providers to build in remote locations (like North Dakota), which risks creating "stranded assets" if technology shifts.ROI Crisis and "AI Washing"The optimistic market narrative is colliding with underwhelming operational data. A landmark 2025 MIT study found that 95% of corporate AI projects fail to generate meaningful returns. Researchers point to a "learning gap"—the inability of AI to adapt to complex, non-linear corporate workflows.The fall of Builder.ai serves as a cautionary tale of "AI washing". Valued at $1.5 billion, the startup claimed its "Natasha" AI could autonomously build software. In reality, the work was manually performed by approximately 700 engineers in India. The company collapsed in 2025 following allegations of 300% revenue inflation and a creditor-led seizure of its cash.Future Scenarios: Consolidation vs. CrashExperts identify two primary ways the current cycle might end:1. Crash Pathways: A "slow crash" driven by ROI disappointment and capex deceleration, or a "fast crash" triggered by credit tightening in the private credit layer, which currently manages over $1.6 trillion in assets.2. The Squeeze (Consolidation): The sector may survive but concentrate value in incumbents who control "choke points"—default surfaces like search, browsers, and operating systems. Giants like Google and Microsoft can bundle AI as a "retention feature" or durning AI into a background feature of existing suites, effectively squeezing out standalone competitors through cross-subsidies.Early warning signals for a correction include shortening contract durations, aggressive discounting for compute capacity, and a shift in corporate language from "announcing new capacity" to "efficiency and utilization".AI tools were used in the translation. Hosted on Acast. See acast.com/privacy for more information.
Become a supporter of this podcast: https://www.spreaker.com/podcast/the-world-between-us--6886561/support.
