When the bubble pops, what do you actually own?

· Source: Artificial Intelligence on Medium · Field: Business & Management — Corporate Strategy & Leadership, Artificial Intelligence & Machine Learning, Capital Markets & Investment Management · Depth: Intermediate, long

Summary

The AI market shows signs of a financial bubble, fueled by circular funding where chipmakers invest in AI companies that then commit trillions to buying chips and cloud services from those same providers. Hyperscaler capital spending is projected to reach \$725-800 billion in 2026, with analysts questioning a \$1.5 trillion payback. Despite this, studies from MIT and BCG indicate 95% of enterprise generative AI pilots yield no measurable return, and only 4% create substantial value. Drawing parallels to the dot-com era's "dark fiber" overbuild, the author suggests that while valuations may crash, underlying AI infrastructure and capabilities will persist. Companies must distinguish between "rented dependency" on vendor APIs and "owned capability" built with in-house models and skilled teams. Chinese open models, offering significantly cheaper alternatives (e.g., 18 cents per million tokens vs. \$4 for American models), enable owned capability despite potential future export restrictions.

Key takeaway

For Directors of AI/ML evaluating long-term strategy, prioritize building owned AI capability over rented dependency. If you rely solely on vendor APIs, your product's resilience is tied to external pricing and access, which can be repriced or revoked. Instead, secure open-weight models on your infrastructure and invest in developing your team's expertise to direct and judge AI systems. This ensures your intelligence remains an enduring asset, insulating your operations from market corrections and potential policy shifts. For your organization, the "bubble pop" then becomes merely weather.

Key insights

Own AI capability, don't just rent it, to survive market corrections and policy shifts.

Principles

Method

Build owned capability by acquiring open models and developing in-house expertise, rather than relying solely on vendor APIs.

In practice

Topics

Best for: Investor, Executive, AI Architect, Director of AI/ML, VP of Engineering/Data, CTO

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence on Medium.