Model Monoculture 2026: How Concentrated Is Enterprise AI, and Do CIOs See the Risk?
Published
Regulators warn that firms built on the same few AI models could fail, or act, in the same way at the same time. We measured how concentrated enterprise AI really is, and whether CIOs are prepared.
Blended research: grounded in surveys, powered by AI (read our methodology)Key statistics
83%–88%
of tech staff at large firms in every industry have worked with OpenAI models; about a third rely on a single AI vendor
Evidence: MeasuredSurveyed: Stack Overflow Developer Survey 2025
88%
of US enterprise LLM API spend goes to three providers
Evidence: CitedPublished: Menlo Ventures 2025
39% vs 8%
of CIOs in a calibrated AI twin panel are confident they could switch AI providers within 90 days, versus those with a tested exit plan
Evidence: Calibrated modelAI-modelled: 22decisions Twin Panel
Get the full report
This page is the summary. The full report adds the complete findings, segment analysis, recommendations and methodology appendix.
From$2,200
By purchasing you agree to our Terms and Licence.
Pay by invoiceSummary
The Financial Stability Board, the Bank of England, and the ECB have warned that concentration among AI providers could push firms towards correlated behaviour. This brief tests that concern inside enterprises: how concentrated AI use really is, and whether CIOs are prepared.
Enterprise AI turns out to be an oligopoly rather than a monoculture: three providers dominate spending, yet most large firms run several models. The bigger risk is shallow diversification: confident CIOs, few tested exit plans, and little attention to shared blind spots.
What's in the full brief
- What regulators say, and which claims the evidence supports or overstates
- How concentrated enterprise AI is, by industry
- Do CIOs see the risk? Strategy, dependence, and exit plans
- The confidence gap
- Sector differences: why banks are ahead
- What CIOs and IT services firms should do
- Methodology and sources
6 pages, PDF.
Every figure is labelled by source: surveyed, AI-modelled, or published.
Written by22decisions Research
Varun Shourie founded 22decisions to enable technology leaders to prepare for, and get ahead in, the AI era.
About 22decisionsGet the full report
This page is the summary. The full report adds the complete findings, segment analysis, recommendations and methodology appendix.
From$2,200
By purchasing you agree to our Terms and Licence.
Pay by invoice- Topics:
- AI risk, AI governance, Third-party risk
- Sectors:
- Banking, Healthcare, Manufacturing, Retail and Consumer Goods, Insurance