Enterprise adoption of lower-cost AI models accelerates amid rising inference costs and billing concerns
Other ● Confirmed 72% confidence first seen
Enterprises worldwide are increasingly shifting from expensive Western AI services like OpenAI and Anthropic to cheaper alternatives, including Chinese AI models and open-source options, driven by sticker shock from usage-based token billing and concerns about data privacy. Organizations are struggling to track and control AI infrastructure spending, with many deploying models at significantly reduced capacity utilization while reconsidering their AI investments.
Decision brief
- What changed
- Multiple surveys and reports indicate enterprises are shifting from usage-based-billed Western AI services (OpenAI, Anthropic) toward cheaper Chinese and open-source models, citing sticker shock from token-based pricing, data privacy concerns, and difficulty tracking AI infrastructure costs. A KPMG survey found 29% of executives struggling to understand scaling costs and nearly half reconsidering deployments, while a separate 107-enterprise survey found 83% report GPU utilization at or below 50% and only 44% can rigorously track compute costs.
- Why it matters
- Vendors are transferring infrastructure cost risk to customers via usage-based billing just as enterprises lack the financial tooling to forecast or control spend, creating budget exposure that Forrester expects 80% of decision-makers to see rise by 2027. This is prompting a strategic reassessment of AI vendor lock-in, data governance, and build-vs-buy decisions, with some enterprises turning to geopolitically sensitive Chinese models purely on cost grounds, raising compliance and security tradeoffs alongside savings.
- Evidence
- Coverage draws on named third-party surveys (KPMG, Forrester, Gartner, Bain & Company) and a 107-enterprise study cited by VentureBeat, corroborated by three trade outlets (The Register x2, The New Stack) and one policy research source (CSET Georgetown) independently describing the same cost-shock and cost-tracking dynamics, suggesting consistent signal across sources rather than a single outlet's framing.
- What remains uncertain
- It is unclear how many enterprises have actually completed migrations to Chinese or open-source models versus merely evaluating them, and whether cost savings hold up once factoring in retraining, support, and compliance overhead; the reliability and independence of vendor assurances on data usage (zero retention claims) also remain unverified by the coverage. The degree to which Chinese model adoption is concentrated in specific regions or industries is not specified.
- Monitor next
- Watch upcoming earnings calls and procurement disclosures from major enterprises for concrete migration numbers away from OpenAI/Anthropic toward open-source or Chinese models, and track whether Forrester's projected 2027 budget increases materialize in actual enterprise IT spending reports.
Analytical support, not advice — assumptions and open questions stated above.
Source coverage
Companies turn to Chinese AI models to cut costs
CSET Georgetown
Sticker shock has execs rethinking this whole AI thing
The Register
The enterprise strikes back: Keep your data, ditch the AI bill.
The New Stack
AI vendors have found someone to pay their infrastructure bills: You
The Register
The AI compute gap: Enterprises are buying infrastructure faster than they can measure what it costs
VentureBeat AI
independent