Companies with rising AI costs are increasingly using tools that tap cheaper models, including some from China, putting pricing pressure on OpenAI and Anthropic
Startups and tech giants alike are mixing and matching AI models to avoid the premium prices charged by industry leaders
Context & Ripple Effects
Related coverage traces a cost-driven alternative supply chain for AI: Chinese developers including 01.ai and DeepSeek have sought lower-cost model development under export controls, and US startups have increasingly adopted open-weight Chinese models for price, customization, and adequate capability.
This report extends that pressure to a broader set of buyers, including large technology companies. It matters because model selection is becoming an operating-cost decision rather than a default commitment to a single frontier-model provider.
First-order effects
- Startups and large technology companies can route workloads across lower-cost models instead of paying premium rates for every AI task.
- OpenAI and Anthropic face immediate pressure to defend enterprise spending through pricing, model efficiency, or clearer performance differentiation.
Second-order effects
- Model-routing and comparison tools gain importance as customers need to match tasks to cost and capability across multiple providers.
- Rivals with lower-cost or open-weight offerings gain a clearer route into enterprise deployments, forcing leading vendors to compete more directly on unit economics.
Third-order effects
- If multi-model purchasing persists, frontier models may become a premium tier within a segmented market rather than the default choice for all AI workloads.
- The competitive boundary could shift from owning the strongest single model toward controlling the tooling, distribution, and cost management layer that determines which model is used for each task.
The trend: Enterprise AI is moving toward portfolio-based model procurement, with cost efficiency increasingly constraining the pricing power of frontier-model vendors.