Model routing squeezes OpenAI and Anthropic revenue

Enterprises are increasingly adopting "model routing" to match tasks with the cheapest adequate AI model, a shift that threatens revenue for OpenAI and Anthropic. Glean CEO Arvind Jain estimates 95% of enterprise AI usage still defaults to expensive frontier models, while Cognition CEO Scott Wu says companies can achieve 5 to 10 times better cost efficiency on routine tasks using smaller models. The trend, driven by CFOs and boards cracking down on AI overspending, undermines the premium-demand assumptions behind OpenAI and Anthropic's valuations.

Model routing squeezes OpenAI and Anthropic revenue CNBC reports that enterprises are increasingly using "model routing," which matches tasks to the cheapest adequate model rather than defaulting to the largest frontier model. CNBC cites Glean CEO Arvind Jain estimating that roughly 95% of enterprise AI usage still routes to the most expensive frontier models today. Cognition CEO Scott Wu is quoted by CNBC saying firms can achieve roughly 5 to 10 times better cost efficiency on routine tasks by using smaller models. CNBC frames the shift as a response to CFOs and boards cracking down on AI overspending and warns that this efficiency trend could undermine the premium-demand assumptions embedded in valuations for OpenAI and Anthropic . What happened CNBC reports that corporate customers are moving toward a practice called model routing , which matches each task to the most cost-effective model rather than defaulting to the largest frontier model. CNBC attributes an estimate from Glean CEO Arvind Jain that roughly 95% of enterprise AI usage is still being run on the most expensive frontier models. CNBC also cites Cognition CEO Scott Wu as saying companies can get about 5 to 10 times better cost efficiency on routine work by using smaller models for those tasks. The story frames the change as driven by pressure from CFOs and boards responding to AI budgets running ahead of expectations. CNBC reports that this dynamic creates a risk to the premium-demand assumptions behind OpenAI and Anthropic valuations. Editorial analysis - technical context Industry-pattern observations: Model routing reduces average cost-per-query by assigning low-complexity requests to cheaper, lower-latency models and reserving frontier models for high-complexity work. Companies that implement per-request selection typically combine lightweight classifiers, heuristics, or ensemble scorers to decide routing, plus instrumentation to measure latency and cost at scale. For practitioners, this increases emphasis on model evaluation against realistic workload mixes, and on building reliable routing logic and monitoring to prevent regressions when cheaper models handle production traffic. Context and significance The CNBC piece positions model routing as a financial control mechanism responding to enterprise budget overruns. If broadly adopted, routing could materially reduce the volume of queries that flow to premium-priced frontier models, which in turn affects revenue pools for providers that price primarily for highest-capability usage. For ML engineering teams, the implication is a greater need to benchmark models by task-level utility and total cost of ownership, rather than optimizing solely for top-line capability metrics. What to watch - •Adoption signals: enterprise vendor disclosures, case studies showing routing at scale, or third-party telemetry revealing shifts in query mix. - •Technical signals: open-source tooling or managed services that simplify routing decisions and model orchestration. - •Commercial signals: any changes to provider pricing, new tiered offerings, or contractual terms that reflect routing-aware consumption. Editorial analysis: Broadly, the trend toward routing reflects a maturing procurement discipline in large organizations, where cost controls and measurable ROI increasingly shape model selection. Practitioners should expect engineering efforts to shift toward operationalization: robust A/B testing, routing logic, and metrics that tie model choices to business outcomes. Scoring Rationale The story signals a notable shift in enterprise procurement and operational practice that could alter usage patterns and revenue mixes for major AI vendors. It is relevant to practitioners but not an immediate technical breakthrough. Practice interview problems based on real data 1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with. Try 250 free problems /problems