New data from Databricks’ AI Index highlights striking trends in enterprise AI usage, including a rapid rise in open-weight model traffic to 31%, 72% of companies using multiple AI providers, and a vast cost disparity between average and top users.
- Open-weight model token share jumped from 7.3% to 30.9% in four months
- 72% of enterprises now deploy multiple AI model providers simultaneously
- Median AI token spend doubled recently, but top users spend 281x more
What happened
The Databricks AI Index tracked token usage from over 10,000 companies routing AI through its Unity Gateway, revealing that open-weight models' share of enterprise AI token traffic rose sharply from 7.3% in July to nearly 31% by November 2026. This surge reflects enterprises shifting bulk operations like classification and summarization to more cost-efficient open-weight models, rather than swapping out proprietary models for their toughest tasks.
Additionally, enterprises have dramatically changed their supplier mixes; the proportion relying on a single AI provider dropped from 64% to 28% in one quarter. The average number of unique models used per customer increased from 4.2 to 7.1 since January, indicating a multi-provider approach where workloads are routed to the best-suited model.
Why it matters
This shift has significant implications for SaaS and AI product builders: with roughly one-third of enterprise token traffic now on cheaper, open-weight models, companies that continue to route all AI calls to premium frontier models risk losing gross margin advantages. Enterprises are optimizing costs by reserving expensive models for high-value use cases and offloading routine, high-volume tasks to cost-efficient open-weight models.
The data also reveals that enterprises move quickly to adopt superior proprietary models when they provide meaningful value. For example, Anthropic’s Claude Opus 5.5 captured 38% of proprietary-model tokens within six weeks of launch, while older but more capable models saw rapid decline in usage. This dynamic underscores that cost-efficiency per completed task—not just per token pricing—is the critical factor guiding enterprise AI adoption.
What to watch next
The ongoing democratization and fragmentation of AI model usage could accelerate innovation and pricing competition across providers. Monitoring how model performance and integration ease influence enterprise mix will be key as companies increasingly treat AI models like cloud regions—deploying multiple options tailored to specific needs and budgets.
Another critical development to track is the widening spending gap among users. Median AI token spending more than doubled recently to $2.49, yet the top 1% of users spend nearly $700 weekly—about 281 times the median. This disparity highlights a power-user segment driving outsized AI consumption and costs, presenting opportunities and challenges for pricing, capacity planning, and service differentiation.