BAT的「AI云销售」，困在MaaS内卷战中

TL;DR - Chinese cloud vendors’ push to rapidly grow model-as-a-service revenue has trapped AI sales teams in aggressive quotas, commoditized model resale, and escalating token-price competition. The report matters because it shows that model quality, low switching costs, and internal delivery capacity—not traditional enterprise-sales relationships—now determine MaaS competitiveness.

• Demand concentrates around whichever models lead specific use cases, such as GLM for coding and Seedance for multimodal generation, making weaker proprietary models difficult to sell.
• Major cloud providers increasingly credit third-party model resale toward sales targets, reducing differentiation as vendors offer the same popular models.
• Near-zero switching costs for token APIs fuel discounting, cloud-credit subsidies, and weak customer retention; cheaper Codex subscriptions add further pricing pressure.
• MaaS deals require coordinated architecture, adaptation, deployment, and discount approval, but conflicting internal KPIs often obstruct sales and delivery.