Kong's Enterprise AI Control Plane
Augusto Marietti, CEO of Kong, on the end of tokenmaxxing
This is a systems integration sale disguised as an AI infrastructure product. Kong is not mainly selling cheaper prompts, it is selling a control point that sits between many internal apps and many models, where a large company can cap token spend, block sensitive data, cache repeat requests, and route simple tasks to cheaper models, which makes the product fit big enterprise budgets and longer, higher value deployments.
-
Kong’s AI gateway extends its API gateway playbook into LLM traffic. The product layer is concrete, rate limiting, token aware controls, and semantic cache, which means a central platform team can enforce rules once instead of every app team wiring model policies by hand.
-
The contrast with OpenRouter is business model, not just product shape. OpenRouter monetizes routing and access across hundreds of models with a 5.5% platform fee on pay as you go and a 5% BYOK fee after the free threshold, which is closer to brokerage economics than enterprise software.
-
The Palantir comparison points to how these deals get won. Large enterprises often need deployment help and deep workflow integration, and Palantir’s model is built around forward deployed engineers, which is why AI infrastructure sold into the Global 2000 tends to support bigger contracts than developer tools that spread through social buzz.
This market is heading toward a split. Lightweight routing will keep looking like a low margin marketplace, while behind the firewall AI control planes will become part of the core enterprise stack, alongside API management, identity, and data governance. That favors vendors like Kong that already own production traffic and enterprise procurement paths.