Three Distinct AI Gateway Markets
Augusto Marietti, CEO of Kong, on the end of tokenmaxxing
This is less one AI gateway market than three different businesses that happen to sit in the same request path. Kong is selling an internal control layer to CIOs and security teams, where the buyer cares about cost caps, audit logs, PII stripping, and routing employee traffic across many models. Public gateways like Vercel and Cloudflare sell developer convenience, while OpenRouter sells model access and price arbitrage more like a marketplace broker.
-
The enterprise workflow is operational, not just developer facing. A company can send every employee prompt through one checkpoint that compresses prompts, caches repeated queries, strips sensitive data, assigns budgets, and routes simpler tasks to cheaper models before any token spend hits the model provider.
-
The sales motion changes because the economic buyer changes. Kong describes its AI gateway as a large enterprise infrastructure sale, with most demand coming from internal guardrails and cost control. Vercel and Cloudflare instead package one API key, routing, failover, and observability for teams shipping external AI features fast.
-
The business model changes too. OpenRouter monetizes routing as brokerage and pricing access, with fees layered on top of model usage. Kong is monetizing governed infrastructure behind the firewall, which makes it closer to security and systems management software than to an LLM reseller.
The category is heading toward clearer separation. Developer side gateways will compete on model breadth, uptime, and pricing, while enterprise gateways will expand into identity, policy, chargebacks, and agent traffic across internal systems. As AI use shifts from isolated chat to company wide workflows, the internal governance layer becomes the more durable control point.