Ramp hits $60B valuation on AI token spend management

Ramp raised $1.85 billion at a $60 billion valuation, and the products it pitched alongside the round are about one thing: your AI bill. The announcement on October 15, 2026 put AI token spend management and model routing next to the corporate card. When a $60 billion finance company decides token costs are a category, token costs are now a line item every operator has to manage.

What actually happened

From Ramp's release:

• The round. $1.85 billion, led by Dragoneer Investment Group, with Thrive Capital, Founders Fund, Sands Capital and Avenir Growth co-leading. Total equity raised is now about $5 billion. Reuters, via KFGO, reported the valuation was $44 billion in June.
• The business. Over $2 billion in annualized revenue, positive cash flow, more than 85,000 customers and $275 billion in annualized purchase volume, up 175% year over year.
• AI Token Spend Management. One view of AI spend across providers, teams, models, users and API keys. AngelList's controller is quoted saying it cut their monthly AI spend by 17.5%.
• Router.com. One endpoint that routes requests across model providers by cost and performance. Ramp claims it cuts inference costs by up to 40%.

Those savings numbers are Ramp's claim and customer quotes. The release gives no method. Treat them as a sales pitch until you test them on your own traffic.

Why AI token spend management matters for your business

Most small teams cannot answer a simple question: who spent what on AI last month, and on which job? The bill shows up as one OpenAI or Anthropic invoice, plus a few seat licenses on cards nobody tracks. That is how a stuck agent loop burns $900 over a weekend.

You do not need a $60 billion vendor to fix it. You need three things:

• One key per job. Separate API keys for the support bot, the product description generator and the dev team. Spend reports by key then tell you which workflow costs money.
• Hard limits. Set monthly caps per key at the provider. An agent that hits a cap fails loud instead of failing expensive.
• A thin routing layer you own. Send simple tasks to cheap models and hard ones to frontier models, through one internal endpoint. If you later buy a router like Ramp's, you swap the backend, not every app.

If you already run Ramp, the token tool may be the cheapest way in. Just keep the routing logic portable. A router owned by your expense vendor is still a vendor.

Sources: Ramp via PR Newswire, Reuters via KFGO.