Ramp at $60B: From Expense Tracking to AI Agent Finance
Ramp has raised $1.85 billion at a $60 billion valuation as it pushes deeper into AI. The round was disclosed on Oct. 15, four months after the company raised $750 million at a $44 billion valuation. TechCrunch
The financing was co-led by Dragoneer Investment Group, Thrive Capital, Founders Fund, Sands Capital and Avenir Growth. Ramp said total funding now stands at about $5 billion.
In my view, capital is not the constraint now. Execution is.
The numbers behind the step-up
Ramp said annualized revenue has crossed $2 billion, up from $1.5 billion in June 2026. Annualized means the pace of recent sales stretched as if held for a full year. Growth is 25% faster than around the same time last year. The company said it now facilitates over $275 billion in annualized purchase volume, up 175% from a year earlier. Customer count stands at 85,000, up from 70,000 in June. TechCrunch
Customers include Visa, Uber, Shopify, Robinhood, Anduril, Notion, Cursor and Figma. The group spans payments infrastructure, logistics, commerce, brokerage, defense technology and productivity software.
The broader context here is why that customer mix helps a spend platform. Transactions across categories, procurement cycles and company sizes provide more data for controls and automation.
In my view, the four-month cadence is unusual even for late-stage companies. Revenue moved from $1.5 billion to $2 billion annualized in that window. The company added 15,000 customers. Purchase volume kept climbing while growth accelerated.
From expense management to agent infrastructure
Ramp has expanded beyond expense management into payments, fraud detection, procurement, vendor management and accounting.
The company has embedded AI agents into procurement, expense management, accounting and budgeting products. Agents are software that can complete defined steps inside other tools. They operate inside approval, coding and reconciliation workflows where policy and spend controls already live. Coding here means assigning costs to the right budget category. Reconciliation means matching payments to records.
Ramp also launched a corporate credit card designed for AI agents to use, giving software actors a payment method under existing corporate controls. In August 2026, the company introduced Router, a tool for model routing. That means choosing which AI model handles each task based on cost, response time and quality. It then launched Glass, a workspace that collects and remembers AI automations, prompts and skills built by an organization's employees. TechCrunch
Ramp described the milestone in a blog post titled "Ramp at $60B: The End of the Beginning" at https://ramp.com/blog/ramp-at-60-billion. Ramp Bloomberg had reported in September 2026 that the company was in early talks for a new round at roughly a $60 billion valuation. Bloomberg
The broader context here is a shift from tracking spend to authorizing it in software. Expense tools began as systems of record. Procurement added workflow. Cards added a payment rail. Once agents can initiate purchases, contact vendors, code invoices and reconcile ledgers, the control point moves. It is no longer the monthly close. It is the moment of delegation, when a human or another agent grants authority to act and spend.
In my view, the questions for technology teams center on identity, policy and observability. An agent with a card acts as a new kind of user. It needs scoped credentials, spending limits, merchant controls and a complete audit trail, a step-by-step record. Model routing needs choices per task against latency, cost and quality rather than per vendor contract. Glass addresses how prompt libraries, skills and automations persist when employees change roles or leave.
In my view, the valuation math points more to platform logic than to card economics. Purchase volume at $275 billion annualized and revenue at $2 billion annualized point to fees on flow plus software subscriptions, with growth still accelerating. The move from 70,000 to 85,000 customers in four months points to broader adoption across mid-market and enterprise, not only deeper use inside existing accounts.
Worth flagging, that same breadth raises the bar for reliability, access controls and reconciliation accuracy. A procurement agent that misclassifies spend creates extra work. One that commits spend incorrectly creates an incident.
In my view, what this could enable if execution holds is a finance stack where policy checks, payment and ledger updates happen in the same loop. That has been the long promise of autonomous back-office software. Credentialed agent payments, embedded workflow agents, routing across models and a shared memory layer map closely to what practitioners say has been missing. The work ahead is less about model capability and more about trust boundaries, exception handling and clean links into ERP systems and data warehouses.
