This week Eric Glyman and the team at Ramp raised $750 million at a $44 billion valuation, and they did it on the back of one of the cleanest pieces of category thinking I’ve seen in a long time.
Their frame is this: for five hundred years, business ran on two pillars of spend. People, and vendors. Then, in the span of roughly two years, a third arrived — intelligence, paid by the token, invisible to every financial system we’d ever built to manage cost. Ramp is making that third pillar visible. They’re right that it exists. They’re right that it’s the fastest-growing line item in business. And they’re right that almost nobody can see it clearly.
I want to build on that frame, because I think it’s correct — and incomplete in a way that matters enormously to anyone who runs an engineering organization.
Every pillar of spend has two sides
Here’s the thing about a pillar of spend. It’s never just a cost. It’s a cost and a return, and a mature organization tracks both.
We’ve always done this with the first two pillars. People show up as wages on one side of the ledger and as performance on the other — and no serious company manages headcount by staring at payroll alone. Vendors show up as contracts on one side and deliverables on the other. We negotiate the price, then we hold the work accountable to it. Cost and return. Both halves. Always.
The third pillar arrived so fast that we’ve only built the first half. Ramp is building the cost side, and building it well. The token bill is becoming visible. What it cost is becoming knowable.
But knowing what AI costs is not the same as knowing whether it worked.
That second half — the return side of the third pillar — is still dark. And it’s darkest exactly where the tokens are being spent first and hardest: inside engineering.
Tokens hit code first
If you want to know where the third pillar is growing fastest, don’t start in finance. Start in your codebase.
Agentic development is where token spend compounds quickest, because it’s where the work itself is being rewritten. A developer who has internalized agentic tooling can compress weeks of engineering into hours. That same developer is consuming tokens at a rate that would have been unthinkable eighteen months ago. The spend and the leverage live in the same place. So does the blindness. Uber recently capped its people at $1,500 a month each on AI tools — after burning through its entire 2026 AI budget in four months.
Ask most engineering leaders three questions and you’ll see the gap. Which of your developers are getting frontier-level results from AI, and which are simply generating frontier-level bills? Are your top performers’ workflows reaching anyone else, or dying on their laptops? Is your AI spend turning into shipped software, or into activity? Most can’t answer with confidence. They have a number from a provider and a feeling in their gut, and nothing in between.
That gap is what we built CodeVine to close. It’s the difference between teams that are AI-active and teams that are AI-productive. One is a line item. The other is an outcome.
Cost visibility is table stakes. Return visibility is the moat.
None of this diminishes what Ramp is doing. The opposite. Cost visibility is the necessary first move, and a $44 billion valuation is the market agreeing — loudly — that the third pillar is real. When the most sophisticated buyers in the world start demanding to see what their AI costs, they’re about ninety seconds away from demanding to know whether it was worth it.
That second question is the one a spend report can never answer. A dashboard can tell a CFO the bill tripled last quarter. It can’t tell a VP of Engineering whether the tripled bill bought ten times the output or ten times the noise. Cost lives in finance. Return lives in the work — in commits, in pull requests, in the velocity of the team and the knowledge it compounds sprint over sprint.
That’s why CodeVine sits at the organizational layer rather than the financial one. We Capture what your developers are actually doing with AI. We Correlate that activity to real engineering output, so spend maps to results instead of to vibes. And we Compound it — the Grafting Engine takes what your best developers figure out and deploys it to everyone else, so the return on the third pillar doesn’t just become visible. It accelerates.
The market validated the question. We built the second answer.
Glyman is right: every company now needs infrastructure for the third pillar. I’d only add one sentence to it.
The companies that win the AI era won’t be the ones who could finally see what their intelligence cost. They’ll be the ones who could prove it was working — and then made it work better, automatically, every single sprint.
Ramp is building the infrastructure to see the bill.
We’re building the infrastructure to make the bill worth paying.
Same pillar. The other half.
— Wells Burke