By Geoff McDonald, CEO, Ambassador
A question I hadn't heard in years is back at the top of our buyer conversations. It isn't about features, integrations, or even price. It's this: will you still be here in three years?
We're hearing it from prospects and clients alike, at companies of every size. And it comes from the people at the very top of the decision. When you're the one signing, this is what keeps you up at night. Before they commit, they want to know how a vendor is funded, who it owes, and what happens to them if it stumbles. The balance sheet has quietly become part of the product.
Look at what happened in AI infrastructure over a few days in October.
On October 14, CoreWeave shares fell more than 15% as yields on some of its bonds climbed to around 15%, a level the market usually saves for companies it is worried about. The company reportedly carried about $52 billion in debt and lease obligations as of June, and its CEO told staff the contracted backlog more than covers them. A yield at that level doesn't mean a default is coming, but it does tell you what lenders now charge for uncertainty.
The same day, Gartner forecast that more than 80% of neocloud providers will stop operating independently by 2030, and that over $3.5 trillion of AI infrastructure spending could be exposed to impairment by the end of the decade if the current pace holds.
Then PitchBook's LCD desk reported where the money is moving. Institutional leveraged loans backing data center infrastructure reached $16.5 billion so far this year, up from $2 billion in 2025. High-yield bond issuance for data center expansion hit $50.1 billion, against $13.6 billion for all of last year. And lenders are asking for collateral, increasingly the GPUs themselves.
When lenders want the chips as security, the people closest to the money are telling you exactly how sure they are.
All of this is landing as rates move the wrong way for borrowers. The Fed raised rates in September for the first time in more than three years, and the minutes show most policymakers expected another increase before year end.
That changes the math for every company carrying heavy debt or a valuation it can't grow into. Some will find a home on someone else's terms. Some will shut down. Others will keep the lights on by cutting the things their customers depend on, like support, roadmap, and service.
The scenario that worries decision makers most is the sale under duress. A vendor that has to be bought, on a timeline set by its lenders, rarely gets to protect its customers in the deal. Pricing resets at renewal. The roadmap stalls. The people you trusted leave. Your contract now belongs to an owner you never chose.
And a big funding round isn't automatically good news anymore. Every raise sets a valuation the company now has to justify, and stakeholders on the buying side are starting to ask what happens if it can't. Fresh capital buys time. It doesn't answer the question.
It's tempting to file this under somebody else's problem. Most companies don't buy GPUs. They buy applications.
Every AI application runs on someone's compute, though, and plenty of application vendors have funded their growth with venture debt and the promise of the next round. Gartner expects less than half of AI revenue to be earned by providers that own significant infrastructure, because most AI-native companies run asset-light. So a lot of the money in AI flows through the application layer, which is exactly where a vendor's funding story is hardest to see from the outside.
Here's the diligence I'd run on any AI vendor right now, including us.
None of these require a finance degree. They require asking out loud the question buyers used to keep to themselves.
A system that runs your referrals, loyalty, retention, and measurement gets more valuable the longer it runs. Every program teaches it something about which customers stay, which ones refer, and what each outcome was worth. That history is the asset. Lose the vendor and you lose more than a tool. You lose years of what your customers taught you.
That's why I've argued the winning move is to out-compound your customers' alternatives instead of trapping them. The same logic runs the other way when you're the buyer. Compounding only works if the partner is still around to keep the record.
The AI buildout is real, and so is the bill for it. Over the next few years a lot of AI vendors will be absorbed, restructured, or gone, and the buyers who asked about the balance sheet up front won't be ripping out a system in the middle of a fiscal year.
Here's where we stand. We run Ambassador with a focus on being cash flow neutral, so our roadmap answers to our customers, not to a lender's calendar. We built it as one system across the whole customer lifecycle, so the record of what your customers do keeps getting more valuable the longer you run it. The programs we can attribute cleanly have driven more than $2.4 billion in attributed revenue to date.
Ask us the same five questions. You should be asking everyone.
The partners worth choosing are the ones still in the room when the results come in.
Grow. Keep. Prove.
Ambassador is The Customer Lifecycle Operating System, orchestrated by HiroAI. It is built so the record of what your customers do keeps compounding for as long as you run it. See how it works.