The Signal · Column 100

When the chipmaker cosigns the loan, talent becomes the constraint

Andrei van Noordt · 31 October 2026 · 2 min read

Nscale wants to build an AI data center in West Virginia, and the financing has taken an unusual turn. According to TechCrunch, Nvidia is weighing whether to cosign the loan. The company that sells the chips may stand behind the debt used to buy them. When investors are nervous about how much money is going into AI infrastructure, the supplier's signature on the guarantee is reassuring.

Most coverage will treat this as a story about capital and circular financing. I read it as a sign of where the real risk now sits. If the chipmaker will take on part of the credit risk, money and hardware are no longer the scarce inputs. Lenders are really worried about whether a facility like this ever earns back its cost. That depends on things no guarantee can cover: whether it gets built on time, whether it runs reliably, and whether it fills with workloads that pay.

When the supplier is willing to cosign the loan, capital stops being the constraint and competence becomes it.

That shifts the pressure onto people. Every facility financed this way needs a small group who can turn a building full of accelerators into dependable, billable capacity. That means infrastructure engineers who understand power and cooling at scale, platform teams who keep utilisation high, and leaders who can prove the return to the people holding the debt. These people were already hard to find before this build cycle started. Every new project draws on the same short list, and the companies financing the hardware are not financing the talent.

There is a second effect for everyone else. As lenders and suppliers finance infrastructure more aggressively, compute gets easier to buy and harder to compete on. The advantage moves to the teams that use it well. That means engineers who know how to get value out of models in production, not engineers who only know how to ask for more GPUs.

For hiring leaders, the practical read is simple. Assume the capacity will be there, and treat the people who make it pay off as your scarce asset. Hire the infrastructure and applied AI engineers who can show a return on compute now, before the next wave of financed data centers starts bidding for the same names.