And yet, in the same window, my 50-person consulting firm hired its first full-stack software developer in January. We hired a second in April. We’re likely to hire a third before the year is out. Big Tech is right that AI is rewriting the developer profession. We’re right to be hiring into it. Both things are true — and the gap between them is the most important workforce story no one is telling.

I have done this before

Sixteen years ago at Microsoft, I helped build a product that severely threatened an entire job category. Power BI was a direct challenge to the elite priesthood of Business Intelligence developers – highly-paid professionals who’d spent years learning a deeply technical craft. Our explicit goal was to unlock professional-grade BI for a much broader audience. Good for the world. Bad, on paper, for the priesthood.

And the product did what it was designed to do. It sidestepped the bottleneck. It ushered in an era of information abundance. It put real BI within reach of people who couldn’t spell SQL.

But the priesthood didn’t disappear. They got busier. A profession formerly numbered in the thousands now numbers in the millions – and the original specialists, the ones who actually understood the hard parts, found themselves more in demand than ever. The total surface area of BI in the world grew by orders of magnitude. Even if most of that new work was beneath their pay grade, the number of complex projects also exploded. Elite professionals were the natural choice for those.

When something becomes cheaper, the world makes more of it. That is the only sentence in this essay that I am sure of, and it is the one that matters.

The math that flipped at our firm

Since P3 Adaptive’s founding in 2013, we’ve been a data-and-dashboards firm, staffed almost entirely with the new breed of BI professional that Power BI made possible. We never hired traditional software developers. The ROI on a so-called full-stack engineer – high salary, narrow problem set, long ramp – simply did not pencil out for a firm our size. AI has reset that calculation. Independent research from METR and Anthropic puts the typical productivity uplift for AI-assisted developers at roughly three times – with task-specific spikes far higher. That is the difference between hiring a developer who can take on one project at a time and hiring a developer who can take on three. It didn’t make sense for us to staff a three-person engineering team. Hiring one developer who can do the work of three, with one salary and one manager? That penciled.

The result is that we’re no longer “just” a data-and-dashboards firm. We embed capabilities directly into our clients’ line-of-business systems instead of standing them up in parallel. We build custom web portals, mobile apps, and AI platforms on top of the Microsoft stack rather than waiting for Microsoft to ship them. We had these ambitions for most of the company’s existence; the economics simply didn’t allow them. They do now.

We are, in other words, exactly the kind of buyer that the developer profession is migrating toward.

Three things are happening at once

The headlines are capturing one of them. They are missing the other two.

First, large software teams will continue to shrink. Big software companies are the traditional employers of developers, and they have to absorb the new economics first. A 2-to-3x productivity gain means a team of ten can ship what a team of twenty used to ship. The payroll savings are obvious; the reduction in coordination overhead may matter more. Big software is going to keep right-sizing. The headlines will keep coming.

Second, a wave of new buyers is entering the market. Companies like ours are the very tip of an iceberg. As more organizations – from SMBs to individual Fortune 500 departments – discover that one developer plus AI is now an in-house software team, the demand side is going to expand at a pace that has no precedent. Total developer employment may very plausibly land higher than its pre-AI peak.

Third, the volume of software in the world is going to explode. The big platforms will continue to ship at roughly the current rate. But the share of the world’s software that is built close to the business — custom-fit to a single company’s workflows and data – is going to dwarf everything that came before. Most software ever written will be written after 2026, and most of that will be written for an audience of one company.

For the next few quarters, the first trend is going to outrun the second. Cutting headcount is a quarterly story that boards understand; standing up brand-new capabilities is a multi-year story they don’t. So we will probably see a downturn in overall developer employment before we see the rebound. The Stanford data on 22-to-25-year-old developers is the early signal.

The rebound is coming anyway. My background as a software product leader is the only reason we caught it early. That won’t be the case for long.

The priesthood is migrating

I helped obsolete an elite profession once. It promptly grew a hundredfold. The members of the old priesthood weren’t put out of work – they became the people you called when the new tools weren’t enough, which turned out to be more frequently than ever.

The developers being laid off this quarter aren’t the end of their craft. They’re the priesthood, watching the gates open. When something becomes cheaper, the world makes more of it. We are going to make a lot more software.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.