Tech Hiring Is Still Frozen. Here’s the Real Reason Why.

A quick glance at the TechCrunch 2025 tech company layoff tracker shows that massive firings across the tech sector have not come close to letting up.

To date, just in 2025, nearly 90,000 layoffs have made the list, and these are coming from just the major tech companies we’ve all heard of. For example, xAI, Rivian, and Oracle are September’s most current “winners.” 

The 2025 layoffs sit on top of 2024’s 150,000 cuts, according to the same list, and that’s after “significant cuts” in 2023 and 2022. Let’s use a nice, round, conservative number, and say there’s been about 300,000 tech jobs cut. 

So quick question: Was there over 300,000 workers worth of fat to be trimmed in the tech industry?

OK, kind of a trick question. I know for a fact that exactly 300,000 of you will likely say “Hell, no!”

But even the most cynical among us have to raise an eyebrow at that large a number. I suspect that even if all these tech companies got profitable – and many of them say they did – they’d have a really hard time growing again without hiring up, at least a little.

That’s a problem that’s only going to snowball. So after four years of massive resource cuts, why is tech hiring still frozen? 

I have a one-word answer. At the end.

Truth: You Can’t Cut Your Way To Growth

Speaking of cynicism, any time you give a short-sighted company a choice: Either cut a less-experienced, not as good, cheaper resource or a more experienced, very good, expensive resource, that company will choose to cut the latter, especially in tech. 

This is for a couple reasons.

1. It’s the unfortunate nature of the tech worker beast. Despite conventional wisdom, becoming a tech worker has a very low barrier to entry. In fact, as time goes on and tech evolves, that barrier drops even lower. Remember the “Learn to Code” movement?
2. For reasons I’ve never understood in 30-plus years of doing tech, the tech company SOP has always been to put as much of the tech department in a black box that gets fed instructions and returns apps. And for reasons I do indeed understand but don’t condone, most every tech worker is fine with this. Which leads to a hard ceiling of value as their salaries continue to increase.

What does this have to do with growth?

About 10 years ago, one of my mentors hit me with this quote:

“You can either grow, or you can make money. One or the other. Not both.”

As the unstoppable force of AI labor replacement met the immovable object of mandated board calls to make more money, cutting experienced tech workers and replacing them with chatbot-armed junior counterparts became a “two-fer.” The company was able to slash the holy hell out of their tech labor costs while also leaning into AI to boost productivity by a seemingly exponential factor.

Oh-ho-ho no. That did not work.

The Shine is Off Of AI-Enhanced Productivity Estimates

Now, the math isn’t adding up. My favorite CTO once told me:

“One great developer = 2.5 average developers. And one average developer = 2.5 junior developers.”

Look, I’m not going to ask you to commit to anecdotal math that one great developer is equal to 6.25 junior developers. But if you’re in tech, in your heart of hearts, you know he’s right. And I’ve found this to be true across the tech spectrum, from QA engineers to database administrators to project managers. 

But even if you don’t subscribe to the ratio, you have to admit that no one is fooling anyone anymore with ChatGPT as a productive 1:1 replacement for tech labor. The reason last month’s MIT study stating that 95 percent of AI projects haven’t produced any ROI shook every tech company leader to their core is because it was a big statement from a credible source that simply shouted something everyone already knew.

Prompting Skills Aren’t Skills

Let’s do a little more math. 

One junior developer plus AI equals… one junior developer.

Crap. Might as well fire her too.

And here we are.

But why?

I’ve had my hands in the guts of AI, including some of the first generative AI, for over 15 years. I was the source of what we were building on the data side, what today you could call “small language models,” but we were even too early to call them that.

So when I first heard about “prompting skills,” I immediately started calling them bullshit. 

To put it as simply as possible, the person with the best “prompting skills” is going to be the person with the most intimate knowledge of the data, the context of the prompt, and the desired accuracy of the result. It’s about knowing what to ask for, what to negatively prompt against, and what guardrails need to be in place. Not to be “the most convincing” or whatever “prompting skills” were being sold as.

In other words, if you put a software developer in front of an AI code assistant, the more experienced that developer, the more productivity they are going to get using the assistant. If you give a junior developer that same tool, they will struggle. If you tell a newbie to “vibe code,” they will take down your system.

And this effect isn’t just in software development or even tech, it’s inherent in every generative AI use case. Everyone who was developing or selling or promoting AI knew this. But the lure of selling 10x or 20x productivity gains, with “no skills necessary”, was too strong to ignore. 

Thus, one junior tech worker plus ChatGPT equals one junior tech worker. Or even 0.8 junior tech workers because they’re also dealing with hallucinations and not experienced enough to catch them.

Tech Hiring Is Frozen Because It’s Paralyzed

Paralysis. That’s the one-word reason why those 300,000 tech jobs, fat or not, aren’t being rehired. 

“It’s frustrating as hell,” said another friend who is a VP of Product and, not uncoincidentally, is leading a team in a black box of junior product managers and junior engineers. “Everyone is overworked. We need help, but the ELT hasn’t figured out what kind of help. So we’re screwed until they figure that out.”

And so if I may speculate, the delay is because it’s really, really hard for an executive leadership team to admit they were wrong. Especially when they’re wrong about promises they were sold, not facts they independently verified. I can’t blame them, because what got sold as “AI” is a new science, it would take months before any sort of verification could be done, but to justify the spend, they had to cut expenses somewhere else.

Like I said, when faced with that decision, they will almost always cut the more experienced, very good, expensive resource. Some tech companies are realizing that now, and tech hiring might be ticking up, almost imperceptibly, but with 300,000 jobs to go, it’ll take a lot more companies to admit their mistakes before we get those lean, experienced, very good tech resources back in place to start fueling growth and unfreezing the hiring cycle.

If you like a little reckless speculation with your tech, please join my email list and get a heads up whenever I stick my neck out.