The tech-job slump represents a mix of cyclical economic forces and structural changes from AI. What’s happening in tech is a preview of what’s ahead for many professions.
A few years ago, software engineering was the hot job everyone wanted. Tech execs were suggesting that everyone should learn how to code. Today it feels like those jobs are gone forever, as exemplified by this recent NY Times article, “Goodbye, $165,000 Tech Jobs. Student Coders Seek Work at Chipotle.” Is this the beginning of the end of tech jobs? Who is at risk next?
Let’s recall that this is not the first time we’ve seen such dire circumstances. MIT, arguably the top computer science school in the world, had about a quarter of its undergrads majoring in Electrical Engineering / Computer Science at the height of the dot com boom. (MIT’s undergraduate enrollment was 4,372 in 1999; with 1025 of them in EE/CS). After the dot com crash in 2000 the number dropped 21%, with only 806 in that major in 2002. This happened because many of those excited about the job opportunities seen the previous years choose to look elsewhere, thinking the bubble burst. That number then climbed back up as the economy and tech industry recovered and boomed again. For 2021 it was 1,479 students, nearly one-third of the undergrads, either majoring in Electrical Engineering / Computer Science alone or one of the new combined majors the school offered (basically computer science along with another discipline like biology, cognitive science, or economics). In other words, we’ve seen cyclical demand in the field, as evidenced by interest in students’ choice of majors. (Ideally, students should choose majors independent of current cyclical market conditions, but the reality if they are very heavily influenced by such factors.)
The past few years tech jobs have faced a confluence of challenges. Some of these are transitory, others are more structural. While some of the causes may follow a standard boom-bust pattern, the structural changes, primarily from AI, will have a long-term impact.
Five Economic Factors Impacting the Tech Labor Market
There are five key economic factors impacting tech hiring today. One is specific to tech, but the other four are not. They divide into a hiring correction, monetary policy, and policy changes.
First, there was over-hiring. During the covid pandemic, companies raced to develop new software. Much was to support the remote-first world we were living in at work. Some demand came from new functionality, e.g., better video conferencing support; companies like Zoom couldn't hire fast enough. In other cases, it meant expanded or shifted workloads; Amazon and Instacart saw huge growth in orders, for which additional software functionality would yield efficiency gains and the opportunity for new market share. Core tech companies, like Google and Microsoft, just saw a longer-term shift to more technology use. In all cases, companies raced to hire more software engineers (as well as related people like QA, product management, marketing, etc.).
When the world opened back up these companies realized they overinvested in tech workers. We saw a huge number of tech layoffs in 2023 and 2024. A few hundred thousand layoffs over two years (slightly more since they started at the end of 2022 and continue through today) isn’t significant for the economy as a whole, which is why unemployment overall in the US remains around 4%. However, it is significant for the tech sector. The labor supply of quality tech workers grew significantly. This factor, unique to tech, is one of the reasons it has suffered more than other fields. While hiring corrections often follow a classic boom-bust cycle, it's not the only factor.
Second, interest rates spiked. VCs and smaller PE firms, the ones who invest in early and later stage startups had a limited supply of money. Many couldn’t raise new funds and consequently held on to their existing capital, not wanting to hold a cashless fund without a new fund committed. Their investors, family offices and large institutions, saw a fed rate of 5%. A AA corporate bond (a very low risk asset) was yielding around 5% the past few years. In other words, the near-risk free return was 5% in a liquid asset. VC funds had been dropping their IRR in recent years to around 12%. Heck, the stock market returns 10% annually (and that’s for mere mortals like you and I who don’t have expert money managers or access to hedge funds and tax loopholes). Why lock in millions of dollars for 7-12 years in a VC or PE fund when you can keep it in the market and do just about as well with much better liquidity?
While VC money did go heavily into AI investments the past few years (53% of all investment in the first half of 2025) the oversized valuations typically mean fewer jobs per invested dollars. Further in the same article we can see the number of investments in pre-seed and seed stage companies has been dropping.
The high interest rates have been choking off the supply of money for startup company formation and growth. Some of the biggest drivers of tech jobs are startup companies. Startup companies typically begin by hiring engineers in the earlier rounds to build the software and then hire up sales and marketing in the later rounds once the software is built and they shift their focus to revenue growth. PE companies invest in established businesses but often use software to reduce costs and improve margins. So, in addition to the excess supply of tech labor, there’s also a reduction in demand stemming from reduced VC and PE investment. Next comes the political headwinds.
Third, amplifying the effects of reduced tech investing, there’s been macroeconomic uncertainty, which companies hate. Throughout 2024, as is often the case in presidential election years, companies held off on strategic investments until they saw who won the election and what economic policy will be. The expectation was that post the 2024-election there would be clarity (and that interest rates would drop). Unfortunately, there’s been anything but clarity. Instead, we have chaotic, mercurial policies of questionable impact whose legality is dubious (at the time of this writing, the US Supreme Court is considering arguments about the legality of tariffs). Companies can’t operate under such conditions.
Fourth, the “America first” isolationism and political volatility have hurt foreign investment in the US. For example, the US courted South Korean company Hyundai Motor Group to build more factories in the US; then ICE raided the factory while under construction. Later, the US allowed the detained workers to choose to stay in the US or leave; most left. According to the same article, South Korean President Lee Jae Myung said South Korean companies are now less likely to invest in the US. I’m sure they’re not the only ones. (For non-US readers, remember the old economic adage, “When the US sneezes, the rest of the world catches a cold.”)
Finally, there’s been general economic folly. Immigration policy has hurt farms and factories alike. Tariff roulette makes it hard for companies to financially plan long term. Other policies create similar chaos. While the companies directly impacted by some of these policies aren’t directly large software consumers themselves (factories may be, but less so those who use immigrant workers), their economic pains ripple through the industry food chains.
The Impact of AI on Labor
The above macroeconomic causes are ones we’ve seen in the past. But now we’re starting to really see the impact of AI. The prior factors will change with economic cycles and elections. The impact from AI is structural and long-lasting.
While AI garnered attention in the headlines, through 2024 AI hadn’t been a big factor in terms of its impact on the labor market. Companies talked a good game, and they explored it, but the actual use of AI was very limited. For example, in software, the DORA report Impact of Generative AI in Software Development found limited use and upside as of 2024. The BBC article The AI job cuts are here - or are they? suggests similar skepticism about AI-motivated layoffs to date.
But that’s the bad news for those struggling to find jobs. AIs impact on the labor market had been muted in the past, but that changed in 2025, which will make job hunting even harder. We’re going to see an impact similar to how automation upended the manufacturing labor market. 87% of the manufacturing jobs lost were due to technology (not due to outsourcing as many politicians falsely claimed for political points) and it’s possible we’ll see AI automation on the scale.
Unlike manufacturing automation, this will affect both skilled and unskilled labor alike. Both software developers and call centers agents are some of the first roles affected. It might seem surprising because we normally think of software development jobs as requiring much more training and knowledge than someone in a call center. It does, but that knowledge isn’t defensible.
What factory workers did 100 years ago is akin to what software engineers do today, so the patterns can be revealing. Back then cars were relatively simple, but it was still a lot of work to build one. Even after the engine was built, which was the innovation, the rest of the car had to be built. This included seats, wheels, steering, chassis, etc. It’s not hard to add seats, but it involved more physical work. Most of the work of the car wasn’t the “complex” engine, but the rest of the car. Ford’s assembly line simplified the work. Much of the work software engineers do isn’t thinking of a clever way to code something, that’s maybe 5-15% of the job (and often less). Instead, most of the time coding is about hooking up different existing systems (using APIs), getting the configurations right, figuring out what didn’t get “wired” up correctly, etc. It’s a lot of “installing seating” to continue the car analogy. AI is great at that.
AI won’t eliminate all the jobs, but it can impact many by automating repetitive tasks. In “Why AI Isn’t (Yet) Ready to Take Your Job” I go into detail about how to understand how much of your job function AI can and cannot replicate. In short, it won’t take your whole job, but it will start to remove some of the grunt work. Consider the time it takes to schedule a meeting and send an agenda, or to send a status update email. For some people that’s a few minutes a month, for others it may be tens of minutes a day. When that time drops to zero it frees you up to do other work. If enough people free up enough time, then you don’t need as many people. Scheduling meetings and sending status emails alone won’t do it, but AI can do much more than that. Understand this isn’t about AI being brilliant or insightful, it’s about AI simply automating grunt work. And we’ve only scratched the surface with agentic AI and industry specific solutions.
It’s not just tech and call centers, but many other professions, too. Anything text based is at high risk. Software is a lot of text (albeit with specialized grammar). Call centers are spoken text. Medical claims and insurance, lots of text heavy documentation.
Finance and accounting involve numbers where correct math matters, and AI hallucinations are very problematic. However, we already have software that can do the math for us; that was built over the last few decades. While reconciliation may not be a task for LLMs, other types of auditing and financial analysis may be.
Creative work is obviously at-risk. (On a personal note, while I write my own articles and primarily use spell checks or occasionally AI to see if I missed something, for well over a year all my blog images are generated by AI.) But again, as noted in “Why AI Isn’t (Yet) Ready to Take Your Job,” it only removes some types of creative work, not all of it.
Law and medicine are interesting cases. Both fields are text heavy, but they also demand complete accuracy, so humans will always need to be in loop for the final review. AI will reduce headcount, but perhaps not as much as in other fields.
Before concluding this section, it should be noted that there has been some pushback in 2025 on just how productive AI will be. In the past few months people have raised issues like “workslop” (such as this CNBC article). It notes that the AI output is a lot of fluff, and while bad work can be easy to spot, workslop looks good as first pass, but on closer inspection has issues. While workslop is a true problem, and may give some false productivity gains, it doesn’t mean AI has no productivity gains at all. (Likewise, overeager execs may see AI work well on small pilot projects and then will not understand why, when used on projects at scale, the productivity won’t be as good.) AI will reduce some demand for tech jobs and other jobs stemming from the true (and not just perceived and overhyped) productivity gains. Jobs will be at risk unless Jevons paradox kicks in as it has in the past for technology.
Eventually things will revert to the mean. Interest rates have already started to come down. Eventually the US will elect less mercurial officials, and potentially even ones who value foreign investment or understand basic economics. Eventually the laid off tech workers will be reabsorbed into the workforce, possibly in different roles, and the impact of AI will start to be understood. But if this will be closer to six months or six years for any of these dimensions is anybody’s guess (although I’m pretty sure it won’t be six months). Tech is just the tip of the iceberg. As AI matures and AI enabled tools expand into other verticals, the same risks apply. It’s just a question of how much of the work is “installing seating” and how much of the other macroeconomic issues still apply as AI enters other verticals.
It’s critical to learn about corporate culture before you accept a job offer but it can be awkward to raise such questions. Learn what to ask and how to ask it to avoid landing yourself in a bad situation.
Investing just a few hours per year will help you focus and advance in your career.
Groups with a high barrier to entry and high trust are often the most valuable groups to join.
