By the Verticalmove Team
In mid-2022, software engineering job postings hit 350% of pre-pandemic levels. Companies were hiring so aggressively that the tech sector's recruitment surge outpaced every other industry in the U.S. economy — including banking, which saw the second-largest boom. By early 2025, those same postings had collapsed to a five-year low: 65% of February 2020 levels, with 3.5 times fewer vacancies than the peak. Software development experienced the steepest hiring decline of any major employment category tracked by Indeed.
Then, something unexpected happened. As of early 2026, software engineering openings have surpassed 67,000 globally — a three-year high that roughly doubles the mid-2023 trough. Postings are up approximately 30% year-to-date. Citadel Securities reports that software engineer job postings are "rapidly rising," up 11% year over year. Tech recruiter openings — a reliable leading indicator of sustained hiring demand — are approaching 2022 peak levels. The conventional interpretation is straightforward: the market crashed, and now it's recovering. That interpretation is wrong. The market that crashed isn't coming back. What's emerging is structurally different — in who gets hired, what they're paid, what they're expected to know, and why companies need them at all.
To understand what's actually happening, you have to see the full arc.
The COVID-19 pandemic triggered an extraordinary hiring surge. As companies raced to digitize operations and expand online services, demand for software engineers exploded. Zero-interest-rate policy flooded the market with cheap capital, enabling startups and Big Tech alike to hire at unprecedented scale. U.S. software engineering and coding jobs, which had grown steadily from roughly 100,000 to nearly one million since the early 2000s, added another spike on top of an already plateauing trajectory. By the peak, open roles exceeded 100,000 at any given time. The era produced the hottest tech jobs market in history.
Then interest rates rose, and the whole structure inverted. Over 264,000 tech employees were laid off globally in 2023. Another 150,000-plus followed in 2024, and at least 127,000 U.S.-based tech workers lost their jobs in 2025. The cumulative toll from 2022 through early 2026, depending on which tracker you reference, exceeds half a million positions. The correction was not gentle, and it was not evenly distributed. Companies that had doubled engineering headcount in eighteen months cut 10-20% in a single quarter.
But here is where the data gets genuinely strange: the layoffs haven't stopped, and neither has the hiring. So far in 2026, tech companies have cut over 91,000 positions globally — approximately 926 per day, an accelerating daily rate. Amazon, which posted record revenue of $716.9 billion in 2025, accounted for more than half of all tracked tech layoffs in early 2026 while simultaneously expanding teams in AI and infrastructure. Meta cut roughly 1,500 employees from its Reality Labs division while growing its engineering headcount to 19% above January 2022 levels — the largest recovery of any Big Tech employer. Google's engineering workforce is up 16% over the same period. These companies are not recovering. They are reorganizing — shedding roles in one part of the org chart while building aggressively in another. The market isn't bouncing back to where it was. It's becoming something it has never been before.
The prevailing narrative around AI and software engineering is simple enough to fit on a slide: AI coding tools make engineers more productive, so companies will need fewer of them. It's intuitive. It's clean. And it fundamentally misreads how technological efficiency works.
In 1865, the British economist William Stanley Jevons observed that James Watt's dramatically more efficient steam engine — which used far less coal per unit of work — did not reduce Britain's coal consumption. It caused consumption to surge, because the efficiency gains made coal-powered industry viable in thousands of applications that previously couldn't justify the cost. Factories, railways, ships, and mills that would never have adopted steam at the old price suddenly could. Total demand didn't shrink. It detonated. This dynamic, now known as Jevons Paradox, has repeated with remarkable consistency across 160 years of technological change. Cheaper transistors did not mean fewer computers. Fuel-efficient aircraft did not mean fewer flights. And AI-assisted coding does not appear to mean fewer software engineers.
The data supports this. A randomized controlled trial at Google found that developers with AI tools finished tasks 21% faster. GitHub and Accenture measured a 55% faster completion rate across 4,800 developers on scoped programming tasks, with pull request cycle times dropping 75%. GitHub's CEO has reported that Copilot now writes an average of 46% of code in files where it's enabled, across more than 20 million users. McKinsey estimates AI could raise overall software engineering productivity by 20 to 45%. And yet, Gartner projects worldwide IT spending will reach $6.15 trillion in 2026, up 10.8% year over year, with AI as the dominant growth driver. Companies are not using these productivity gains to do the same work with fewer people. They are using them to do dramatically more work — building products, features, and internal systems that were previously uneconomical to attempt.
This is what we call the Efficiency Trap. The executives who interpret AI productivity gains as a mandate to cut engineering headcount are making a category error. They are optimizing for cost reduction when the structural opportunity is in expanded capacity. The companies that will win this cycle are not the ones doing the same work with fewer engineers. They are the ones recognizing that cheaper software development has expanded the universe of what's worth building — and staffing accordingly, with the right engineers at the right level.
There is, however, a critical nuance that the Jevons framing alone doesn't capture. AI doesn't just make software development faster. It changes which part of the engineer's job carries the most value. Atlassian's internal data shows that engineers spend only about 16% of their week actually writing code. The rest goes to architecture, coordination, debugging, testing, deployment, and documentation. AI tools are augmenting all of these — but they are augmenting them in a way that rewards depth of understanding, not surface-level proficiency. Consider AI's place in the history of software abstraction. Assembly language gave way to compiled languages. Compiled languages gave way to interpreted languages like Java and C#. Intelligent IDEs introduced code completion. LLMs are simply the next abstraction layer — a powerful one, but an abstraction layer nonetheless. And every previous layer raised the floor of what you could build without deep computer science knowledge, but none of them eliminated the need for engineers who understood what was happening beneath the surface. The engineers who believe they no longer need to understand data structures, algorithms, and design patterns because AI will handle it are not riding the wave of progress. They are being commoditized by it. The engineers who treat AI as a power tool built on top of deep fundamentals — the ones who can evaluate AI output, catch architectural flaws, and make judgment calls that no model can replicate — are the ones whose value is compounding.
The most dangerous assumption in engineering compensation right now is that "the market" is a single thing you can benchmark against. It isn't. The Bureau of Labor Statistics projects that employment for software developers will grow 15% from 2024 to 2034 — much faster than average — with roughly 129,200 annual openings and a median salary of $133,080. In the same projection set, the BLS forecasts that employment of "computer programmers" — the narrower, more routine coding role — will decline 6%. The median tells you nothing. The bifurcation tells you everything.
In our work, we are seeing compensation reductions of 10-15% for roles like platform engineering and data engineering, and reductions of 25% or more for commodity roles such as generalist full-stack and front-end positions. Signing bonuses have effectively vanished across these categories. Robert Half data confirms the broader trend: tech salary increases hit just 1.6% in 2025 — the lowest in at least fifteen years — and 66% of employers cite economic uncertainty as the reason for keeping budgets tight, up from 17% the prior year. Meanwhile, on the other side of the split, engineers building frontier AI models — large language models, multimodal systems, novel architectures — are commanding seven-figure total compensation packages. AI/ML roles saw 88% year-over-year hiring growth, with AI engineers earning a 12% salary premium over general software engineers according to Ravio's compensation data. LinkedIn's Emerging Jobs Report shows demand for AI-fluent software engineers surged nearly 60% year over year, with compensation premiums of 15-25% for developers proficient in AI frameworks. This is not a correction with a uniform direction. It is a market cleaving in two.
The visible applicant pool has never been larger. Yet the specific talent companies actually need — senior engineers with deep CS fundamentals, AI fluency, stable tenure, and the ability to own architectural decisions — is genuinely scarce. What's flooding the market is a different profile entirely.
We are seeing a significant influx of software engineers requiring immigration assistance, including H-1B transfers, at a moment when federal policy has made sponsorship materially more expensive and administratively burdensome. The current administration's public tightening of work visa programs is not theoretical — it is creating real friction in hiring pipelines. Companies are increasingly factoring visa status into their hiring calculus, pivoting toward candidates who are green card holders or U.S. citizens. This doesn't mean sponsored talent is less capable. It means the process cost and uncertainty are reshaping which candidates get prioritized, and a large segment of the applicant pool is facing a structural headwind that has nothing to do with their skills. The result is a labor market that looks oversupplied at the surface and undersupplied for the roles that matter most.
During the ZIRP era, many companies treated engineering capacity as something you could rent. Consulting firms, contingent workforces, and offshore development shops provided flexible headcount that could scale up or down with project demand. That model is reversing — particularly for anything involving AI.
We are seeing companies actively pull work away from consulting engagements and contingent arrangements in favor of full-time hires. The driver is straightforward: AI-related engineering work produces intellectual property, and companies do not want that IP walking out the door at the end of a contract. The shift from treating engineering as interchangeable labor to treating it as proprietary competitive advantage fundamentally changes the build-versus-buy calculus. For CFOs evaluating engineering spend, this means the relevant comparison is no longer "FTE cost versus contractor cost." It is "FTE cost versus the risk of your AI capabilities being replicated by the consulting firm's next client."
This is the observation that no one in the market is willing to say out loud, so we will.
For the last 25 years, the principle that A players recruit A players and great companies are built by exceptional talent was treated as gospel. It drove how the best companies sourced, assessed, and closed candidates. In the last three years, that principle has quietly eroded. Too many companies have settled into a posture of "good enough" — defaulting to candidates who applied online rather than running intentional, targeted searches for the best talent at the most innovative companies.
The mechanism behind this erosion is not laziness. It is fear. In an era of aggressive cost-cutting, the mid-level managers and directors who would normally advocate for premium hires have gone quiet. They are protecting their own positions. Suggesting that the company spend more on recruiting — hiring a specialized search firm, paying above-market comp, investing in a longer process to land a transformative candidate — feels like career risk when the prevailing message from leadership is efficiency and restraint. The result is a vicious cycle. The companies that most need A players right now — the ones navigating a genuine structural transition in how software is built and deployed — are precisely the ones too afraid to go hire them. They default to the applicant pool they already have rather than going to market for the talent they actually need. And the gap compounds. Every quarter spent with a "good enough" engineering team is a quarter where competitors with stronger talent are pulling ahead.
The ZIRP-era hiring market created a set of candidate behaviors that the current market no longer tolerates. During peak demand, engineers could job-hop every eighteen months without consequence, show up to interviews unprepared, and still field multiple offers. Those dynamics are gone.
Companies are screening for fundamentals again: strong educational backgrounds, demonstrated tenure of meaningful duration, genuine mission alignment, and preparation that signals respect for the process. Candidates who show up early, ask thoughtful questions, demonstrate intellectual curiosity about the company's problems, and follow up with substance are the ones advancing. Entry-level hiring rates for engineering have dropped approximately 73% according to Ravio's data, not because companies don't need junior engineers, but because AI is automating the routine tasks that historically justified those roles — and because the bar for what "entry-level" means has risen accordingly. The share of AI and ML roles in the tech job market grew from 10% in 2023 to roughly 50% by 2025. The candidates who are thriving are the ones who treated the downturn as a signal to invest in depth — not the ones who assumed AI would substitute for the knowledge they never built.
The software engineering labor market isn't shrinking and it isn't recovering. It is bifurcating.
On one side, commodity engineering roles are being compressed by AI tooling, global talent supply, immigration policy friction, and a buyer's market that has pushed compensation down 10-25% from peak levels. On the other, the engineers who can architect complex systems, evaluate AI output with informed skepticism, and make judgment calls that no model can replicate are more valuable than at any point in the last decade. The gap between these two sides is not narrowing. It is accelerating.
The question for every hiring leader reading this is not "how many engineers do we need?" It is "which engineers — and are we brave enough to go get them?" Because the Efficiency Trap works both ways. AI makes great engineers dramatically more valuable and average engineers dramatically more replaceable. The companies that understand this will use the current market dislocation to acquire talent that would have been unreachable two years ago. The companies that don't will optimize for short-term cost savings, staff their teams with whoever applies, and wonder in eighteen months why their competitors are shipping faster.
Your stability is in your marketability. That's true for individual engineers, and it's true for the companies that hire them. The organizations with the most marketable engineering teams — the ones who invested in depth over convenience, fundamentals over hype, and intentional recruiting over "good enough" — will be the ones still standing when the next abstraction layer arrives.
The macro data tells one story. Our searches tell a more granular one — and they confirm the bifurcation is real.
Over the past twelve to eighteen months, the composition of our engineering engagements has shifted meaningfully. We are running fewer searches for commodity roles — generalist UI engineers, full-stack developers, and the types of positions that defined the high-volume hiring of 2021 and 2022. In their place, we are seeing growing demand for engineers building AI-native products, embedded software for physical devices, and infrastructure that supports proprietary model development. The hard-tech and frontier AI segments of the market are where urgency is highest and where our clients are willing to invest most aggressively in talent quality.
The compensation data from our placements mirrors the split. Roles on the commodity side of the market have come down 10-25% from their peaks, with signing bonuses effectively gone. Roles on the AI and systems side are commanding packages that would have been outliers even during the boom — in some cases, seven-figure total compensation for engineers building frontier models. Our clients who are hiring most successfully right now share a set of common traits: they have a clear thesis on where AI fits in their product and organization, they are willing to pay for specificity over generalism, they are pulling IP-critical work in-house rather than outsourcing it, and they are running disciplined, intentional search processes rather than relying on inbound applicants.
The market has not returned to normal. It has reorganized around a new set of priorities. Companies that are building with conviction, hiring with intentionality, and treating engineering talent as a strategic asset rather than a cost center are the ones defining what comes next. If your engineering org design is shifting and you need senior talent who can lead through the transition, we'd like to hear what you're building.
hello@verticalmove.com
Verticalmove is a specialized talent acquisition partner that places senior individual contributors, leaders, and executives at PE-backed, venture-backed, mid-market, and enterprise companies across 10+ industry verticals.