Tech Layoffs and AI: The 2026 Reality Check

May 10, 2026

The full data-driven analysis of tech companies firing engineers, citing AI as the reason, and whether the numbers support the narrative.

Written by Yuma Heymans (@yumahey), who has been writing code since age six and built HeroHunt.ai, the world's first AI Recruiter. He sits at the intersection of AI, hiring, and engineering, watching both sides of this disruption unfold in real time.

In the first week of May 2026, Coinbase fired 700 employees citing the need to become "lean, fast, and AI-native." Two days later, Cloudflare cut 1,100 jobs (20% of its workforce) while posting record quarterly revenue of $639.8 million. The same week, PayPal announced plans to eliminate 4,760 positions (20% of staff), Upwork cut 24%, and Freshworks dropped 500 employees after its CEO said "over half our code is written by AI" - Fast Company.

This was not a bad week. It was a normal week in 2026.

Total tech layoffs in 2026 have reached approximately 128,270 people across 286 layoff events as of May 10, averaging roughly 1,002 job losses per day. Over 45 CEOs have explicitly cited AI as the reason. Shares routinely climb on the announcements. The capital being freed from payroll is being poured into AI infrastructure, with the four largest tech companies alone spending a combined $725 billion on AI capex in 2026 - 24/7 Wall St..

This guide is the full analysis. Not the headline, not the hot take, but the actual data: which companies are cutting, how many people, what the CEOs are saying versus what the numbers show, and whether AI is genuinely shrinking engineering teams or whether something else is going on. It covers the specific layoff events, the productivity data, the compensation shifts, the counter-arguments, and the case studies of companies that tried replacing workers with AI and had to reverse course. Every claim is sourced.

The May 2026 Wave: What Just Happened
The Full Layoff Timeline: Every Major AI-Cited Cut
What the CEOs Are Actually Saying
The "AI Excuse" Problem: Potential vs Performance
The Productivity Data: What AI Actually Delivers
Junior Engineers: The Real Casualties
The Restructuring of Engineering Teams
Compensation: Two Markets in One Profession
Companies That Tried Full AI Replacement and Failed
The Vibe Coding Disruption
What the Startups and VCs Are Doing
The Counter-Argument: Why Engineering Is Not Dying
What Is Actually Happening (The Synthesis)

1. The May 2026 Wave: What Just Happened

The first two weeks of May 2026 produced a concentrated burst of AI-attributed layoffs that felt, to many in the industry, like a tipping point. But understanding what happened requires looking at each event individually, because the details reveal a more complicated story than "AI is replacing engineers."

Coinbase fired 700 employees on May 5, roughly 14% of its workforce. CEO Brian Armstrong sent the termination memo at 6:55 a.m., stating the company must become "lean, fast, and AI-native." He said AI now lets engineers "ship in days what used to take a team weeks." The restructuring eliminated "pure managers" in favor of "player-coaches" and introduced "AI-native pods" that could include one-person teams directing AI agents. Armstrong then warned that mass layoffs are coming to "every company" as AI reshapes corporate America. Coinbase stock gained after the announcement - CNBC.

Cloudflare cut 1,100 employees on May 7, 20% of its 5,156-person workforce, making it the first mass layoff in the company's 16-year history. CEO Matthew Prince stated: "Today's actions are not a cost-cutting exercise or an assessment of individuals' performance; they are about Cloudflare defining how a world-class, high-growth company operates and creates value in the agentic AI era." He cited the tipping point as November 2025, when teams began seeing "massive productivity gains, team members that were two, 10, even 100 times more productive than they had been before." Internal AI usage had increased more than 600% in three months. Employees were running thousands of AI agent sessions daily - TechCrunch.

The critical detail: Cloudflare reported record quarterly revenue of $639.8 million, up 34% year-over-year, in the same earnings call. The layoffs happened during record growth, not decline. Stock dropped 24%, but that was attributed to guidance concerns, not the cuts themselves - CNBC.

The same week brought cuts at PayPal (4,760 jobs, 20% of workforce), Upwork (145 jobs, 24%), Freshworks (500 jobs, 11%), and Bill (up to 30% headcount reduction). Each cited AI as the driving factor. Together with Cloudflare and Coinbase, these companies eliminated over 8,000 positions in a single week - TechRadar.

What makes this wave different from previous tech layoffs is not the scale. The 2023 post-ZIRP corrections were larger in absolute numbers. What is different is the stated rationale. These companies are not claiming financial distress or market downturns. They are claiming that AI has made a meaningful portion of their workforce unnecessary, and they are making that claim while posting record revenue.

The affected employees report a specific kind of whiplash. Many describe working in organizations that spent the previous year encouraging AI adoption, providing internal training, celebrating teams that shipped features faster using AI tools, and then using those same productivity gains as the justification for cutting the teams that achieved them. The message employees hear is: "Thank you for becoming more productive. Your reward is that we no longer need you." Whether that interpretation is fair or not, it shapes the narrative among engineers and creates a chilling effect on genuine AI adoption at the individual contributor level.

The severance packages have been notably generous in most cases, which tells its own story. Cloudflare offered full pay through end of 2026, healthcare through end of 2026, and equity vesting through August 15. Block offered 20+ weeks depending on tenure, equity vested until end of May, 6 months healthcare, corporate devices, and an extra $5,000. These packages suggest the companies know the cuts are not about individual performance and want to maintain employer brand reputation in a market where they will need to attract top AI talent going forward.

2. The Full Layoff Timeline: Every Major AI-Cited Cut

The May 2026 wave did not appear out of nowhere. It is the latest acceleration in a trend that has been building for over a year. Here is the complete timeline of major layoffs where AI was explicitly cited as the primary reason, with verified numbers and sources.

Block (Square/Cash App), February 26, 2026: The event that set the template. CEO Jack Dorsey cut approximately 4,000 employees, roughly 40% of the workforce, reducing the company to just under 6,000 people. Dorsey said: "I think most companies are late. Within the next year, I believe the majority of companies will reach the same conclusion and make similar structural changes." Block's CFO cited 18 months of AI capability gains that enabled the company to "move faster with smaller, highly talented teams." Shares soared up to 24% after the announcement. Reported gross profit: $2.87 billion in Q4, up 24% year-over-year - CNN.

Oracle, March 31, 2026: The largest and most controversial. An estimated 20,000 to 30,000 employees were cut, roughly 18% of the 162,000 workforce. Termination emails were sent at approximately 6 a.m. with no prior warning. The most damaging detail: according to TIME, Oracle ran a deliberate data-collection program asking employees to document their workflows to train AI systems, then used the results to make those employees redundant. One employee, a 30-year veteran technical writer, was called while driving to the hospital for back surgery and told she was laid off. $300,000 worth of her stock units vanished overnight. The company had posted a 95% jump in net income ($6.13 billion) last quarter and committed to $156 billion in AI infrastructure buildout - TIME.

Meta, announced April 2026, effective May 20: 8,000 employees (10% of workforce) plus 6,000 open roles frozen. The internal memo explicitly cited AI automation as enabling reductions, particularly in content moderation, customer support, software testing, and certain engineering roles. Simultaneously, Meta raised 2026 capex guidance to $125-$145 billion for AI infrastructure - CNBC.

Microsoft, April 2026: Offered voluntary buyouts to up to 8,750 employees (7% of US workforce), the first voluntary buyout program in the company's 51-year history. Set 2026 capex at $190 billion for AI infrastructure - Fortune.

Salesforce, September 2025 and February 2026: Cut approximately 5,000 customer service roles across two rounds. CEO Marc Benioff said "I need less heads" to explain why AI agents are reducing staffing needs. Salesforce's Agentforce product handled 1.5 million customer conversations while human agents handled another 1.5 million, achieving roughly the same satisfaction scores. Benioff also announced Salesforce would hire no more software engineers in 2025 - CNBC.

Pinterest, 2026: Cut 15% of workforce, explicitly shifting resources toward AI-driven products - HR Executive.

Chegg, May and October 2025: Cut 248 employees (22%) in May, then 388 employees (45%) in October. Stock crashed 99% from its 2021 peak of $113.51 to an all-time low of $0.48. CEO admitted ChatGPT was destroying customer growth. Cited "new realities of AI" - CNBC.

Company	Date	Jobs Cut	% of Workforce	Revenue Trend
Block	Feb 2026	~4,000	40%	Record profit
Oracle	Mar 2026	~30,000	18%	+95% net income
Meta	Apr-May 2026	8,000	10%	Growing
Microsoft	Apr 2026	~8,750	7%	Record revenue
Salesforce	Sep 2025 - Feb 2026	~5,000	Mixed	Growing
Coinbase	May 2026	700	14%	Growing
Cloudflare	May 2026	1,100	20%	Record revenue
PayPal	May 2026	4,760	20%	Growing
Pinterest	2026	~15%	15%	Growing
Freshworks	May 2026	500	11%	Growing

The pattern in this table is the story. Nearly every company that cited AI as the reason for layoffs was posting strong or record financial results at the time. This is not a recession-driven correction. It is an operating model shift executed during profitability.

In aggregate, layoffs explicitly linked to AI in 2025 reached 55,000 people, more than 12 times the number attributed to AI just two years earlier. In 2026, the pace has accelerated further, with April alone seeing 83,387 announced job cuts, of which AI was cited as the primary reason for 21,490 - Challenger, Gray & Christmas data via Fortune.

Amazon has cut at least 30,000 jobs since October, representing roughly 10% of its corporate and tech workforce. Google has ongoing reductions of roughly 1,500 positions. Even Apple, which has been relatively quiet on the layoff front, is restructuring internally while planning to hire 20,000 new people focused on AI and R&D, a signal that even "growth" companies are shifting composition, not just adding uniformly.

Bloomberg projects AI-related job displacement could affect up to 502,000 roles economy-wide in 2026. The Challenger, Gray & Christmas data shows that April 2026 was up 38% from March in announced cuts. The acceleration is not slowing. Each major announcement creates competitive pressure for other CEOs: if your competitor cut 20% and their stock went up, the board starts asking why you have not done the same. This herd dynamic explains the clustering of announcements in a single week, a phenomenon familiar from every previous tech layoff cycle but now dressed in AI language.

3. What the CEOs Are Actually Saying

The CEO rhetoric around AI and headcount has shifted dramatically in 2026. Where executives once hedged with vague references to "operational efficiency," they are now making explicit, sometimes provocative claims about AI replacing human workers. These statements matter because they set the narrative for Wall Street, influence competitor behavior, and shape the lived experience of hundreds of thousands of workers.

Dario Amodei (Anthropic CEO) has been the most aggressive. At Davos in January 2026, he said: "I have engineers within Anthropic who say I don't write any code anymore. I just let the model write the code, I edit it." In March 2025, he predicted AI would write "90% of code in 3-6 months" and "essentially all of the code in 12 months" - Yahoo Finance. That timeline has passed without those predictions materializing at the scale he described, but his statements provided cover for other CEOs to make similar claims.

Sam Altman (OpenAI CEO) was more measured but directionally similar: "My basic assumption is that each software engineer will just do much, much more for a while. And then at some point, yeah, maybe we do need less software engineers." He added that in many companies "at least half" of code is already written by AI - Slashdot.

Mark Zuckerberg (Meta) stated: "Our bet is sort of that in the next year, probably maybe half the development is going to be done by AI, as opposed to people." He envisions every engineer becoming "more of like a tech lead" who leads "an army of AI agents" - TechTimes.

Satya Nadella (Microsoft) confirmed: "Today, over 30% of Microsoft's code is already being generated by AI." Microsoft expects this to hit 60% by 2026 - TechTimes.

Sundar Pichai (Google) announced that 75% of all new Google code is AI-generated and approved by engineers, up from 50% the previous fall. Yet in the same statement, he pledged to hire more engineers through 2026, calling AI "an accelerator, not a replacement" - Fast Company.

Jensen Huang (Nvidia) pushed back forcefully: "If we convinced all the young college graduates to not be software engineers, and it turns out the United States needs more software engineers than ever, that's hurtful." He accused some CEOs of having a "God complex" about AI apocalypse warnings - Fortune.

The contrast between Pichai and Huang on one side and Amodei and Zuckerberg on the other is revealing. The companies that sell AI (Anthropic, OpenAI) have the strongest incentive to promote the narrative that AI replaces workers, because that narrative drives adoption of their products. The companies that make the hardware (Nvidia) or that need engineers to build AI features (Google) have more reason to temper the narrative. This does not mean either group is lying. But the incentive structure is worth acknowledging when evaluating their claims.

There is also a generational dimension to the CEO statements. Many of the executives making the boldest claims about AI replacing engineers are not engineers themselves, or if they were, they have not written production code in years. Amodei, to his credit, was a researcher before becoming a CEO. But when a CEO says "my engineers tell me they don't write code anymore," that is a claim being filtered through organizational hierarchy. The engineers "not writing code" at Anthropic are doing something: they are reviewing AI-generated code, designing systems, making architectural decisions, debugging edge cases, and maintaining quality. What they are not doing is the line-by-line typing that constitutes a small fraction of what engineering actually involves. The CEO framing makes it sound like the engineer's job disappeared. The reality is that the job changed shape.

Freshworks CEO Dennis Woodside's statement is worth examining closely: "Over half our code is written by AI." This is almost certainly true by the metric of lines generated. But it obscures the question of value. If AI writes 100 lines of boilerplate and a human writes 10 lines of critical business logic, the AI wrote "91% of the code" but the human wrote most of the value. The metric of "percentage of code written by AI" is inherently misleading because it treats all code as equal, which any engineer knows it is not.

Marc Benioff offered perhaps the most honest, contradictory sequence. He first said "I need less heads" to justify Salesforce layoffs, then months later contradicted himself by saying companies cutting jobs are "hiding behind a convenient scapegoat... AI" - Fortune. Both statements serve his business interests at different moments, which is the point.

4. The "AI Excuse" Problem: Potential vs Performance

This is where the data gets most uncomfortable for the narrative. A Harvard Business Review study surveying 1,006 global executives in December 2025 found a striking disconnect between stated reasons and actual evidence.

The numbers: 39% of executives made low-to-moderate headcount reductions in anticipation of AI capabilities. 21% made large headcount reductions in anticipation of AI. But only 2% made large reductions based on actual AI implementation that had proven it could replace the work. Meanwhile, 44% said generative AI was the most difficult AI technology to assess for economic value - Harvard Business Review.

Read that again: 60% of executives who cut jobs citing AI did so based on potential, not demonstrated performance. Only 2% had actually proven that AI could do the work before eliminating the people doing it.

The HBR article states it plainly: "Job losses are real, but they precede any proven productivity gains from generative AI." Companies are firing people based on what they believe AI will be able to do, not what it has proven it can do. This is speculative restructuring dressed up as technological inevitability.

Economists have noticed the pattern. About a quarter of March 2026 layoff cuts were attributed to AI and automation, while the other three-quarters trace back to cost discipline, business-unit restructuring, and the lagging effect of a decade of cheap-capital hiring. As one analyst put it, firms are "trying to dress up layoffs as a good news story rather than a bad one, by pointing to technological change instead of past overhiring" - CBS News.

There is a perverse incentive at work. Wall Street rewards AI-narrative layoffs. Block's stock surged 24% after cutting 40% of its workforce. Coinbase stock gained on its announcement. The market is telling CEOs that firing people "because of AI" is better received than firing people "because we over-hired during zero interest rate policy." The outcome for the affected workers is the same, but the stock price response is dramatically different.

Fortune published an analysis titled "The problem with using AI as an excuse to cut jobs" that documented the pattern and warned it could backfire: companies that cut too deep lose institutional knowledge, create execution gaps, and ultimately spend more rehiring when the AI tools fail to deliver the promised productivity - Fortune.

A Chinese court has already ruled that companies cannot terminate employees just to replace them with AI. This regulatory response has not yet appeared in the US, but the legal and ethical framework is evolving - Tom's Hardware.

None of this means AI is not genuinely increasing productivity. It clearly is (as the next section shows). But it does mean that the scale of layoffs attributed to AI substantially exceeds the scale of proven AI productivity gains. The gap between rhetoric and reality is where thousands of jobs are being lost.

The timing pattern further supports the "excuse" interpretation. Many of the companies making AI-attributed cuts hired aggressively during 2020-2022 when interest rates were near zero and growth-at-all-costs was the dominant strategy. Meta grew from 48,000 employees in Q1 2020 to over 87,000 in Q3 2022 before starting cuts. Microsoft, Google, and Amazon followed similar trajectories. These companies needed to right-size regardless of AI. But "we're restructuring for the AI era" plays better with investors than "we over-hired during the bubble." The proof is in the stock prices: Block's shares surged 24% after cutting 40% of staff because the market heard "AI efficiency," not "past mismanagement."

The paradox at the center of this is that both things can be true simultaneously. AI is genuinely making some work faster and some roles less necessary. And companies are also using AI as rhetorical cover for cuts they would have made anyway. Disentangling the two requires looking at each company individually, which is why the case-study approach matters more than the aggregate numbers.

5. The Productivity Data: What AI Actually Delivers

So what does the data actually say about how much more productive engineers become with AI tools? The answer is: real gains, but smaller and more nuanced than the CEO narratives suggest.

GitHub Copilot, the most widely adopted AI coding tool, reached approximately 20 million total users by July 2025, with 4.7 million paid subscribers (up 75% year-over-year). It is deployed at approximately 90% of Fortune 100 companies. GitHub's own research, involving 4,800 developers, found that developers complete tasks 55% faster using Copilot. The tool generates an average of 46% of code written by users, with Java developers reaching 61%. Pull request time decreased from 9.6 days to 2.4 days among Copilot users - Panto.

By mid-2026, 41% of all code written across the industry is AI-generated - Index.dev.

Those are impressive numbers. But the quality picture introduces critical caveats.

GitClear analyzed 211 million changed lines of code and found concerning trends. Code duplication ("copy/pasted" lines) rose from 8.3% to 12.3% between 2021 and 2024. New code revised within two weeks of commit grew from 3.1% in 2020 to 5.7% in 2024, indicating increasingly premature or low-quality commits. Refactoring dropped from 25% of changed lines in 2021 to less than 10% in 2024 - GitClear.

Developer experience tells a consistent story: 45% report debugging AI-generated code takes longer than debugging human-written code. 66% encounter "solutions that are almost right, but not quite" requiring manual fixes. Projects that rely heavily on AI-generated code show a 41% increase in bugs - Stack Overflow Developer Survey 2025.

AI Coding Tool Impact (Measured)

The chart above captures the duality. AI makes engineers faster at generating code, but it introduces new costs in debugging, quality assurance, and rework. The net productivity gain is real, but it is not the "2x, 10x, even 100x" that Cloudflare's CEO claimed. Independent measurement by DX, which surveyed engineering leaders, found that most organizations land in the 8-12% productivity improvement range. McKinsey's broader estimate is 20-45% across the software lifecycle - DX.

Perhaps the most revealing data point: a study by The Pragmatic Engineer surveyed 900+ engineers and found that participants felt 20% more productive with AI, but their actual demonstrated results showed a 20% decrease in productivity on average - Pragmatic Engineer. The perception of productivity gains is not matching the measured reality.

The trust paradox is stark. 84% of developers use or plan to use AI tools. But only 33% trust AI accuracy, 46% actively distrust it, and a mere 3% "highly trust" AI outputs. 76% will not use AI for deployment or monitoring. 69% reject AI for project planning - Stack Overflow.

Anthropic's own 2026 Agentic Coding Trends Report offers an honest internal perspective: developers use AI in roughly 60% of their work but can "fully delegate" only 0-20% of tasks. Approximately 27% of AI-assisted work is work that would not have been attempted at all without AI, suggesting AI is creating new work, not just replacing existing work - Anthropic.

This is the nuance that gets lost in CEO press releases: AI makes engineers faster, but not as fast as the layoff numbers imply. A 10-40% productivity gain does not justify a 20-40% headcount reduction without accepting a proportional reduction in output or quality. Companies that cut deeper than their actual productivity gains support are taking a bet on future AI improvements, not acting on present reality.

There is an additional dimension that productivity studies rarely capture: the type of work AI accelerates versus the type it does not. AI excels at generating boilerplate code, writing tests for existing code, producing documentation, converting between languages, and scaffolding standard patterns (forms, CRUD operations, REST APIs). AI struggles with understanding nebulous business requirements, designing scalable architectures, making high-stakes technical trade-offs, debugging complex distributed systems, and maintaining architectural coherence across large codebases - Pragmatic Engineer.

The distinction matters because it explains why productivity gains feel so different depending on who you ask. A junior engineer doing mostly boilerplate work might genuinely experience a 50-70% speedup. A senior engineer doing mostly architecture and debugging might experience a 5-15% speedup (or even a productivity decrease if they spend time debugging AI-generated code). An internal OpenAI example illustrates the upper bound: a three-person team used Codex to ship 1,500 pull requests and roughly a million lines of code without writing a single line manually, but this was for a specific type of well-defined task, not general engineering work - Unite.AI.

The most honest summary comes from Anthropic's own research: developers can "fully delegate" only 0-20% of tasks. The rest requires human judgment at some stage: defining what to build, reviewing what was built, testing whether it actually works, and deciding whether the solution is architecturally sound. This is why the "AI will write all the code" narrative from Amodei and others has not matched reality even at their own companies. Writing code was never the bottleneck. Deciding what to write, and ensuring what was written actually works correctly, has always been the hard part.

6. Junior Engineers: The Real Casualties

If there is one area where the data is unambiguous, it is the impact on early-career engineers. Junior developers are being disproportionately affected, and the numbers are not subtle.

Stanford's 2026 AI Index Report found that employment among software developers aged 22-25 fell approximately 20% from its 2022 peak by September 2025. The share of juniors and graduates in IT employment has dropped from approximately 15% to just 7% over the past three years. Entry-level hiring at the top 15 tech companies fell 25% from 2023 to 2024, with the decline continuing into 2026 - Stanford Digital Economy Lab.

Critically, all age groups above 26 continue to grow. Stanford researchers describe this as "seniority-biased technological change": AI substitutes for junior labor while leaving senior roles intact.

The mechanism is straightforward. AI has eliminated the specific tasks junior developers were historically hired to perform: boilerplate code, test generation, documentation, CRUD scaffolding, basic bug fixes, and routine data processing. Senior engineers now use AI to handle those tasks themselves, without delegating to a junior team member. A team of five senior engineers with AI tools now does what previously required a team of eight, including three junior developers - CIO.

The employment data tells the story in starker terms. Computer science graduates now face a 5.8% unemployment rate, higher than the general unemployment rate. CS graduates rank 7th highest in unemployment among all college majors at approximately 6.5%, worse than philosophy majors (3.2%), art history majors (3%), and journalism majors (4.4%) - Yale Insights.

This creates a pipeline problem that the industry has not yet grappled with. If junior roles disappear, where do tomorrow's senior engineers come from? David Ellis, SVP of Talent Transformation at Korn Ferry, has warned: "It would be a mistake to stop hiring young, entry-level people. These are the fastest adopters of new technology." But the market is not listening. The rational short-term decision for individual companies (cut juniors, arm seniors with AI tools) creates a collective long-term crisis (no experienced engineers in five years).

Goldman Sachs quantified the damage: AI is eliminating roughly 16,000 net US jobs per month (25,000 substituted, 9,000 added through augmentation). Entry-level workers under 30 bear the brunt, facing a 3.3 percentage-point wage gap increase for those in high-AI-substitution occupations. Displaced tech workers face longer job searches and 3%+ pay cuts, with earnings gaps that widen for a decade - Fortune.

Jensen Huang's warning feels prescient here: if we convince an entire generation that software engineering is a dead end, and then discover we need more engineers than ever, the damage will be difficult to undo.

The data from McKinsey reinforces the scale of the problem. One in three companies plans to reduce headcount in the coming year, with software engineering among the functions with the sharpest expected declines. But these same companies report that they cannot find enough senior engineers with AI expertise. The market is simultaneously shedding junior talent and desperately competing for senior talent, creating a compression that has no historical parallel in the tech industry.

For anyone trying to navigate this market, whether as a company building an AI team or as a recruiter sourcing talent, the junior engineer dynamic is the most important underlying force. It explains why the talent pipeline is narrowing at the entry point even as the exit point (senior roles, AI-specialized roles) widens. It explains why compensation for experienced AI engineers keeps climbing: there are fewer people entering the pipeline to eventually become those senior engineers. And it explains why platforms like HeroHunt.ai are seeing increased demand for AI-powered sourcing, because finding the right senior and mid-level talent in this compressed market requires searching across 1 billion+ profiles rather than relying on the shrinking inbound pipeline of junior applicants.

7. The Restructuring of Engineering Teams

What is happening inside the companies that are cutting is not simply subtraction. It is restructuring. The engineering team of 2026 looks fundamentally different from the engineering team of 2023, and understanding the new shape explains why headcounts are dropping even as output stays constant or grows.

The new team archetype emerging across the industry is leaner and more senior-weighted. Senior engineers design agent systems and set guardrails. Mid-level engineers review AI-generated code. The tasks that juniors used to perform are handled by AI tools. Total headcount decreases while expertise density increases. Teams that adopted this model report reducing cycle time by 40-70% with AI integration - CIO.

Block offers the most dramatic example. After cutting 40% of its workforce, the company reported that developer productivity improved 40% per engineer through AI tool adoption over the prior 18 months. The restructuring was not speculative: they measured the productivity gains first, then made the cuts. This is the exception, not the rule (as the HBR data shows), but it demonstrates the model that other companies aspire to.

Coinbase's approach is particularly illuminating. CEO Brian Armstrong did not just cut headcount. He restructured how teams operate. The "AI-native pods" he described can include one-person teams where a single engineer directs AI agents that handle the responsibilities previously held by engineers, designers, and product managers. This is not a subtle efficiency improvement. It is a fundamentally different organizational model - Fortune.

Shopify's approach was less dramatic in execution but potentially more influential. CEO Tobi Lutke's April 2025 memo did not announce layoffs. Instead, it declared: "Before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI." He added: "What would this area look like if autonomous AI agents were already part of the team?" AI usage was then added to performance reviews. Within eight months, Meta, Microsoft, Google, and Nvidia had adopted similar approaches - CNBC.

The practical effect is that companies that once hired five juniors now hire two mid-level engineers and a Copilot subscription, getting comparable output for significantly less cost. The "missing rung" problem this creates, where entry-level roles drop 20-35% globally, removes precisely the on-ramp that new engineers need to build the experience that makes them valuable - Stack Overflow.

For companies hiring in this new environment, the recruiter's job has changed too. Traditional recruiting tools and approaches built for high-volume junior hiring are less relevant when the need is for fewer, more experienced, AI-fluent engineers. Platforms like HeroHunt.ai are adapting to this shift, using AI sourcing across 1 billion+ profiles to identify the specific senior and AI-specialized talent that companies now need, rather than filling entry-level pipelines.

8. Compensation: Two Markets in One Profession

The compensation data reveals that "software engineering" is no longer a single market. It has split into two distinct labor markets with very different trajectories, and understanding this split is essential for making sense of what is happening.

Market 1: Traditional software engineering. Tech salaries increased an average of 1.6% in 2025, down from 2.9% in 2024 and 3.5% in 2023, the lowest growth in 15 years. Average US software engineer salary in 2025 was $105,683, approximately 18.79% lower than in 2023 - Ravio. Overall tech salary growth is approaching stagnation for generalist roles.

Market 2: AI-specialized engineering. PwC's 2025 Global AI Jobs Barometer documented a 56% wage premium for AI skills, up from 25% the prior year. The premium doubled in 12 months. AI/ML roles command a 12% premium at the IC level versus non-AI roles. LLM fine-tuning specialists earn $195,000-$350,000. Frontier-lab software engineers command $600,000-$795,000 in median total compensation - Ravio.

Software Engineer vs AI Engineer Salary Premium

The divergence shown in this chart is the clearest indicator of what the market actually values. Companies are paying dramatically more for AI-skilled engineers while traditional engineering salaries stagnate or decline. The gap is not narrowing. It is accelerating.

At the extremes, the numbers are staggering. Meta offered AI researchers signing bonuses reported at up to $100 million. OpenAI distributed approximately $1.5 billion in total retention bonuses to roughly 1,000 employees. LinkedIn pays AI engineers approximately $288,050 at entry level versus $225,000 for non-AI roles. Intuit staff-level AI engineers reach approximately $917,000 in total compensation versus $515,000 for non-AI staff at the same level - Inc..

The implication is clear: the industry is not devaluing engineering broadly. It is repricing engineering based on AI relevance. Engineers who can build, deploy, and optimize AI systems are in a seller's market. Engineers who do the work that AI tools are learning to automate face a buyer's market. Same profession, two entirely different economic realities.

National Association of Colleges and Employers expects CS major starting salaries to increase nearly 7% year-over-year in 2026, even as entry-level jobs decline. This suggests that the surviving entry-level positions are paying more, but there are far fewer of them. The floor is being raised, but the door is getting narrower.

The bifurcation has geographic dimensions as well. North America averages $285,000 for senior AI engineers. Western European AI engineers earn $60,000-$145,000. Eastern European engineers (Poland, Romania, Ukraine) earn approximately $40,000, representing a 60-70% cost savings versus the US. Latin American AI engineers average approximately $40,800 - Axiom Recruit. Companies doing AI-attributed layoffs in the US are simultaneously hiring AI engineers in lower-cost markets. The headline is "we cut 20% of engineering," but the fine print often includes "and hired equivalent capacity at one-third the cost in a different geography." This reframing changes the analysis significantly: some "AI layoffs" are really offshoring layoffs with an AI narrative attached.

9. Companies That Tried Full AI Replacement and Failed

The "AI replacing workers" narrative needs to be examined against the evidence of companies that actually tried it and discovered the limits. These case studies are essential reading for any CEO considering a similar move, and for any engineer trying to assess the actual threat.

Klarna is the most instructive case. In 2024, CEO Sebastian Siemiatkowski announced that the company's AI chatbot (built in partnership with OpenAI) did the work of 700 customer service agents. The company reduced headcount from approximately 5,000 to 3,800. The market celebrated. Then it backfired.

Customer satisfaction deteriorated on complex interactions. Repeat customer contacts jumped 25%. The AI could not handle nuanced problem-solving, lacked empathy, and failed on non-formulaic responses. Projected cost savings did not fully materialize. Siemiatkowski eventually admitted: "We focused too much on efficiency and cost. The result was lower quality, and that's not sustainable." Klarna quietly reversed course, shifting to a hybrid model with AI handling routine queries and humans handling escalations. 55% of companies surveyed now report regretting AI-driven job cuts - Digital Applied.

Duolingo provides another walkback. The company announced an "AI-first" approach and eliminated a significant portion of its contractor workforce. CEO Luis von Ahn publicly stated the company would "gradually stop using contractors to do work that AI can handle." One week later, after significant backlash, he publicly walked it back: "I do not see AI as replacing what our employees do." He admitted the controversial memo "did not give enough context" - Fortune.

Oracle's situation raises ethical questions beyond business outcomes. Employees were asked to document their workflows to train AI systems, effectively building the tools that would replace them. A survey of 272 laid-off employees found 62% were over 40, and 22% had worked there for more than 15 years. One employee described it: "It really makes you feel used and abused. They're having you do something, it's recorded, and then they're going to replace you with whatever you just built" - TIME.

The broader survey data supports the pattern. Over half of UK business leaders who rushed to replace human jobs with AI say they now regret it - Fast Company. The regret cycle follows a predictable arc: announce AI replacement, cut staff, discover AI cannot handle edge cases and complexity, suffer quality degradation, and either quietly rehire or endure the consequences.

The Oracle case deserves deeper examination because it reveals a dynamic that other companies have not had to confront publicly. When employees are asked to document their workflows to train AI, they are essentially being told to automate themselves. The 272 surveyed former employees described a betrayal that goes beyond normal layoff distress: they feel complicit in their own replacement. The fact that 62% were over 40 and 22% had more than 15 years of tenure suggests the company targeted experienced, higher-cost employees whose accumulated knowledge could be extracted and encoded into systems. TD Cowen analysts estimated the cuts freed up $8-10 billion in free cash flow for data center projects, reframing the layoffs not as AI optimization but as a direct capital transfer from payroll to infrastructure.

This dynamic, where the money saved from salaries is redirected into AI infrastructure, is not unique to Oracle. The four largest tech companies plan to spend $725 billion on AI capex in 2026, up 77% from $410 billion in 2025. This spending has to come from somewhere. For many companies, the "somewhere" is the payroll line item. The layoffs are not just about AI making workers unnecessary. They are about freeing up capital to build the AI systems that companies hope will eventually make workers unnecessary. The sequence matters: the cuts are funding the replacement, not the replacement justifying the cuts.

This does not mean AI replacement never works. It means that the companies rushing to cut deepest and fastest, driven by Wall Street incentives and competitive pressure, are more likely to overshoot than those taking a measured approach. The measured approach (use AI to augment existing teams, gradually shift role composition, maintain institutional knowledge) produces sustainable gains. The dramatic approach (cut 40% of workforce, announce AI-native transformation) produces headlines but often requires course correction.

The most telling data point across all these case studies is the 55% regret rate among companies that rushed AI-driven job cuts. More than half look back and wish they had moved more slowly. This should give pause to any CEO contemplating a Coinbase-style restructuring. The companies that will be studied as success cases in two years are not the ones making the boldest cuts today. They are the ones making the most measured transitions: using AI to augment their best people, gradually redefining roles as AI capabilities mature, and maintaining enough organizational knowledge to course-correct when (not if) the AI tools fail to deliver on their most ambitious promises.

The lesson from Klarna in particular deserves its own emphasis. When Siemiatkowski said "we focused too much on efficiency and cost, the result was lower quality," he was articulating a failure mode that applies far beyond customer service. Engineering quality is harder to measure than customer satisfaction, which means the quality degradation from premature AI replacement in engineering may take longer to surface. But when it does surface, in the form of increased bugs, architectural decay, security vulnerabilities, and technical debt, the cost of correction will be far higher than the savings the layoffs produced. The Oracle employees who were asked to train their replacements were encoding their explicit knowledge. The implicit knowledge, the judgment, the institutional context, the understanding of why things were built a certain way, disappears with the people.

10. The Vibe Coding Disruption

Beyond the layoff story, there is a parallel disruption that is changing who can build software at all. "Vibe coding," a term coined in February 2025 by Andrej Karpathy (OpenAI co-founder, former Tesla AI lead), refers to creating software with AI where the user does not necessarily understand the code being produced. MIT Technology Review named it one of the "10 Breakthrough Technologies of 2026". Collins English Dictionary made it the Word of the Year for 2025 - Harvard Gazette.

The market impact is real. Lovable, a leading vibe coding platform, reached $300 million ARR and a $6.6 billion valuation. Job postings requiring AI coding tool experience increased by 340% between January 2025 and January 2026, while postings for pure implementation roles declined by 17% - AI Builder Club.

Non-coders are now shipping real, working applications through prompting. The pool of potential "developers" is expanding from millions to potentially hundreds of millions. This matters for the engineering job market because it blurs the line between who needs a dedicated engineer and who can build something functional themselves.

But the quality gap persists. Bug density in projects with unreviewed AI-generated code is 23% higher than in human-overseen projects. The skill of 2026 is not writing code. It is "looking at AI-generated code and spotting errors." This is a skill that requires deep engineering knowledge, which is why senior engineers remain valuable while the entry points for new engineers narrow. The irony is that the people best positioned to use vibe coding effectively are the experienced engineers who do not need it, while the non-engineers using it most enthusiastically are least equipped to catch its mistakes.

For the engineering job market, vibe coding represents a third vector of disruption beyond layoffs and team restructuring. The first vector (layoffs) removes existing jobs. The second vector (restructuring) changes which jobs exist. The third vector (vibe coding) reduces the need for certain jobs to exist in the first place. When a product manager can build a working prototype without involving an engineer, the scope of what requires dedicated engineering shrinks. When a non-technical founder can ship a minimum viable product using AI tools, the startup's first engineering hire happens later and fills a different role (architecture, scaling, security) rather than basic feature development.

The combined effect of all three vectors is not the elimination of software engineering but the raising of its floor. The minimum level of skill, judgment, and architectural thinking required to add value as an engineer is rising rapidly. Tasks that once required engineering skill (basic CRUD applications, simple data pipelines, standard API integrations) can now be done by non-engineers with AI tools. What remains for professional engineers is the work that requires genuine expertise: system design, performance optimization, security architecture, complex debugging, and the judgment calls that determine whether a system will scale or collapse.

11. What the Startups and VCs Are Doing

The startup ecosystem provides a forward-looking indicator of where the engineering labor market is headed, because startups are less encumbered by legacy team structures and more willing to experiment with new models.

The data from Y Combinator is striking. For about a quarter of current YC startups, 95% of their code was written by AI. YC CEO Garry Tan said: "You don't need a team of 50 or 100 engineers. You don't have to raise as much. The capital goes much longer." YC's current batch is the "fastest growing and most profitable in fund history" because of AI - CNBC.

Founding teams are shrinking. 36.3% of new ventures in 2026 are solo-founded. Solo founders are building products that would have required teams of 10-20 people three years ago. The examples are specific: Maor Shlomo built Base44 alone, sold to Wix for $80 million in 6 months (250,000 users, profitable). Danny Postma's HeadshotPro reached $3.6 million ARR as a solo operation. 38% of seven-figure businesses in early 2026 are led by solopreneurs who replaced hires with AI workflows - Solo Business Hub.

Dario Amodei predicted in May 2025 that the first billion-dollar one-person company would emerge by 2026, giving it 70-80% odds. Whether that specific prediction materializes, the directional trend is clear: the minimum viable team for a software company is approaching one person. A full solopreneur tech stack costs $3,000-$12,000 per year, a 95-98% reduction versus traditional staffing, enabling operating margins of 60-80% - NxCode.

AI captured close to 50% of all global startup funding in 2025, with $202.3 billion invested. Mega-rounds of $100 million or more accounted for 79% of AI funding. The capital is flowing disproportionately toward companies that are building with AI, not toward companies that are building software the traditional way.

For the engineering job market, this means two things. First, the total number of engineering jobs at startups is declining per company (smaller teams, same output). Second, the remaining jobs are dramatically more valuable and harder to fill, because they require the judgment, architecture, and AI-orchestration skills that make small-team AI-augmented development work. The recruiter serving this market needs to find engineers who can be the entire engineering function, not just a contributor within a large team.

The math is striking when you put it together. A full solopreneur tech stack costs $3,000-$12,000 per year, a 95-98% reduction versus traditional staffing, enabling operating margins of 60-80%. A YC-backed startup that would have raised $2 million to hire 8 engineers can now raise $500,000 and hire 2, using AI for the rest. The capital efficiency improvement flows directly to investors (who get more ownership per dollar invested), to founders (who dilute less), and to the small number of engineers who do get hired (who command premium compensation because they need to be exceptional rather than adequate).

This is not a bubble or a temporary distortion. It is a structural change in the economics of building software companies. The companies being built today on smaller teams with AI leverage will not suddenly need to triple their headcount when they scale. The efficiency is permanent, even if the specific AI tools change.

12. The Counter-Argument: Why Engineering Is Not Dying

The layoff data and CEO rhetoric paint a bleak picture, but the full analysis requires the counter-arguments, and there are strong ones backed by data.

Morgan Stanley Research estimates the software development market will grow at 20% annually, from $24 billion in 2024 to $61 billion by 2029. The Bureau of Labor Statistics still projects 15% job growth for software developers from 2024 to 2034, significantly faster than average. Software developers held about 1.7 million jobs in 2024, with roughly 129,200 openings projected annually - Morgan Stanley.

The historical pattern strongly supports the counter-argument. Every major wave of automation in software has produced the same fears, and every time, jobs grew. In the 1950s, compilers were feared to eliminate programming jobs. Instead, coding became easier and cheaper, creating far more demand. In the 1980s and 90s, IDEs and modern compilers made coding 5-10x faster, producing an explosion in software complexity and jobs. In 2008 and after, Stack Overflow and open-source libraries provided instant knowledge, increasing the developer population rather than shrinking it - Coursera.

The mechanism is called the Jevons Paradox: when a resource becomes cheaper and more efficient to use, total consumption of that resource increases rather than decreases. If AI makes software development cheaper, organizations will not do the same amount of work with fewer people. They will do more work. They will build products they could not previously justify, automate processes they previously ignored, and expand into markets that were previously too expensive to serve with custom software.

CNN published an analysis in April 2026 titled "The demise of software engineering jobs has been greatly exaggerated," citing CIO plans to increase software spending by 3.9% in 2026 and the observation that the surge in AI-generated code is creating bottlenecks in other stages of the software lifecycle (review, testing, deployment) that require more human oversight, not less - CNN.

Sundar Pichai's position embodies this argument. Despite Google generating 75% of new code with AI, he pledged to hire more engineers, not fewer. His reasoning: AI accelerates what engineers can do, which means more ambitious projects become feasible, which means you need more engineers to oversee and direct the work, not fewer - Fast Company.

Apple plans to hire around 20,000 people focused on R&D, silicon engineering, software development, and AI/ML, with a $500 billion US investment commitment - Fortune. This is the single largest committed hiring number from a Big Tech company, and it cuts directly against the "engineers are becoming obsolete" narrative.

Stack Overflow's analysis makes the structural argument: every platform shift (internet, mobile, cloud) initially triggered displacement fears but ultimately created more software jobs. The internet did not eliminate print designers. It created web designers, then UX designers, then product designers. Mobile did not eliminate web developers. It created iOS developers, Android developers, cross-platform developers, and an entire mobile ecosystem. Each new platform expanded the total demand for people who build software, even as specific skills within each platform became obsolete - Stack Overflow.

The Jevons Paradox argument is strengthened by what is already happening at the edges. Companies that could never justify hiring a software team (small businesses, non-tech enterprises, individual creators) are now using AI tools to build custom software. This expansion of the customer base for software development services could create more engineering jobs (in support, customization, integration, quality assurance) than AI eliminates in pure coding roles. Whether this expansion happens fast enough to offset the contraction is the open question.

Fortune reported that small businesses will hire nearly 1 million graduates in 2026, and some of the hottest roles are "gloriously AI-proof": service technicians, field roles, hands-on positions that require physical presence and human judgment - Fortune. This is cold comfort for CS graduates specifically, but it suggests the overall labor market is healthier than the tech layoff headlines imply.

The counter-argument has limits, though, and they are significant. The BLS also projects that Computer Programmers (those performing routine coding from specifications) will decline 6% from 2024 to 2034. Programming jobs have already plunged 27.5% over two years, placing the profession among the 10 hardest-hit of 420+ occupations tracked by BLS. Programmer employment has fallen to levels not seen since 1980 - Washington Post.

The distinction between "software developer" (growing) and "computer programmer" (declining) in BLS categories captures the real dynamic: the nature of the work is changing, and the jobs that survive are different from the jobs that are being eliminated. But this distinction offers cold comfort to the workers in the declining category, who are told their profession is "growing" while their specific role disappears. The aggregate statistics hide the lived experience of the people being displaced.

The question of whether the Jevons Paradox will hold this time, whether making software cheaper to produce will increase total demand enough to offset the productivity gains, is genuinely open. Every previous automation wave in software followed this pattern, but every previous wave automated a step in the process (compiling, debugging, testing) rather than the core activity itself (generating code). AI is the first technology that automates the act of writing code, which is qualitatively different from automating the tools around code. Whether the historical pattern holds through this qualitative shift is the trillion-dollar question, and anyone who claims certainty in either direction is selling something.

13. What Is Actually Happening (The Synthesis)

After examining the layoff data, the CEO statements, the productivity numbers, the compensation trends, the case studies, and the counter-arguments, a specific picture emerges that is more nuanced than either "AI is replacing engineers" or "everything is fine."

Here is what the data actually supports.

First, total software engineering employment is still growing, but the composition is changing dramatically. Senior and AI-specialized roles are expanding. Entry-level and generalist roles are contracting. The profession is being restructured from a pyramid (many juniors, few seniors) into a diamond shape (few juniors, many seniors, AI handling the base). The total number of jobs may continue growing as Morgan Stanley and BLS project, but the types of jobs and who gets them are shifting rapidly. The BLS distinction between "software developers" (growing 15%) and "computer programmers" (declining 6%, already down 27.5% in two years) captures this perfectly. The programmers who translate specifications into code are being displaced. The developers who design systems, make architectural decisions, and manage complexity are in growing demand - BLS.

Second, many layoffs attributed to AI are about cost discipline and post-pandemic over-hiring. The HBR data showing only 2% of large headcount reductions are based on actual AI implementation (versus 60% based on anticipation) is the strongest counter to the CEO narrative. Companies over-hired during the zero-interest-rate era and are using AI as a more palatable explanation for corrections that were coming regardless. Wall Street's enthusiastic stock price response to "AI-driven efficiency" layoffs versus its negative response to "we over-hired" layoffs creates a perverse incentive to frame every cut as AI-related.

Third, AI productivity gains are real but overstated. Measured gains of 8-12% at most organizations (with McKinsey's broader estimate at 20-45%) are meaningful but do not justify the 20-40% headcount cuts some companies are making. Companies cutting deeper than their measured productivity gains are making a bet on future AI improvements, not acting on present evidence. Some of these bets will pay off. Others, as Klarna and Duolingo have demonstrated, will not.

Fourth, the role is transforming, not disappearing. The shift is captured in a phrase that appears across multiple sources: "Programming as typing code is being automated. Software engineering as designing, debugging, and deciding under uncertainty is not." Engineers are becoming system architects, AI orchestrators, and quality evaluators. The work is different, but it is not gone.

Fifth, the junior pipeline crisis is real and underappreciated. A 20% decline in employment for developers aged 22-25, CS graduate unemployment higher than philosophy majors, and entry-level hiring down 25% at top companies represents a structural break in how the industry develops talent. If this continues for another 2-3 years, there will be a severe shortage of experienced engineers when the current cohort of seniors retires or moves to management. No one is solving this problem yet.

Sixth, the CEO incentive structure matters. AI companies want adoption, so they promote the replacement narrative. Hardware companies want demand, so they promote the augmentation narrative. Public company CEOs want stock price gains, so they frame layoffs in whatever terms the market rewards. None of these actors are neutral observers. The engineers affected are the ones with the least voice in the narrative.

Seventh, the capital flow tells the real story. The $725 billion being invested in AI infrastructure by the four largest tech companies is not coming from new revenue. Much of it is coming from the salary budgets freed up by layoffs. This is not "AI efficiency funding AI progress." It is a deliberate capital reallocation from human labor to machine infrastructure, and the companies are making an enormous bet that the returns on that infrastructure will exceed the returns they were getting from the employees they replaced. Some of those bets will pay off spectacularly. Others, if the AI infrastructure does not deliver the promised productivity (and the measured data suggests it often does not), will result in companies that are understaffed, over-invested in infrastructure they cannot fully utilize, and scrambling to rehire the talent they let go. The Klarna case is a small-scale preview of this dynamic.

The Reality Gap

The chart above captures the gap between what CEOs are claiming (or implying through their actions) and what independent measurement shows. Productivity gains of "2-100x" collapse to 8-12% in measured studies. AI-generated code percentages are real but lower than executive claims. And the share of companies that cut jobs based on proven AI implementation (versus anticipated potential) is vanishingly small at 2%.

The honest answer to "is AI replacing software engineers?" is: not yet, not broadly, but the role is changing fast, and certain categories of engineering work are genuinely being displaced. The honest answer to "are these layoffs really about AI?" is: some are, most are not, and the ones that are tend to overshoot what the current technology justifies. The honest answer to "should engineers be worried?" is: senior engineers with AI fluency should not be worried. Junior engineers and generalists should be concerned. Everyone should be adapting.

The companies that will look smartest in two years are not the ones cutting deepest today. They are the ones investing in the right mix: augmenting existing teams with AI, restructuring toward higher-leverage roles, maintaining the pipeline for future talent, and making measured bets on AI capability rather than speculative cuts based on CEO conference rhetoric.

The defining question is not "will AI replace engineers?" It is "how fast will the role transform, and who benefits from the transition?" The data suggests the transformation is real but the pace is being artificially accelerated by market incentives and CEO posturing. The workers being displaced today are real people with real consequences. The productivity gains being claimed are real but smaller than advertised. And the companies that cut too deep too fast are discovering, as Klarna and Duolingo did, that the gap between AI's promise and its performance still requires human judgment to fill.

For engineers navigating this environment, the practical advice is the same regardless of which interpretation you favor: build AI fluency, move toward system design and architecture, and understand that the value of writing code from scratch is declining while the value of orchestrating, reviewing, and deciding what to build is increasing. For companies navigating it, the advice is equally clear: use the data, not the narrative. Measure your actual AI productivity gains before making headcount decisions based on someone else's press release.

For recruiters and hiring leaders, the synthesis has direct operational implications. The talent market is bifurcating, and recruiting strategies must bifurcate with it. Sourcing for senior AI-fluent engineers requires competing on compensation, speed, and employer brand against every other company trying to hire the same people. Tools like HeroHunt.ai, which use AI to source from over 1 billion profiles and automate outreach through its AI Recruiter Uwi, are built precisely for this compressed market where finding the right senior talent requires searching broadly and moving fast.

At the same time, companies need a strategy for the junior pipeline problem. The companies that figure out how to hire and develop early-career talent in the AI era, creating new kinds of entry-level roles (AI output reviewer, prompt engineer, AI-assisted developer) that build the skills tomorrow's senior engineers will need, will have a structural advantage in five years when the talent shortage hits its most acute phase. The companies that simply stop hiring juniors because AI handles the tasks juniors used to do will find themselves unable to replace their senior engineers when those engineers eventually leave.

This analysis reflects data available as of May 10, 2026. The situation is evolving weekly. Company-specific numbers, revenue figures, and market data should be verified against the latest available reporting before making decisions.