How to read this. The four panels are four lenses on one event. The first is the structural signal: tech-sector employers (Information; computer-systems design) have been shedding payroll since late 2022 while total employment grows, and CompTIA finds tech occupations across all industries growing and forecast up 1.9% in 2026. Tech jobs are not disappearing so much as dispersing out of tech companies into everyone else’s payroll, while the tech companies themselves convert operating headcount into capital equipment. The $700 B AI build-out is explicitly funded by the same announcements that cut staff. The age panel shows where the adjustment lands: the divergence between developers aged 22–25 and everyone older opens within months of ChatGPT’s launch and widens steadily. It is a within-occupation split (the Stanford authors control for firm-level shocks and find the same pattern), so it cannot be explained by which companies or industries were hit. Incumbents are kept and even added; the pipeline that would replace them is being cut. The noise panel is a warning about resolution: the household-survey-derived occupation numbers that drive headlines swing ±250 k in a month; the payroll series moves single-digit thousands. Most month-to-month drama is sampling noise. The postings panel is the paradox that resolves the restructure: record layoff announcements and a three-year high in job postings in the same quarter is what reallocation looks like: different people are being hired than fired.

What is measured and what is press. The indexed panel and the bars are measured series (BLS CES payrolls via FRED, seasonally adjusted). Tech unemployment (2.9%) and the occupation diamonds are CompTIA computations on BLS microdata, real but survey-volatile. Job-posting counts are CompTIA/Lightcast tallies reported in press releases; the February value is derived from CompTIA’s stated +9.7% March month-on-month change, and the postings-versus-2020 split (all tech −36%, ML engineers +59%) is taken from secondary reporting. The age panel is measured: ADP administrative payroll microdata as published in the Stanford Digital Economy Lab’s Canaries dashboard (a ~4.6 M-worker balanced firm sample, not a survey), downloaded 23 Jul 2026. Layoff figures are Challenger, Gray & Christmas announcements: intentions, not realised separations, and multi-year programs announced at once.

How to read this. The bars understate the shift because the adjustment is happening at the intake, not the exit: TCS’s cuts are ~2% of its workforce, but Wipro’s fresher guidance fell to 7,500–8,000 with recruits publicly reporting onboarding deferrals of seven months and more. A sector can add 135 k net jobs while the entry pipeline quietly closes: the same pattern as the US, where employment of developers aged 22–25 fell ~20% while every older cohort grew. The cuts are not financial distress: they attack the industry’s founding economics. If an AI agent does the work of ten tier-1 BPO analysts, the cost advantage of Bengaluru over Philadelphia disappears; industry reporting describes generative voice agents matching human operators on speed, accent and empathy across the $40 B voice segment, with the impact visible in hiring freezes rather than layoff announcements. NASSCOM’s own strategic review describes providers moving from per-head (FTE) pricing to outcome-based contracts, the clearest tell that vendors themselves expect revenue and headcount to decouple permanently.

What is estimated. India has no BLS equivalent; there is no measured monthly series to plot. NASSCOM aggregates are annual self-reported industry figures (FY26 = Apr 2025–Mar 2026). Company cut totals are press tallies of announcements (TCS announced ~12,000 in Jul 2025 and the tally grew to ~23,500 across 2026 reporting); Infosys disputes the “mass layoff” framing of its ~950. The Gartner $80 B figure is global, not India-specific, though India holds the largest share of the world’s outsourced contact-centre seats. Treat all India panels as directional.

What survives the noise: four structural trends. (1) Capital for labour: profitable companies cut staff explicitly to fund AI infrastructure; US hyperscaler capex ~$700 B in 2026, roughly double 2025. (2) The entry rung closes first: US developer employment aged 22–25 −20% since Nov 2022 while every cohort over 30 grew (ADP); Indian fresher intake deferred and shrunk, so experienced staff are kept while pipelines are cut. (3) Occupation and sector diverge: tech occupations grow (dispersing into non-tech employers) while tech sectors shrink, so IT as a function is leaving IT as an industry. (4) Pricing decouples from headcount: outcome-based contracts replace per-head billing in Indian services; the AI-skill wage premium (~28%, with AI-engineer salaries ~$206 k and rising) concentrates income in a smaller, more expensive workforce.