A friend who runs a small verification team in Bengaluru called last month. Half joking, half scared, he said: "We are not being laid off. We are being automated in place." Same employer. Same badge. Fewer people expected to ship the same tape-out—and more Copilot in the loop.
The short version (if you have two minutes):
- Highest pressure now: manual QA, junior RTL, L1 helpdesk, BPO voice, junior CRUD developers.
- More runway through 2026–27: chip bring-up and post-silicon debug, bedside nursing, solution architects with real customer trust.
- Your move this week: find your role in our action table below and do that one step—do not wait for HR to name the problem.
That is what we keep hearing in 2026: not always a pink slip on Friday, but a team that quietly stops backfilling when someone leaves. Senthil and Ian (Ian is a Principal Cloud Architect at Oracle) wrote this because the press releases say "efficiency" while the org chart tells a sharper story.
What the headlines got right (spring 2025 – spring 2026)
Ian watched the Oracle cuts land in real time: pre-dawn emails, badge access gone. The same week, enterprise buyers on his accounts asked which support and DBA tiers they could trim after rolling out AI assistants. He was not just reporting the news. He was living inside it. Meta cut roughly 8,000 roles and dropped 6,000 open reqs. Microsoft offered buyouts to about 8,750 US staff. Amazon removed 30,000 corporate jobs since October 2025. TCS ended FY26 down 23,460 net; Infosys shed 8,440 in Q4 alone; Cognizant's Project Leap set aside up to $270M for severance.
Layoffs + salary pressure
| Company / role | Reported impact | Pay trend (2024→2026) |
|---|---|---|
| Oracle | Est. 20,000–30,000 cuts globally | — |
| Meta | ~8,000 layoffs; 6,000 reqs cancelled | — |
| Amazon | 30,000 corporate roles since Oct 2025 | — |
| TCS / Infosys / Cognizant | Net bench shrink; Project Leap 7k–15k est. | — |
| Manual QA tester (India) | — | −15% to −20% |
| Junior front-end developer | — | −10% to −15% |
| L1 helpdesk / IT support | — | −8% to −12% |
| Chip bring-up / post-silicon engineer | — | +5% to +12% |
| Registered nurse (US) | — | +10% to +15% |
Cut totals from news and SEC filings; pay ranges are directional estimates from compensation surveys and staffing commentary. Your employer and city may differ.
Layoffs are only half the story. Many workers kept their jobs but lost leverage—the pay rows show who is negotiating from weakness.
CNBC noted more than 92,000 tech layoffs in the first four months of 2026. Treat every total as announced cuts—companies still say "restructuring" when they mean margin defence or offshoring. The direction, though, is hard to argue with.
Reporters who interviewed Indian IT workers through early 2026 kept hearing the same split: teams shrank, but only a minority were told "AI" out loud. The rest heard utilisation, pyramid reshape, or budget optimisation—with Copilot on the project anyway. — Pattern reported across Times of India, Outlook Business, and employee forums, 2026
India's cities: where openings cooled fastest
Nationally, Xpheno's June 2026 outlook put active tech openings at about 93,000—a 28-month low, down 14% month-on-month (The Hindu, June 2026). Entry-level slots (0–2 years) fell to roughly 10,000, down 44% year-on-year in the same report.
Xpheno does not always publish a full city split, but hiring platforms and staffing commentary in 2026 consistently show the same hierarchy: NCR and Bengaluru softening fastest, Hyderabad and Pune following, Chennai somewhat steadier thanks to manufacturing and auto-tech accounts.
Estimated active IT openings by hub (June 2026, directional)
| City / hub | Est. openings | YoY trend (reported / estimated) |
|---|---|---|
| Bengaluru | ~28,000 | −20% to −25% (testing, AMS-heavy) |
| NCR (Gurugram, Noida) | ~14,000 | −22% to −28% (telecom & auto testing first) |
| Hyderabad | ~15,000 | −15% to −20% |
| Pune | ~12,000 | −12% to −18% |
| Chennai | ~9,500 | −10% to −15% |
City figures are approximations derived from national Xpheno totals and typical hub share reported in Indian business press—they are indicative, not an official census.
NCR is bleeding visibly in telecom and automobile testing benches—the kind of work AI-assisted test generation eats first. Bengaluru feels it in verification and application maintenance. If you are in either city in a pattern-match role, the job boards are not lying to you.
Is your role at risk? A five-point check
Answer honestly. Count your Yes answers:
- Does your work output look much the same every week? (Yes = risk)
- Could a smart intern do 80% of your tasks after a week of training? (Yes = risk)
- Do you spend more than half your time in Excel, Jira, or ticketing systems? (Yes = risk)
- Is your industry talking about "efficiency" more than "growth"? (Yes = risk)
- Has your team stopped backfilling when people leave? (Yes = risk)
Three or more Yes: treat your résumé and skills like a live project this month—not next quarter.
Which jobs face the highest risk—and what to do this week
Stop hiding behind the label "IT." Below is a practical map: job risk (from automation and headcount pressure) plus one concrete move you can make in the next seven days. Scores are our editorial judgment from news, vendor roadmaps, and field conversations—not destiny.
Roles, risk, and your next step
| Role | Sector | Job risk | What to do this week |
|---|---|---|---|
| Silicon validation / post-silicon debug | Semiconductor | Very low | You are scarce. Mentor a junior on lab bring-up; visibility beats panic. |
| Nurse (bedside) | Healthcare | Very low | Safe for now. Stack certifications that pay (critical care, languages). |
| Solution / enterprise architect | Cross-industry | Low | Schedule one customer workshop on AI governance—not slides, decisions. |
| Design verification (DV) engineer | Semiconductor | High | Own one corner case end-to-end (spec → closure). Document what the AI missed. |
| Junior RTL / IP integrator | Semiconductor | High | Volunteer for clock-domain crossing review on a real block—make yourself the niche owner. |
| Python backend (routine APIs) | Software | High | Design an API with auth, rate limits, and failure modes—show senior judgment. |
| Database administrator (routine tuning) | Cloud / enterprise | High | Map one production incident to root cause; autonomous DB clouds still need humans in war rooms. |
| QA manual tester | Software | Very high | Automate one flow in Playwright or Cypress. Put the repo link in your next stand-up. |
| Junior full-stack / CRUD developer | Software | Very high | Ship one feature with tests + observability. "I used AI" is not a skill; ownership is. |
| PHP / WordPress agency dev | Software | Very high | Rebuild one client plugin in Python or pick a vertical (payments, healthcare forms). |
| Technical writer | Cross-industry | Very high | Rewrite one doc set for a regulated audience (safety, finance). Accuracy beats volume. |
| L1 helpdesk | IT services | Very high | Close five tickets via script or KB article. Show tier-2 instincts. |
| BPO voice support | IT services | Very high | Move toward written async support or domain specialty (insurance, clinical). |
| Data entry clerk | Non-tech | Very high | Learn basic reconciliation or customer-facing coordination in your industry. |
Priya's team in Pune: verification without the glamour
Take Priya (name changed), a design verification engineer in Pune. Her group had twelve people in 2024. Today? Five. Nobody held a town hall about AI. When seniors left, the reqs simply disappeared. AI assistants now spit out testbench scaffolding overnight. Priya spends her days reviewing machine output and arguing about coverage holes—not writing tests from scratch. She says she feels less like an inventor and more like a supervisor for a very fast, very literal intern.
That is the chip-industry pattern Ian sees from the enterprise side too: verification and integration budgets are where EDA vendors pitch AI hardest. If your résumé is only "ran regressions," the next regression may not need you—but engineers who own closure on nasty failures still do.
Other silicon notes, briefly: physical design and DFT are script-heavy (high risk). Analog and formal verification sit in the middle. Field application engineers and post-silicon debug remain among the safest on the floor—customers still want a human when the board smokes.
Jobs that barely existed two years ago
Destruction dominates the news cycle. Creation is real but narrower—and usually requires retraining, not a weekend certificate.
Jobs shrinking vs jobs growing (2024–2030)
| Role | Type | Outlook |
|---|---|---|
| AI audit / fairness specialist | Emerging now | Banks, insurers—test models for bias and drift |
| Industrial prompt engineer | Emerging now | EDA, CAD, ERP workflows at chip and factory IT shops |
| LLMOps engineer | Emerging now | Monitor model cost, latency, regression in production |
| AI compliance officer | Emerging now | EU AI Act, sector rules, client contracts |
| AI / ML specialist | Growing (WEF outlook) | Among fastest-growing categories through ~2030 |
| Cybersecurity analyst | Growing | AI widens attack surface; humans still accountable |
| Robotics / automation engineer | Growing | Factories, logistics, warehouse systems |
| Care-economy & green-energy roles | Growing | Demographics and policy, not hype cycles |
Reality check: a manual tester cannot become an ML specialist in six weeks. Most new roles sit above the automation layer—governance, integration, accountability—not beside it on the same pay grade.
Retraining: honest time and money (India)
Articles love the word "reskill." Few say how long it takes or what it costs. Here is a blunt table for common pivots we hear about from training providers and hiring managers—not a guarantee.
Pivot paths (approximate)
| Current role | Realistic target | Min. time (months) | Cost (₹, approx.) |
|---|---|---|---|
| Manual QA | QA automation (Playwright / Cypress) | 3–4 | 50,000–80,000 |
| Junior RTL integrator | DV lead with AI-assisted flows | 4–6 | 80,000–1,20,000 |
| L1 helpdesk | Junior SOC or cloud support | 8–12 | 1,50,000–2,50,000 |
| PHP web dev | Python backend + basic data/AI literacy | 5–8 | 1,00,000–1,80,000 |
Pivoting is possible. It is neither quick nor free—and employers rarely fund it unless you are already on their short list.
Three countries, three different shocks
United States
CEOs name AI because Wall Street listens. Cuts cluster in corporate tech, support, and routine white-collar work. BLS payroll data through 2026 still show hiring in health care, leisure, and local government—so far this is a sector story, not a whole-economy collapse. Safer bets for now: licensed trades, bedside nursing, senior security incident response, data-centre electricians, and sellers who own a P&L line.
India
Beyond the MNC headlines, the bench is the signal: TCS −23,460 net FY26, Infosys −8,440 in Q4, Cognizant Project Leap, Oracle India in the March wave. Fresher pipelines are thinner—TCS guided ~25,000 campus offers for FY27 versus 40,000+ in peak years. Higher risk: manual testing, AMS maintenance, voice BPO, body-shop Java/.NET without domain depth. Lower risk: regulated delivery leads, bank cybersecurity ops, on-site client architects, healthcare IT with real compliance mess.
China
Beijing talks about employment stability while factories automate. The pattern is often reassignment, not headlines—at least for now.
In Shenzhen, BYD and other large manufacturers have been replacing warehouse inventory checkers with computer-vision systems since roughly 2024–2025, according to Chinese business press summaries. Workers were often moved to battery or assembly lines rather than marched out in one day. That is the Chinese model in miniature: automate the task, keep the payroll somewhere else in the plant, avoid a youth-unemployment scandal. Gig delivery, low-end content farms, and repetitive inspection still face the sharpest pressure. Roles tied to semiconductor self-sufficiency and EV supply chains look safer—policy-backed, capital-intensive, hard to offshore overnight.
What still buys you time
Four patterns recur in every country we looked at:
- Accountability. Someone signs the filing, holds the license, or faces the regulator.
- Scarce context. You know the fab quirk, the hospital politics, the client's audit history.
- Physical reality. Heat, weight, latency, a smoking board.
- Compound skills. DV plus security. Nursing plus geriatrics plus Spanish. Oracle ERP plus public-sector procurement.
Back to Bengaluru
Remember my friend in verification? He called again last week. The team is down to three. He is not panicking the way he was in February. He is learning emulation debug—hours on Palladium, chasing failures the regression AI cannot reproduce. "That part still needs a human with stubbornness," he said. "At least for this tape-out."
AI could hijack a lot of work starting in 2026. It will not hijack all of it on the same calendar. The question is not whether AI will touch your job. It will. The question is whether you will be reviewed by a machine—or be the human who reviews the machine's work.
Your turn
We can read SEC filings all day. Your desk is the ground truth.
- Has your team shrunk without anyone saying "AI"?
- Which row in our table is wrong for your city?
- Are DV engineers, PHP developers, or support agents actually safer where you work?
Leave a comment below. Mention your city and role if you can—Senthil and Ian read them.
Sources
- The Hindu — Xpheno active tech jobs outlook (June 2026)
- CNBC — AI cited in U.S. layoff announcements (June 2026)
- CNBC — Meta & Microsoft cuts (April 2026)
- BBC — Meta job cuts (April 2026)
- U.S. SEC — Oracle filings
- U.S. SEC — Cognizant Project Leap
- IMF — AI and employment exposure (2024)
- U.S. BLS — Employment Situation
- Times of India — Top five IT firms FY26
- Outlook Business — Fresher hiring (2026)
- World Economic Forum — Future of Jobs Report 2025 (press summary)
Disclaimer: Analysis and opinion, not career or financial advice. Salary, city, and retraining figures are directional estimates. Layoff announcements are not the same as completed separations.
Comments
Write your view after login. Your comment appears only after the moderator approves it.
No approved comments yet. Be the first to share a view.
Or use email and password below.