AI Is Reallocating Tech Jobs, Not Deleting Them

Amazon confirmed a fresh round of cuts affecting fewer than 1,000 employees across customer service, marketplace support and retail engineering, with staff in the US, India and the UK impacted. It follows roughly 30,000 roles eliminated from late last year into January. The twist: the company is simultaneously reaching out to former employees, including some it laid off earlier, about open roles in AWS and its AI businesses, even as CEO Andy Jassy prepares a record $220 billion AI infrastructure spend. Internally, employees flooded Slack asking about severance, medical leave and whether an outside offer during the 90-day notice period forfeits their package.

Teksands take: This is the cleanest statement of the 2026 hiring model we have seen. Amazon is not shrinking; it is sorting. Retail support and legacy engineering get defunded, AWS and AI get funded, and the company is willing to rehire its own alumni to fill the funded side. For TA leaders the operational lesson is blunt: your layoff pool and your hiring pool are different populations. Do not assume a big-tech restructuring will solve your AI pipeline. It will solve your ops pipeline, and it will raise the bar you can set there.

H-1B registrations by TCS, Infosys, HCLTech, Wipro, LTIMindtree and Tech Mahindra plunged 92% for FY27 to 1,763 from 21,919 a year earlier. The Trump administration's layered restrictions — $100,000 fee proposals, a wage-weighted lottery, OPT and STEM-OPT fee proposals and tighter compliance scrutiny — are dismantling the traditional Indian IT onsite pathway. Indians took roughly 70% of the 406,348 approved H-1B petitions in FY2025, with computer-related roles at 62%. Senior, highly paid engineers are relatively insulated under wage-weighting; early-career workers and the volume-based staffing model are squeezed hardest.

Our take: A 92% collapse is not a slowdown, it is a demolition. The onsite pyramid that Indian IT built its pricing model on has lost its supply line, and the work is being pulled back into India. That is good news for Indian delivery depth and bad news for anyone hiring generalist engineers in Bengaluru, Hyderabad or Pune, because the same profiles that used to get shipped abroad are now competing locally. The selective part matters: wage-weighting protects senior talent and squeezes juniors. Screen for depth, not pedigree.

TCS hired 10,000 freshers in Q2. That is 4,000 fewer than Q1, and it is being called a strategy.

TCS onboarded about 10,000 campus graduates in the July-September quarter, down from roughly 14,000 in Q1, taking H1FY27 fresher intake to 24,000. More telling than the number is the language: CHRO Sudeep Kunnumal said the company is 'hiring to demand', with continued intake tied to specialised AI roles rather than a fixed annual fresher target. Headcount rose by a net 4,258 to 598,056, attrition eased to 13.3%, and annualised AI revenue crossed $3.1 billion, now over 10% of total revenue. Learning hours jumped 17% sequentially to 17.1 million.

For two decades, Indian IT ran on a predictable fresher pyramid: hire tens of thousands, train, deploy, bill. TCS is now openly converting to a demand-pull model, and AI revenue crossing 10% of the business is the reason. The fresher firehose is narrowing and what remains is skill-specific. If your campus strategy still assumes bulk intake and long bench cycles, you are planning for a market that no longer exists. Note the learning hours too — 17.1 million — because that is the real cost of the transition, and most mid-tier firms are not funding it.

Four signals worth watching

The lesson: This is a hiring story disguised as an immigration story. The six Indian IT firms named are the same firms that move thousands of engineers to US client sites every year. With PERM frozen, the green-card carrot is gone, and senior H-1B talent has less reason to stay in the US. Expect a measurable reverse flow into Indian tech hubs over the next two to four quarters. Smart TA teams should be building pipelines for exactly this cohort now — agentic AI, cloud and enterprise architecture skills, US client exposure, and suddenly open to India-based roles.

One door closes in the US, another opens in Jammu

Hiring for the funded side of the ledger

The roles being funded right now — AI engineering, forward deployed engineering, cloud and platform, cybersecurity, deployment and simulation work — are the hardest to assess and the slowest to fill. Teksands works with GCCs, IT services firms and product companies to build hiring motions for exactly these profiles: structured evaluation instead of keyword screening, and pipelines that hold up when the requisition is capability-driven rather than volume-driven.

Run a ten-person redeployment audit this week

Pick ten people in your organisation whose current roles sit on the defunded side of the market — support, manual QA, routine processing, legacy retail or ops engineering. For each, write down one funded role they could plausibly reach within two quarters and the specific gap between them and it. Then check whether your learning spend is actually pointed at those gaps. TCS logged 17.1 million learning hours in a single quarter while cutting fresher intake; that is what a demand-pull model costs to run. If your redeployment plan is a generic course catalogue rather than ten named people and ten named destinations, it is not a plan.