Same country, same month, two headlines that don’t seem to belong in the same story. Tata Consultancy Services confirms it cut 23,460 jobs in FY26, its largest reduction ever, and points to an “AI-first” delivery model as the reason. Weeks later, an eight-person research outfit called Sarvam AI closes a $234 million round that values it at $1.5 billion, with HCLTech alone writing a $150 million check for a stake. Both are true. Both are happening to the same workforce, in the same cities, at the same time. India isn’t being disrupted by AI so much as split by it, and which half of that story you land in depends on what you were doing five years ago.
The IT services model that built India’s outsourcing boom ran on billable headcount: more engineers on a client account meant more revenue, predictably, quarter after quarter. AI-first delivery breaks that math directly, and TCS isn’t alone in feeling it. Infosys has quietly cut over 950 jobs this year, shut a Pennsylvania back-office site, and trimmed campus recruitment, while the Nifty IT index is down more than 20% in 2026 as investors price in the same thing employees are living through. None of this arrives as a press release. It shows up as a “performance exit,” a bench that never gets refilled, a fresher offer letter that quietly doesn’t materialize.
Set that against what Rest of World reported this year from the other end of the same industry: a cluster of Indian labs building what they call frugal AI, models designed from the ground up for a market where bandwidth, hardware, and money are all constrained. Sarvam AI, founded by Vivek Raghavan and Pratyush Kumar out of AI4Bharat at IIT Madras, built SarvamM, a 24-billion-parameter model trained across ten Indian languages, and has since shipped open-weight models scaling up to 105 billion parameters. Krutrim, launched by Ola co-founder Bhavish Aggarwal, took a rockier path to the same goal, and Aggarwal has since admitted publicly that its early models weren’t close to global benchmarks. The design philosophy both are chasing borrows explicitly from Aadhaar and UPI: infrastructure built cheap enough and open enough that a billion people can actually sit on top of it. The problem it’s solving is concrete rather than aspirational. Tokenizing a query in an Indian language can cost roughly five times what the same query costs in English, which makes every American-built model an expensive and clumsy fit for the country’s own languages.
The government has decided this gap is worth subsidizing directly. The IndiaAI Mission has onboarded more than 38,000 GPUs through its compute portal, rented out to startups, MSMEs, and academic labs at roughly 42% below market rate, with an official target of 100,000 GPUs by the end of the year. Sarvam was the first startup selected to build a sovereign LLM under the mission. This is industrial policy aimed squarely at not repeating the outsourcing playbook: build the infrastructure domestically this time, rather than renting it from Redmond or Mountain View and providing the labor underneath it.
The workforce caught between these two stories isn’t abstract, and it’s the same demographic that has spent two decades building the diaspora’s professional identity around the words “I work in IT.” A Blind survey of over 1,200 verified tech professionals this year found that 53% had witnessed colleagues return to India from the US, with reverse migration highest among staff at Amazon, Walmart, and Uber. Higher H-1B fees and AI-driven restructuring at US tech firms are pushing that migration, and the landing isn’t soft: one returning Google employee described pay abroad at roughly a fifth of what it had been in the US. A 2025 thematic analysis of Indian IT professionals facing AI-driven displacement, published in a peer-reviewed occupational health journal, names the pattern plainly: emotional shock, erosion of professional identity, and what researchers term “perceived organizational betrayal” among workers who did everything the outsourcing economy asked of them.
What’s absorbing some of that shock is India’s own Global Capability Centre boom, on track to add over 510,000 jobs in 2026, growing 18-27% annually against 4-6% for legacy IT services, with nearly two-thirds of new roles now requiring AI or automation skills. It’s a real cushion, but not a return to the old bargain. A GCC job in Bengaluru pays a fraction of a US tech salary, and everyone taking one knows it.
So the honest answer isn’t “disrupted” or “next AI lab.” It’s that the generation whose parents’ career currency was proximity to American clients is now watching its own kids choose between prompt-engineering someone else’s model at a GCC desk, or building the sovereign one down the street from where TCS just handed out its layoff notices.
Sources
- Rest of World — 2026
- Business Standard — Indian IT stocks and Nifty IT index, June 2026
- Metaintro — TCS and Infosys AI reckoning, 2026
- Inc42 — Indian AI startup funding soars 4x YoY in H1 2026
- Business Standard — Sarvam raises $234M at $1.5B valuation
- Business Standard — India’s AI compute race and subsidised GPUs
- Abhishek Gautam — IndiaAI Mission: 34,000 GPUs at Rs 150/hour for startups
- American Bazaar — Laid-off H-1B workers return to India, face lower pay, July 2026
- PMC — Psychological impacts of AI-induced job displacement among Indian IT professionals: a Delphi-validated thematic analysis
- BusinessToday — India’s GCC hiring boom scales new peak, July 2026
SEO
- SEO title: Is India Still the World’s Back Office, or Its Next AI Lab?
- Meta description: TCS cut 23,460 jobs while Sarvam AI hit unicorn status. Inside the split identity of India’s IT-diaspora generation as outsourcing gives way to sovereign AI.
- Focus keyword: India sovereign AI outsourcing
- Secondary keywords: India IT layoffs AI, Sarvam AI Krutrim, IndiaAI Mission GPU, H-1B reverse migration India
- Slug: india-back-office-or-next-ai-lab
- Category: Culture & Diaspora
- Tags: sovereign AI, India IT layoffs, Sarvam AI, Krutrim, H-1B diaspora, IndiaAI Mission