AI Daily Brief: Reliance's Sovereign AI Bet, and the Anthropic Scare Rattling Indian IT

Three things happened this week that, read together, sketch the same fault line: who builds the AI that runs in India, and who's left exposed when someone else controls the switch. Reliance answered with infrastructure. Indian IT's frontier-model partners got a reminder of how exposed they still are. And a new hiring study quietly confirmed what that exposure is already costing entry- and mid-level engineers.

Ambani's answer: build the AI backbone yourself

At Reliance Industries' 49th AGM on Friday, Akash Ambani told shareholders that a sovereign AI backbone in Jamnagar is taking shape, with an initial fleet of Nvidia's newest GB300 GPUs equivalent to more than 75,000 H100s going live by the end of 2026, a number Reliance says could eventually scale past 200,000. The pitch isn't just capacity: "Reliance Intelligence will provide sovereign AI hosting within India, offering full model transparency and portability," Ambani said, language aimed squarely at enterprises wary of depending on infrastructure they don't control.

Mukesh Ambani framed it as a matter of national interest, not just corporate strategy: "India should not be a mere consumer of AI created elsewhere. It must become a creator, adopter, and a global leader in AI." On the consumer side, that means a clutch of India-first apps, Jio Bharat IQ, JioHealth IQ, JioLearn IQ, JioKrishi IQ, Jio's AI call assistant, and a smart-home layer called Jio Teleframe, all built across 22 Indian languages rather than English first. Reliance isn't walking away from the foreign labs either: Google's Gemini-powered AI Pro is already free for Jio users, and a new Meta tie-up will host Llama models domestically for Indian enterprises. Building your own backbone and renting someone else's frontier models aren't contradictory in Reliance's telling; they're the same hedge.

The Anthropic curb that revived India's sovereignty debate

That hedge looks more prudent given what happened a week earlier. A June 12 directive from the US government forced Anthropic to restrict access to its Fable 5 and Mythos 5 models for non-US citizens, including its own non-citizen staff, and since there was no practical way to enforce that selectively, the company pulled both models worldwide. The timing stung: it landed a day after Tata Consultancy Services announced it would train 50,000 employees on Claude and stand up a dedicated AI business unit around it. Infosys and HCLTech have made similar bets on Anthropic and OpenAI over the past year.

The episode reopened a fight that's been simmering in Indian tech circles for a while. Zoho's Sridhar Vembu used it to renew his case for sovereign AI self-reliance; Infosys chairman Nandan Nilekani has long argued the opposite, that India's edge is being the "use-case capital of the world" rather than building "one more LLM." What's changed is that the build-it camp now has real capital behind it: HCLTech is leading a $150 million slice of Sarvam AI's $300 million Series B, valuing the startup at $1.5 billion and making it India's newest AI unicorn, with HCLTech taking a 10.46% equity stake. It's the first time an Indian IT major has bought into a homegrown frontier-model company rather than just reselling someone else's. Vembu and Nilekani are still arguing in public; HCLTech just wrote a check that says which side it's hedging toward.

What AI is actually doing to the IT job ladder

While the boardroom debate runs on theory, a new study gives the clearest picture yet of what's happening underneath it. FindMyCollege analysed 37,553 active IT job postings collected in early 2026 against a 53,788-listing baseline from January-March 2024, and found total IT hiring down 30% over that period. The cut isn't falling evenly. The 3-to-5-year band, historically the largest single slice of Indian IT hiring, has fallen from 56.5% to 45.8% of postings, an 11-point drop in a single year. Fresher postings actually grew 25%, but the median fresher salary fell 19%, from ₹3.25 lakh to ₹2.62 lakh.

Look closer at the skill data and a more specific story emerges than "AI is coming for jobs." Postings demanding Python fell 32%, JavaScript 54%, C# 52%, Oracle 71%, yet pay for each of those skills rose 20 to 35% over the same period. That's not a market shrinking so much as one consolidating: fewer roles, doing more, paid more, while the easy entry-level rungs that didn't require much specialization quietly disappear. The one category growing outright, data and AI postings, is up 46% to 2,443 listings, still just 6.5% of the market. AI hasn't replaced the IT job market yet. It's hollowing out the middle of it while everyone waits to see if the new category grows fast enough to matter.

Markets and AI money: the India close

Friday gave a preview of how jittery investors are about that AI-and-IT story. Accenture trimmed its FY27 constant-currency revenue growth guidance to 3-4% (core commercial growth to 4-5%) and flagged AI as an increasingly meaningful demand driver, which the market read backwards: if AI is the demand driver, the old discretionary-spending pipeline Indian IT relies on is drying up faster than expected. Nifty IT fell more than 6% intraday, and the top ten listed IT names erased over ₹1.6 lakh crore in market value within minutes of the open.

Set against that, Sarvam's $234 million raise this week is a reminder that AI money in India isn't all flowing one direction. It's just flowing somewhere else: out of the listed services giants AI might disrupt, and into the smaller, harder-to-value bet that someone can build the alternative before that disruption fully lands.