Finance & economics | Still Indians, actually India’s IT sector is surviving artificial intelligence Though the technology is making life still harder for many graduates Aug 6th 2026 IF THE HYPE about artificial intelligence is to be believed, India’s tech industry should bear the brunt of any incoming wave of AI job losses. Before the breakthroughs that made chatbots functional, tech types quipped that AI stood for “actually Indians”, as seemingly whizzy tech was really the work of humans in Hyderabad or Bangalore. But now the sort of tasks outsourced to India—software development and routine business processes such as payroll—are among those AI is best placed to take over from people. Yet evidence of job losses is not easy to come by. That is partly because data are patchy. Around 6m Indians work in information technology, according to Nasscom, the industry body. Almost anywhere else, that would be a vast number (roughly equal to the population of Denmark). But it is just a tiny

slice, around 1%, of India’s workforce. It is also a group that statistics agencies find hard to sample, says Rosa Abraham, an economist at Azim Premji University. IT workers are rich by Indian standards and often live in gated communities to which survey-takers cannot get access. Zealous security guards turn away those without prior approval. But it is also because AI is not making much of a dent. AI boosters can point to job cuts at India’s IT giants, including Tata Consultancy Services (TCS) and Infosys. For years employment at companies in the Nifty IT index, a group of listed big consultancies, grew by around 6% a year. Since the release of ChatGPT in late 2022 headcount has declined—from 1.71m in 2023 to 1.66m at the end of June, according to The Economist’s estimates. But that is not solely down to chatbots. Aggregate headcount is still 9% above the pre-pandemic trend (see chart), and the consultancies are pulling back after a hiring boom. The Indian operations of Capgemini, Cognizant and Accenture, three big foreign firms, display a similar pattern. Hiring in another form of Indian white-collar work is still booming. The number employed in global capability centres (GCCs), captive back offices handling everything from research and IT to HR and finance, has risen from 1.4m in 2019 to 2.4m this year, Nasscom estimates. Anthropic (yes, the AI darling), Costco, an American retailer, and L’Oréal, a French cosmetics

giant, are among the latest arrivals. Aggregate IT employment rose by half between 2018 and 2025; GCCs’ share of it went up from 35% to 40%. AI is, however, speeding up two shifts that were already under way. The first is from outsourcing to insourcing. Pari Natarajan, a consultant, says companies divide tasks between “core” and “context”. Context functions, such as making accounts payable or providing tech support, can be hived off to outsourcers and are easiest to automate. Core functions, however, are the business-critical sort from which a firm gets its competitive edge. As tech becomes more central to companies, more functions shift from context to core. GCCs have thrived as companies have come to see IT as fundamental rather than a mere supporting function. AI is adding to this. Mr Natarajan gives the example of a bank: when IT was handling payments, they could be outsourced; but if an AI is making lending decisions, it needs to be internal. AI coding tools are allowing companies to move more IT functions in- house: Starbucks, for instance, is developing its own version of the point-of- sale software it used to buy off-the-shelf. That work may fall to staff in its Indian GCC, due to open later this year. The second reinforced trend is the plight of recent graduates. Many have marched alongside the Cockroach Janta Party, a movement which has been protesting in Delhi about the lack of opportunities for the young. “What AI changes is the mix of talent we need,” says Vinod Kumar, chief people officer at Amtech, an American software company with a GCC in Bangalore. There is less demand for those workers with basic skills to do repetitive work. Instead there is a focus on engineers who can work effectively with AI. Naukri, a job portal, reports that in the past year overall demand for IT workers is down by 3%, while that for people with AI or machine-learning skills has risen by 25%. Even if India’s IT companies remain a source of opportunities for more experienced engineers, the underemployed young face intense competition for good jobs. The number of students in India has risen from 34m in the early 2010s to 45m in 2024. Entry-level salaries for computer-science graduates in IT consulting have been frozen at around 310,000 rupees ($3,200) a year for nearly 20 years.

Graduates have noticed, says Neeti Sharma of TeamLease, a recruitment company, and are upskilling. For those who do, the rewards are substantial: she reckons graduates from top Indian universities who secure a place on a specialised AI programme can earn twice as much as their peers studying generic computer science. AI has not eaten India’s software services yet. But it is eyeing it hungrily. ■ This article was downloaded by zlibrary from https://www.economist.com/finance-and-economics/2026/08/06/indias-it-sector-is- surviving-artificial-intelligence

Finance & economics | Buttonwood Investors in Situational Awareness deserved to lose their shirts The plight of the hedge fund says more about them than about its 24-year- old founder Aug 6th 2026 Readers of J.R.R. Tolkien’s “The Lord of the Rings” trilogy meet Frodo Baggins, its hero, when he is still in his tweens. This, Tolkien explains, is what hobbits call “the irresponsible twenties between childhood and coming-of-age at 33”. So Frodo’s cousin Bilbo waits until he has reached this ripe age before giving him a ring with the power, it turns out, to conquer the world. Any investor who entrusted Leopold Aschenbrenner with their money and is now cross with him should take note. Mr Aschenbrenner is in his mid-20s and runs Situational Awareness, a hedge fund—named without irony—

which posted a loss of 67% in July. He launched it in 2024 after writing a 165-page essay with the same title, arguing that artificial “superintelligence” would transform society before the decade was out. Situational Awareness (the screed) took Silicon Valley by storm. More surprising, investors responded by shovelling money to Situational Awareness (the fund). Within months Mr Aschenbrenner, then aged 22 or so and having never worked as a trader, was managing hundreds of millions of dollars. By July this year he was said to be overseeing as much as $45bn. Then many of the artificial-intelligence stocks in which Situational Awareness had taken positions cratered. On July 30th news broke that it had offloaded most of its listed shares to Citadel, a much larger investment firm, at lightning speed. Fund managers do not do this sort of thing unless they face a choice between a quick sale—inevitably at a hefty discount to market value—and a high chance of going to the wall. Mr Aschenbrenner wrote to his investors that he would “fight another day”. For everyone else, the question is whether his fall from grace portends something worse, since past stock-market busts have often been heralded by whizzy funds turning wobbly. Ken Griffin, Citadel’s boss, has form for snapping up bargains when they do. In the run-up to the global financial crisis of 2007-09, Citadel bought distressed portfolios from Amaranth Advisors and Sowood Capital, two hedge funds, just before each collapsed. A decade beforehand, the implosion of the star-studded Long-Term Capital Management (LTCM) fund came to epitomise the leverage-fuelled boom of the 1990s, which culminated in the dotcom bubble. What, then, to make of Situational Awareness’s plunge? Regulatory filings suggest the fund was dominated by exposure to firms providing the electricity and hardware that power AI, such as Bloom Energy, CoreWeave and Sandisk. All were doing stonkingly well until a few weeks ago. But between their recent peaks and Citadel’s intervention early on July 30th, each lost over half its value. Mr Aschenbrenner was hardly alone in such bets; over 80% of fund managers responding to Bank of America’s latest monthly survey named “long global semiconductors” as the most crowded

trade. The big worry is that this trade’s whiplash-inducing reversal will have set other investment firms teetering, too. Few, though, can have been run as recklessly as Mr Aschenbrenner’s. He reported returns of 439% for the first half of 2026. Astonishingly, this would leave the fund still sitting on a handsome year-to-date profit even after the past gut-churning month. It is simply not possible to make such a return without either enormous luck or borrowing so much that you risk bankruptcy. Leverage is what ruined LTCM in 1998, but comparisons between that fund and Situational Awareness are a stretch. In its first four years LTCM never lost money more than two months in a row. Its partners had been trading for Salomon Brothers, an investment bank, for two decades with similar results. Two of its directors had shared a Nobel prize for their work on pricing derivatives. LTCM was still wrong to borrow as much as it did, and its failure might have taken much of Wall Street with it. But its strategy wasn’t bonkers. Two decades before starting Situational Awareness, meanwhile, its founder was a toddler. And two years before, he was writing Tolkien-esque things like “the free world must prevail” and “I can see it. I can see how AGI will be built”. (Thankfully, everything that Buttonwood wrote in his early 20s has since been scrubbed from the internet.) Good for Mr Aschenbrenner for chancing his arm. The mystery is what on earth his investors were thinking. Perhaps they weren’t. In which case, that would be more worrying than anything about Situational Awareness’s humbling.■ This article was downloaded by zlibrary from https://www.economist.com/finance-and-economics/2026/08/04/investors-in- situational-awareness-deserved-to-lose-their-shirts

Finance & economics | Free Exchange How—and how much—should central banks talk? The art is to say enough, and no more Aug 6th 2026 THE ORACLE of Delphi had a gift for being right. Seated in Apollo’s temple, the Pythia delivered her prophecies in riddles. When Croesus, king of Lydia, asked whether he should attack Persia, she replied that he would “destroy a great empire”. So he did: his own. The philosopher Heraclitus described the Delphic method as one that “neither conceals nor reveals, but gives a sign”. Her commanding pronouncements demanded attention; their ambiguity made them almost impossible to prove wrong. For much of the 20th century central bankers cultivated a similar mystique. Their authority rested on the impression that they were omniscient and omnipotent. Since in truth they were neither, obscurantism was useful.