The first country targeted by AI-driven short selling has emerged.

On the Start of Spring in February 2026, San Francisco-based Anthropic quietly released a press announcement, launching an enterprise-grade AI tool.

Thousands of kilometers away in Mumbai, the K-line of India's Nifty IT index plummeted like a snapped kite, dropping nearly 6% that day. It was the most brutal massive negative candle the Indian market had witnessed since the pandemic-triggered circuit breaker in March 2020.

Three months later, OpenAI announced an investment of over 4 billion dollars to build a huge enterprise AI deployment team. Indian IT stocks immediately fell another 3.7% in response.

Over the past year, every cutting-edge release pushed out by Silicon Valley giants after hitting the enter key has triggered a precisely targeted earthquake in the Indian stock market.

In this market once known as a "safe haven for foreign capital", funds are fleeing with extreme determination. Throughout the first half of 2026, more than 23 billion dollars in foreign capital rushed out, dragging the foreign ownership ratio down to 14.7% — the last time such a dismal figure was recorded was 14 long years ago.

The Nifty IT index, hailed as India's version of the "Hang Seng Tech Index", has been drifting downward for 18 months, with a cumulative pullback of 49%, nearly halving its value. The ten largest IT giants have silently evaporated more than 19 trillion rupees on their balance sheets.

This sum is roughly enough to cover 40% of India's entire national fiscal budget last year.

But there is never a crash without a cause in this world.

Over the past three decades, India has reaped the full benefits of its demographic dividend by answering calls from global clients and patching low-level code, but it has also unknowingly stood on the opposite side of technological progress.

When the price tag per unit of Token has become cheaper than human labor along the Ganges River, the outsourcing assembly line that once supported countless Mumbai's middle class has suddenly lost all meaning in the face of cold computing power.

India, the country with the most abundant carbon-based human resources on Earth, not only failed to seize the incremental dividends of the silicon-based era, but also became the first unlucky one to be harvested by the AI scythe.

01 The End of the 20-Year "Labor Arbitrage Game"

To understand India's current pain, we must first understand how it succeeded in the past. The rise of Indian IT outsourcing, at its core, can be traced back to a bug called the "Millennium Bug".

In 1999, the old systems of the financial, aviation, and power sectors in Europe and the United States, which used two digits to record the year, faced the brink of system collapse. Western enterprises urgently needed a massive number of programmers to inspect and modify these outdated legacy codes. This work was not highly technical, but the workload was extremely overwhelming — it was essentially "digital-era brick-moving".

Indians keenly spotted this opportunity. Leveraging their three major advantages of "good English, low wages, and the ability to work overnight", a group of Indian IT enterprises such as TCS, Infosys, and Wipro rose rapidly, firmly establishing India as the "World's Back Office".

According to a 2025 report from the National Association of Software and Service Companies (NASSCOM) of India, the size of India's outsourcing industry has surged to an astonishing 2800 billion US dollars, directly supporting 5.67 million IT engineers.

If we count the families behind each engineer, as well as the catering, logistics, and property services that revolve around them, this entire industrial chain supports the livelihoods of nearly 25 million middle-class people in India.

More critically, this is India's only pillar industry that can earn foreign exchange on a large scale. The combined exports of IT services and BPO account for nearly a quarter of India's total exports of goods and services.

However, if you break down the business model of Indian outsourcing, you will find it extremely simple, even fragile: charging by headcount, billing by the hour.

A US programmer has an annual salary of 150,000 dollars, while an Indian engineer earns 15,000 to 20,000 dollars a year; a US customer service representative earns 40,000 dollars annually, while an Indian counterpart only makes 6,000 dollars.

Indian outsourcing companies accept orders at prices far lower than those in Europe and the United States, then send the work back to their home country to pay local wages, profiting from the labor price difference in between.

This was a perfect arbitrage game that Indians played for 20 years, reaping huge profits — until the emergence of AI overturned the entire table.

A 2025 study by researchers from Carnegie Mellon University and Stanford University delivered a fatal blow: AI agents can complete tasks 88.3% faster than humans. On the cost side, the median salary for engineering and data roles in India in 2025 was 22,000 dollars, but the annual subscription cost for AI programming tools only ranges from a few hundred to a few thousand dollars.

When a tireless AI that does not need social security contributions can write code 88% faster than you at a fraction of your cost, the foundation of the "World's Back Office" collapses instantly.

The chill has spread to every Indian programmer's workstation.

On April 29, 2026, global IT service giant Cognizant officially launched its transformation plan codenamed "Project Leap", setting aside 200 to 270 million dollars just for severance pay. Although specific figures were not disclosed, media reports revealed that 12,000 to 15,000 employees worldwide would be laid off, the vast majority of whom are in India.

This is not an isolated case. US real estate technology company Opendoor directly closed all its offices in Chennai and Bengaluru, India; French pharmaceutical firm Sanofi simply transferred the audit of procurement orders previously handled by Indian outsourcing teams to SAP's AI Agent.

The impact is also reflected in financial reports. Industry leader TCS saw its US-dollar-denominated revenue fall to 30 billion dollars in FY26, a 0.5% year-on-year decline at constant exchange rates, marking its first annual negative revenue growth in many years; Wipro's full-year revenue was only 10.5 billion dollars, down 1.6% year-on-year at constant exchange rates, almost stagnant.

Even for the most resilient player Infosys, although its revenue exceeded the 20 billion dollar mark for the first time, its constant exchange rate growth rate was only 3.1%, far lower than the 13.7% compound annual growth rate of the past decade.

The wave of layoffs is clearly visible. In 2025, around 245,000 tech layoffs occurred globally, with India ranking second with 19,000 layoffs. It is worth noting that India accounts for far less than 7% of global tech employment, but contributed 7.8% of the world's layoffs.

What is even more terrifying is the reversal of trends. In the previous fiscal year (FY25), India's top 5 IT companies still added a net 12,718 employees; by FY26, these 5 companies had a combined net reduction of 6,981 employees. TCS alone cut more than 23,000 jobs, with its total employee count dropping from a peak of 614,000 to below 580,000. The last time TCS saw such a large-scale net layoff dates back to the 2008 global financial crisis.

A country that built its middle-class dream on writing code is being ruthlessly pulled back to reality by AI.

02 Why Did India Fail to Stay in the Game?

With their old jobs taken away, logically speaking, India, which has so many technically skilled engineers, could have easily turned to grab new opportunities in the AI era. But the reality is that India did not even get a seat at the "cake-splitting table".

US asset management firm Altimeter estimated that the global AI net profit would reach 637 billion dollars in 2026. The US took 49%, South Korea took 35%, and the two countries together captured 84% of the global AI profit.

The remaining 16% was divided among regions such as Chinese Taiwan, mainland China, Japan, and Europe. India's name is nowhere to be found on this long list of profit-sharing parties.

Many people simply attribute India's missed AI opportunity to insufficient policy investment or lack of computing power, but these are only superficial phenomena. The real problem is deeply rooted in India's industrial path over the past decades.

Looking back at the last century, Japan, South Korea, and China all went through arduous industrial upgrading. Japan's path evolved from low-cost automobiles to high-quality automobiles, and then to semiconductor materials; China's path moved from contract manufacturing to consumer electronics, and then to internet products and AI.

You will find that every step of East Asian countries was about "building things". What about India's path? IT services, IT services, and nothing but IT services. India directly skipped the industrialization stage, "leapfrogging" to develop the service industry.

This is not because Indians are inherently averse to opening factories, but because their own institutional framework has tightly constrained them. After gaining independence in 1947, India implemented a bizarre "License Raj" system — any region that wanted to build new factories, expand production, or even adjust product lines had to apply for approval from the central government. This system essentially protected vested interests, making it impossible for new entrants to obtain licenses.

By the time the restrictions were finally relaxed in 1991, East Asian countries had already divided up the entire low-end manufacturing market. In the 1990s, when foreign exchange was extremely scarce, India was "forced" to enter the global IT outsourcing business.

Having missed the global manufacturing industrial chain division, India naturally became a spectator in the wave of AI infrastructure construction.

If hardware manufacturing is not feasible, what about the software model track? It is also difficult to achieve breakthroughs. This logic is consistent with why India missed the internet era.

In the internet era, the US gave birth to Google, Amazon, and Meta, while China produced Alibaba, Tencent, and ByteDance. What about India? The country with the largest number of programmers in the world only ended up nurturing a bunch of outsourcing companies.

The core reason is that India's domestic market cannot form an experimental field for product iteration. A successful software company needs to go through the flywheel of "large domestic market → scale effect → product iteration → globalization".

India seemingly has 969 million internet users, but their spending power is extremely low. At present, India's per capita GDP is only 2,800 US dollars, and wealth is highly concentrated, with around 228 million people still living below the poverty line.

The vast majority of Indian users have limited payment capacity, which determines that Indian internet companies cannot easily replicate the growth path of relying on domestic market scale that US and Chinese companies have followed.

Large technology companies in the US and China usually complete commercial validation in their domestic markets first, then expand overseas.

Take the US as an example. SaaS giants usually first find a large number of high-paying customers in the domestic market before expanding overseas. Large US enterprises are not only willing to purchase software, but also act as early customers, helping startups polish their products repeatedly.

The situation in India was the opposite in the past. A large number of local enterprises have low levels of digitalization, are highly price-sensitive to software, have fragmented procurement processes, and are more accustomed to customized development and manual services. As a result, Indian SaaS companies often have to directly compete in the US market before completing product validation domestically.

This also explains a peculiar phenomenon: India's SaaS ecosystem has flourished, with nearly 20,000 companies accounting for one-fifth of the global total, but very few of them have grown into 10-billion-dollar giants.

A market that cannot even sustain asset-light internet platforms has no reason to support large model development that often burns billions of dollars in investment.

Therefore, India's technology giants have simply not invested in AI.

The five leading outsourcing companies including TCS and Infosys occupy the top five positions among Indian IT firms, with a combined market capitalization once exceeding 500 billion dollars, but their R&D investment is surprisingly low.

In the 2008-2009 fiscal year, TCS's R&D expenditure only accounted for 0.2% of its revenue, while Wipro's was 0.19%. Fifteen years later, this ratio has barely changed. In FY25, TCS's R&D ratio was 1%, while Infosys and Wipro both stood at 0.5%. In comparison, Microsoft's R&D ratio is 12%, Google's is 14%, and Meta's is as high as 25%.

Where did all these profits go? The answer is: returned to shareholders.

In the 2020-2025 fiscal year, the top 5 IT companies returned about 4.8 trillion rupees to their shareholders, accounting for 87% of their combined net profit. This dividend payout ratio exceeding 80% is almost unique among global technology companies.

The underlying helplessness is that two mountains — domestic conglomerates and Wall Street — are pressing down on them.

Take TCS as an example. Its parent company, the Tata Group, operates a large number of heavy-asset, highly cyclical, and even perennial loss-making businesses such as steel and automobiles, which employ hundreds of thousands of people who cannot be laid off. TCS must maintain an extremely high dividend payout ratio, continuously channeling its earned US dollars back to the parent group to fill financial gaps.

On the other hand, Wall Street prices these outsourcing companies as "bond-like" assets with high dividend yields. If they dare to pour profits into uncertain large-model R&D, capital will immediately vote with their feet.

Outsourcing giants must distribute profits to their shareholders, and the capital market favors stable cash flow, granting them higher and higher valuations. In this self-reinforcing vicious cycle, no one has the incentive or dares to take risks for innovation.

At the end of the day, India's "absence" in the AI era is an inevitable outcome of its industrial path development.

03 When Cheap Labor Becomes a Burden

Having said so much, what exactly is the development status of India's AI industry?

As of the first half of 2026, there are only three recognized AI unicorns in India.

Sarvam AI is the only company that is genuinely working on foundational models. In June 2026, it completed its Series B financing at a valuation of 1.5 billion US dollars, but its revenue in FY26 was a mere 5.4 million US dollars.

Krutrim once claimed to compete with OpenAI, but within two years it removed its AI assistant from the market, suspended chip R&D, drastically downsized its team, and pivoted to selling AI cloud services, with 90% of its revenue coming from related-party transactions within its parent group.

The third company, Neysa Networks, operates in the computing power rental business.

The combined valuation of these three unicorns developed with national efforts is less than 4 billion US dollars. In the global AI frenzy, this scale is not even enough to qualify as a market participant.

The latest forecast from the International Monetary Fund (IMF) has officially validated this transformation: it downgraded India's GDP growth rate for the 2026-27 fiscal year to 5.8%, marking the largest downgrade since the pandemic. The only core reason cited is the permanent contraction of the country's pillar IT service exports.

This is not just India's story; it is a brutal economic metaphor.

When an economy has long regarded "cheap labor" as its core competitiveness and built a massive interest distribution system around it, it is essentially betting against technological progress. The faster technological change occurs, the more severe the backlash it will suffer.

With the world's largest pool of cheap labor, India has instead become the most heavily burdened economy in the AI era. The country with the largest number of programmers has thus been locked in the twilight of the old era.

This article is from the WeChat official account "Silicon-Based Observer Pro", authored by Silicon Observer, and published by 36Kr with authorization.