Labour market in times of technological changes

The impact of AI on the economy, especially on the labour market, is most likely to be the defining political economy issue of this generation. 

In this context, it is useful to understand the trajectory of the evolution of jobs with technological changes over the last century or so. Daron Acemoglu and Pascual Restrepo have shown that roughly half of America’s employment growth between 1980 and 2010 came from the creation of entirely new occupations. They argue that the automation effect of the displacement of workers is offset by the reinstatement effect arising from the creation of new occupations. 

David Autor, Caroline Chin, and Anna Salomons have a paper which examined the substantive content of emerging job categories (or new work) over the 1940-2018 period in the US, where it comes from, and its effect on labour demand. Augmentation innovations are those that increase capabilities, quality, variety, or utility of the outputs of occupations, thereby generating new demands for worker expertise and specialisation. They constructed a database of new job titles linked both to US Census microdata and to patent-based measures of occupations’ exposure to labour-augmenting and labour-automating innovations. 

We find, first, that the majority of current employment is in new job specialties introduced after 1940, but the locus of new work creation has shifted—from middle-paid production and clerical occupations over 1940–1980, to high-paid professional and, secondarily, low-paid services since 1980. Second, new work emerges in response to technological innovations that complement the outputs of occupations and demand shocks that raise occupational demand; conversely, innovations that automate tasks or reduce occupational demand slow new work emergence. 

Third, although flows of augmentation and automation innovations are positively correlated across occupations, the former boosts occupational labour demand while the latter depresses it… Employment and wagebills grow in occupations exposed to augmentation innovations and contract in occupations exposed to automation innovations… augmentation innovations increase occupational wagebills by boosting both employment and wages suggests that ‘new work’ may be more valuable than ‘more work’—plausibly because new work demands novel expertise and specialization that (initially) commands a scarcity premium… we establish that the effects of augmentation and automation innovations on new work emergence and occupational labour demand are causal. Finally, our results suggest that the demand-eroding effects of automation innovations have intensified in the last four decades while the demand-increasing effects of augmentation innovations have not.

Routine task-intensive occupations gained substantial new titles between 1940-80, and few between 1980-2018.

What most stands out from this figure is the shifting fortunes of routine task-intensive occupations—both blue-collar occupations such as operative and kindred workers, metal workers, and mechanics; as well as white-collar occupations such as shipping and receiving clerks; stenographers, typists, and secretaries; bank tellers and bill and account collectors; and library attendants and assistants. 

So what does this all mean for the labour market in the times of AI? In a much discussed essay, Dario Amodei, CEO of Anthropic, sounds alarm that AI could displace half all white collar jobs in 1-5 years.

The pace of progress in AI is much faster than for previous technological revolutions. For example, in the last 2 years, AI models went from barely being able to complete a single line of code, to writing all or almost all of the code for some people—including engineers at Anthropic. Soon, they may do the entire task of a software engineer end to end… it just implies the short-term transition will be unusually painful compared to past technologies, since humans and labor markets are slow to react and to equilibrate... AI will be capable of a very wide range of human cognitive abilities—perhaps all of them. This is very different from previous technologies like mechanized farming, transportation, or even computers. This will make it harder for people to switch easily from jobs that are displaced to similar jobs that they would be a good fit for… 

AI is increasingly matching the general cognitive profile of humans, which means it will also be good at the new jobs that would ordinarily be created in response to the old ones being automated… Across a wide range of tasks, AI appears to be advancing from the bottom of the ability ladder to the top. For example, in coding our models have proceeded from the level of “a mediocre coder” to “a strong coder” to “a very strong coder.” We are now starting to see the same progression in white-collar work in general… AI, in addition to being a rapidly advancing technology, is also a rapidly adapting technology… Early in generative AI, users noticed that AI systems had certain weaknesses… But pretty much every such weakness gets addressed quickly— often, within just a few months.

However, in a recent issue, The Economist disputed such alarming prognostications.

We analysed employment and wage trends across more than 100 large white-collar occupations in America since the second half of 2022. Employment across the sample has risen by 4% and real wages by 3%. To get a sense of AI’s impact on different roles, we used occupational descriptions to classify white-collar roles into four groups depending on the bundles of tasks involved: technical specialists, managers and co-ordinators, care workers, and back-office employees. We then tracked employment in each group starting in late 2022, using six-month moving averages. 

Roles that combine technical expertise with oversight and co-ordination have enjoyed the biggest gains. Employment among project managers and information-security experts has risen by 30% or so. Other occupations which combine deep expertise in maths-related fields with problem-solving are also thriving. So are jobs which involve interpersonal care work and those which demand judgment and co-ordination. Only routine back-office work has shrunk. Over the past three years or so the ranks of American insurance-claims clerks have shrunk by 13% and those of secretaries and admin assistants by 20%.

It also finds AI generating all-new jobs - data annotators, forward-deployed engineers, chief AI officers, and mainly those without settled names (“other occupations”). 

On the impact of AI on jobs, the Yale Budget Lab has a meta-study that compares evidence from across studies. It uses seven different measures of occupation-level AI exposure calculated by researchers and urges caution in reading too much into the findings. These measures are based on human and AI assessments of whether a job’s constituent tasks can theoretically be performed by an LLM, linking tasks with AI-related patents, and using real-world data on how LLMs are being used to carry out particular work-related tasks. 

Its headline findings:

AI exposure metrics broadly agree with each other, but they disagree with each other more on highly exposed occupations. The key point of disagreement between different AI exposure metrics is in the magnitude of exposure, not whether an occupation is exposed. Occupational exposure to AI is not indicative of a jobs AI will automate out of existence. Rather, it indicates places in the labor market where AI could have an impact.

The study uses each occupation’s exposure and variance across the various scores, with low variance pointing to a consensus on exposure. They regressed the two and found greater disagreement for highly exposed occupations, “driven more by how much an occupation is exposed more than whether it is exposed.”

Clearly, occupations focused on computational, text-based, or administrative work tend to have both higher variance and higher average exposure, whereas manual fields like construction and maintenance have lower disagreement and variance. 

The high degree of variance and disagreements point to the perils of forming opinions on the trajectory of AI’s evolution and its impact on the labour market. The meta-study urges caution in drawing conclusions “about where AI disruption to the labor market could be going”. 

John Burn-Murdoch and Sarah O’Connor in the FT point to a few perspectives that are important while considering the impact of AI. Specifically, they point to the nature of the AI exposure and the regulation of AI adoption. 

It is more than 20 years since David Autor and his co-authors argued convincingly that, like the waves of technological change that came before, computerisation threatens jobs where workers are mainly performing tasks to meet a specification, but is a complement to those who determine the specification. Viewed through this lens, the AI revolution may pose less risk to (or even benefit) a software developer who exercises considerable autonomy over what they work on and how they do it, than to a warehouse worker who loses out to a new generation of AI-enhanced robots, or a retail sales assistant whose store is closed as technology drives ever more commerce away from brick and mortar stores and onto the web…

Or consider the role of regulation. As we have written previously, AI models can now evaluate medical scans more accurately than experienced radiologists, but regulatory barriers and insurance policies have made it virtually impossible for fully autonomous systems to be used. Meanwhile, laws have sprung up across the US prohibiting AI tools that “provide services that constitute the practice of professional mental or behavioral healthcare (such as therapy)”. Whatever one’s views on the rights or wrongs of these particular cases, they are clear demonstrations that vulnerability to occupational displacement in the age of AI comes down to far more than “Is AI capable of performing the tasks that make up your job?”

In conclusion, on the overall likely impact, the central questions revolve around three trends - labour automation (associated displacement), labour augmentation (and associated redeployment), and the emergence of new work categories. What will be the relative impacts of the three? Will the first far exceed the second and third? Or will they offset the first? What will be the pace of the first? Will the second and third lag the first significantly?

It is impossible to answer any of these with a degree of confidence. We may only be able to wait and watch how the trends play out and respond accordingly. 

Weekend reading links

1. China is rising up the ladder on higher education and research.

In 2010, only one mainland Chinese institution ranked in the top 50 of the QS World University Rankings, a closely watched global league table. By 2025, that number had risen to five, and they were positioned higher up the table... A report by the US-based Center for Security and Emerging Technology found that in 2019, a group of 10 elite Chinese universities each had a budget exceeding $5bn a year... Holden Thorp, editor-in-chief of the Science family of journals, says 14 per cent of papers accepted in Science in 2025 were from China, the second-largest share after the US, at 45 per cent... As China has climbed the rankings, many critics have pointed to the industrial scale of fraudulent or poor-quality research, driven in part by incentives that reward publication volume in tenure and promotion decisions... In 2025, Ivan Oransky, co-founder of Retraction Watch recorded nearly 3,000 retractions of Chinese-authored papers from journals, compared with 177 for US authors.

And this on R&D spending

China is close to surpassing the US in total expenditure on R&D, with China spending $781bn and the US $823bn in 2023, according to data from the OECD. By contrast in 2007, China spent $136bn compared with $462bn by the US. The OECD also calculates that China spends, on average, $305,000 on R&D costs per researcher, which is more than the European average of $268,000.

2. Ed Luce has a good description of the war in Middle East.

The war threatens to turn into a contest over which can hold up longer — Iran’s ability to produce drones versus America’s capacity to intercept them.

3. Emmanuel Macron has outlined a new more aggressive and collaborative nuclear strategy for France.

Laying out a new concept, “forward deterrence”, he offered to extend French protection “into the depth of our continent”. Successive French presidents since Charles de Gaulle have consistently alluded to the “European dimension” of France’s “vital interests” covered by its independent force de frappe. Macron has now suggested concrete steps to materialise this. These include the temporary deployment of French nuclear-armed aircraft, the Rafale, to allied countries, as well as participation in deterrence exercises. The conventional forces of allies may also take part in France’s nuclear activities. Macron mentioned seven — Germany, Poland, Sweden, Denmark, Greece, Belgium and the Netherlands — with whom this strategic dialogue has already started. This is all new and quite remarkable. Together with the Northwood Declaration, which committed France and the UK in July 2025 to an unprecedented level of co-ordination in nuclear policy, Macron’s speech signals a major development in the history of European defence.

4. Insurance premiums rocket 12-fold on ships sailing through the Strait of Hormuz. 

Premiums jumped as high as 3 per cent of the cost of a ship on Wednesday, up from about 0.25 per cent before the war... Typical prices in the high-risk region now ranged from 1 to 1.5 per cent of the cost of a ship, while ships linked to the US, UK and Israel had been quoted prices as much as triple those rates, Marsh broker Dylan Mortimer told the FT.

5. On Nvidia's staggering 75% gross profit margin.

Nvidia makes its money selling physical chips that must be manufactured in fabrication plants the company does not own. Nvidia’s most advanced AI chips rely heavily on fabrication by Taiwan Semiconductor Manufacturing Company. Its most profitable products, including the H200, Blackwell and the next-generation Rubin architecture, are made on TSMC’s advanced 4 and 3 nanometre production processes. There is currently no alternative capable of manufacturing those designs at the same scale, performance and yield. In chipmaking, whoever controls advanced manufacturing has the greatest leverage. TSMC decides how much it charges for each wafer and how much advanced production capacity each customer receives... Nvidia designs the architecture and the software. TSMC builds the chips and spends more than $40bn each year expanding and upgrading fabrication capacity. Nvidia enjoys software-like margins without bearing the cost of constructing and upgrading factories. TSMC keeps its most advanced factories running at full capacity. For now, that balance has held. Nvidia’s valuation, however, assumes it will continue to hold... At current revenue levels, every one-point move in gross margin would represent about $2bn in annual gross profit, which would be enough to move earnings forecasts... Today, the chip industry revolves around two critical bottlenecks: ASML in advanced lithography equipment and TSMC in advanced chip manufacturing. Credible alternatives are rare and the barriers to entry are vast. Nvidia’s advantage sits above manufacturing, in chip design and in the software ecosystem built around its processors.

6. Brilliant essay by Dean Ball, a former White House policy adviser who wrote the Trump administration’s A.I. strategy, on the consequences of the US Government's decision to cancel the Department of War's contract with Anthropic to use its AI system Claude in classified contexts and designate the firm a "supply chain risk". Specifically, the US DoW objected to two restrictions in the contract with Anthropic (a contract that was renegotiated with these restrictions by the current Trump administration) that Claude could not be used for mass surveillance on Americans and it could not be used to control lethal autonomous weapons which are can identify, track and kill targets with no human in the loop at any point in the process. 

The Department of War’s rational response here would have been to cancel Anthropic’s contract and make clear, in public, that such policy limitations are unacceptable... War Secretary Pete Hegseth has gone even further, saying he would prevent all military contractors from having “any commercial relations” with Anthropic... Essentially, the United States Secretary of War announced his intention to commit corporate murder... the message sent to every investor and corporation in America: do business on our terms, or we will end your business. This strikes at a core principle of the American republic, one that has traditionally been especially dear to conservatives: private property...

This threat will now hover over anyone who does business with the government, not just in the sense that you may be deemed a supply chain risk but also in the sense that any piece of technology you use could be as well... No entity with meaningful ties to government business would use DeepSeek, simply because the regulatory risk was too high. Now that the government has applied this regulation to an American company, the regulatory risk simply exists for all software... this could end up making AI less viable as a profitable industry... Simply for having different ideas, expressing those ideas in speech, and actualizing that speech in decisions about how to deploy and not deploy one’s property. Each of these things is fundamental to our republic, and each was assaulted.

7. The economics of Iran-Israel/US war.

The United States is dominating the skies above Iran. But math is not necessarily on America’s side. Iran is using low-cost drones for precision attacks in the Middle East. The United States and its allies have air defense systems capable of intercepting a vast majority of Iranian ballistic missiles and drones, which are sophisticated yet costly... The cost ratio per shot, per interception, is at best 10 to one. But it could be more like 60 or 70 to one in terms of cost, in favor of Iran... Iran’s Shahed drones are triangle-shaped loitering munitions, roughly 11 feet long... They are small enough to be launched from the back of a truck, making them relatively easy to hide and tough to hunt down. The long-range version of the Shahed drone, known as the 136, can travel roughly 1,200 miles... Built with off-the-shelf commercial electronics, each Shahed is said to cost $20,000 to $50,000 to manufacture, depending on the model... The gold standard in missile defense, the Patriot air defense system, uses interceptors that can cost more than $3 million per shot and are in limited supply. For instance, Lockheed Martin delivered just 620 PAC-3 interceptors in 2025, which broke a record for production.

8. One of the biggest casualties of the war is UAE, Dubai and Abu Dhabi. This about Dubai.

Since the US and Israel launched the war a week ago, the United Arab Emirates, which for years has enjoyed spectacular success as a global entrepot, has been the target of about two-thirds of all ordnance fired by Iran across the Gulf. For years, the UAE’s brand — and that of Dubai in particular — was underpinned by its claim to be an island of stability in a dangerous neighbourhood. Tech billionaires, influencers and holidaymakers alike were pulled in by factors ranging from favourable tax treatment to winter sunshine and a location where east meets west, convenient for Europe, Africa and Asia alike... Iranian attacks have also peppered military, infrastructure and energy targets in Saudi Arabia, Bahrain, Kuwait and Oman, confirming long-held fears that Tehran would lash out at its US-allied neighbours if the regime’s survival was at risk... The UAE’s multi-layered defence system’s interception of 93 per cent of more than 1,100 incoming missiles and drones has limited casualties and damage...

Transforming itself from fishing village to regional trade hub in the 1970s, the rise of Emirates airline — linking cities across the continents through its ever-expanding airport — kick-started a tourism industry that made Dubai the world’s sixth most visited city last year. In the aftermath of the 9/11 attacks in the US, funds from Muslim-majority nations flowed to the Gulf. Dubai opened its property market to foreigners, fuelling the city’s first real estate boom. The global financial crisis rattled the region in 2009, shrinking Dubai’s debt-strewn economy and prompting bailout loans underpinned by Abu Dhabi. But the influx of money and people precipitated by the Arab popular uprisings of 2011 boosted its economy further. When the coronavirus pandemic struck in 2020, the government locked down harder and reopened faster than others, fostering a relaxed, safe environment that attracted a new generation of newcomers: social media influencers, cryptocurrency investors and hedge fund managers. The war on Ukraine in 2022 brought Russians seeking sanctuary, while higher UK taxes lured a wave of wealthy residents and long-term residency programmes incentivised all foreigners to put down roots.

9. Formula 1's spectacular growth since its takeover by Liberty Media.

F1’s accounts for 2025 show a sport in rude health. Annual operating profit rose 28 per cent to $632mn as revenues — across media rights, sponsorship, fees from promoters and hospitality — increased by 14 per cent to $3.9bn. In 2017, when Liberty Media acquired the sport, F1 made a $37mn operating loss on revenues of $1.8bn. Big brands continue to flock to Formula 1... F1 now has 10 global partners, versus four only four years ago. Sponsors are also flooding to racing teams... The most lucrative partnerships can make more than nine figures for F1 teams. Global appetite for attending races is stronger than ever. Tickets for this weekend’s opening race in Melbourne sold out in minutes when they went on sale in September, setting the tone for the rest of the season. Last year, 6.75mn fans attended races, up 4 per cent from the year before... Bankers and financial analysts think there is still plenty of room for further growth. According to data from sports news group Sportico and investment bank Houlihan Lokey, F1 teams were valued at around 6.1 times revenues last year, compared to 10.3 times for franchises in the NFL and 11.9 for those playing in the NBA.

This comes at a time when Formula 1 has introduced sweeping changes to the cars from this year.

Owing to new rules introduced by the sport’s governing body, cars this year will all be shorter, narrower and lighter, wings will be simpler and aerodynamic redesigns will result in flatter vehicle floors, which F1 says will increase the scope for different driving styles. While engines have been hybrid for years, the balance of power between petrol and battery has been altered significantly, meaning that cars will now rely on electric power around 50 per cent of the time. The changes have been brought in for two main reasons. One is to shift F1’s approach to sustainability by leaning more into battery power and using so-called e-fuel made from carbon capture, municipal waste and non-food biomass or a combination. The other is to attract more car companies into the sport. The new engines will be “more road-relevant”, F1 says, pointing out that Ford, Audi and, from 2029, General Motors have all been enticed into producing engines as a result of the reforms.

Two new teams, Cadillac and Audi, are entering the race this year.

Some thoughts on startup innovation scaling - hospital solutions

The Ken has an article on how the health systems in India are adopting AI applications, specifically ambient AI transcription apps (always-on AI systems that use contextual interpretation to transcribe speech without explicit prompts). The article highlights several important insights about not only AI-adoption but also generally startups in India. 

The idea is simple: use AI transcription tools as scribes to document patient consultations and integrate them into the patient and hospital management workflows, thereby improving efficiencies and quality of care. Besides, “ambient AI could become the layer on which a full AI stack in diagnostics, predictive health, and ICU optimisation” can be built. 

Apart from the inherent productivity-enhancing value of a digital scribe, the felt need in India is the sheer volume of patient load faced by doctors. An Indian doctor sees, on average, 30 patients compared to three for the US doctor.

This is also because India has a doctor for every 811 people, rising to nearly 11,000 in rural areas, compared to one for 300 people in the US. 

This patient load has naturally led to the search for methods to optimise consultations, especially by adopting ambient AI scribes. The well-heeled hospital chains have preferred to use the mature foreign solutions instead of relying on Indian startups. 

The article describes the challenges faced by ambient AI scribe startups in India.

Most hospitals in the country do not have electronic health records, known as EHR, that can integrate such tools… Where EHR systems do exist—mostly in private hospital chains—they aren’t standardised, making the integration of AI-scribe tools into easy-to-use digital infrastructure a custom engineering project for each hospital… An AI scribe can… allow a doctor to see two to three more patients an hour, a tangible capacity gain for high-burden Indian hospitals... After adopting Augnito, a voice-to-text tool from the British firm Scribetech, 35% of Apollo’s doctors saw more patients in 2024… Apollo has deployed Augnito across 37 of its facilities since 2022–23, giving nearly 4,000 doctors access to it… 

The tools would need to be highly precise, though, and customised for the Indian context. Transcription errors can impact drug dosage, change patient outcomes, affect insurance claims, and even invite malpractice lawsuits… A medical journal estimated in 2024 that there had been a 400% increase in medical-negligence cases in the previous few years… At a price point of Rs 600–1,500 per doctor per month, AI scribes need wide adoption to break even. Building AI tools is expensive, as model training and GPU costs are high…

Beyond the big players, however, it will take much more to convince doctors to adopt these tools than just a promise of less clerical work... such tools are hardly affordable for a non-chain clinic… Selling to big chains is hard for a new company, however. “Apollo’s actual deal at a corporate level is with Microsoft,” a hospital industry expert says, requesting not to be named. “They have also bundled in another voice solution, Nuance Dragon, to improve documentation.”… So startups like Dawnbreak and Eka Care started with selling their tools to hospitals that didn’t have any EHR at all… Instead of integrating their tools into existing systems, ambient AI firms are looking to provide lightweight tools that hospitals can use piecemeal…

India’s hospital-information-system landscape is fragmented… there are some 2,000 EHR systems compliant with the Ayushman Bharat Digital Mission… Unlike in the US, where Epic and competitor Cerner command 70% of the market, each EHR system in India is different. For makers of AI scribes like Dawnbreak and Eka Scribe, this means building custom-integration solutions for each client rather than a mass product. Dawnbreak, in one year of its existence, has managed to build compatibility with four EHRs out of nearly 2,000… Indian EHR companies like Healthplix and Docpulse safeguard their databases. If they open their APIs up, they lose their competitive edge. Their clients are locked in long-term contracts, leaving them unable to change their systems or integrate any AI tools. 

This is a good case study on the problems with scaling startup innovation in India. 

1. AI has undoubted potential for significant productivity improvements, including in public systems. Like scribing, triaging of outpatient (OP) cases coming to a primary health centre (PHC), community health centre (CHC), district hospitals, and medical colleges is an area where AI can play a significant productivity enhancing role. In all these places, OP cases come to doctors with limited or no triaging. Further, as we have seen, the daily OP load in these hospitals (at least the better ones among them) is multiples of what a doctor can manage, leaving them overburdened and stressed. The result is inefficient use of the doctor’s time, inadequate diagnosis time, incorrect diagnosis, wrong OP referrals, and so on. 

An AI-based triaging application where the symptoms are entered at the OP-registration, nurse and doctor-level, can dramatically improve work conditions, increase hospital productivity, and enhance the quality of treatment. Triaging is already one of the early emerging successes of AI, with examples like Bank of America’s digital assistant “Erica”, which handles billions of client interactions and has reduced call centre volumes by 40 per cent. 

2. However, the promise of AI is most likely to be constrained in sectors like healthcare and others where health and public safety are critical factors. In these regulated areas, vertical use cases of AI adoption (agentic solutions that can be outsourced specific tasks) is likely to be slower. The regulatory struggles of autonomous driving systems is an illustration. 

Even a clear and credible demonstration that AI is more accurate than the current human-intermediated approach will not be sufficient. Notwithstanding all its flaws, the human psychology and political economy is such that society will demand a very high, near 100%, accuracy from any electronic/digital system that seeks to replace a human-intermediated system. 

3. There are some important market insights here. Econ 101 would have it that since health care has inelastic demand, and also given the sustained high economic growth rates, one would have imagined a large supply side of hospitals in India who deploy such solutions. Similarly, one would have imagined that Indian startups would have grabbed the opportunity presented by developing AI solutions on patient triaging, consultation scribing, diagnostics, EHR, etc. 

I’m not sure about whether the Indian market can support the demand for such apps and services at the price points required to sustain domestic innovation. Sample this on the limited consumption potential of the Indian economy, and the challenge of making money in the country.

While India’s population of 1.4bn offers enviable scale, its market has proven difficult to monetise. According to Sensor Tower, Indian internet users downloaded 24.3bn apps in 2024 and spent 1.13tn hours on them, but total spending was just $1bn.

The advantage domestic startups have is their lower price point. But any scaling pathway for startup innovation that relies on price point may be no scaling pathway at all. A business model that relies on a low price point does not generate the cash surpluses required to finance the significant R&D investments required to refine such products. The net result is that genuinely innovative companies remain elusive. 

I blogged here about the demand-side constraint arising from the deeply price-sensitive nature of consumers and the small size of the consumption class with disposable incomes. It also does not help that Indian firms, including startups, do not have a culture of investing in R&D beyond that required to grow their ongoing businesses. 

4. The dominant narrative on startups, shaped by the Silicon Valley giants, is that of scaling by growing exponentially. But contrary to this, apart from killer apps and the few platforms, the main scaling pathway for ambient scribing startups like Eka Care or Dawnbreak may well be through large IT companies already serving the same or similar market segments. The vast majority of these solutions, and not just in health, are limited in their scope as stand alone applications. But this changes dramatically once they are integrated with a larger ecosystem platform to leverage network effects. 

These startups will struggle to get the big users, large hospital chains like Apollo and Max, to replace their bespoke legacy solutions or those supplied by established foreign vendors. This is a daunting market access challenge that even the startups with great solutions will face in markets like India. 

It raises the important point about a model of the digital economy where startups develop innovations which in turns scales through large firms. This not only makes the large firms even larger, but also maximises value capture by them.

It also raises the question of whether the startups should pursue getting the big hospital chains to become their investors. This will also align the incentives of the hospitals to integrate these solutions with their EHR and legacy systems. 

5. It is here that the failure of India’s software behemoths to build on their first-mover and other competitive advantages assumes significance. Both TCS and Infosys have long experience in global hospital systems management, including multi-year, multi-billion-dollar contracts. Hospital tasks management applications should have been a natural area of business development for these IT majors. But Indian software firms have struggled to break out from their services-led business model and embrace products and solutions which require high R&D investments. 

IT services still dominate with exports set to reach $210bn this financial year, India Ratings and Research forecasts. It has been a powerhouse industry for India but as IT services presented so much low lying fruit, the sector sucked up tech talent and capital from elsewhere. India’s SaaS sector in particular punched below its potential as a result. Software majors treated their services businesses as cash cows, deploying a small share to intellectual property assets. The 10 largest IT services companies had consolidated profits of $114bn in the past decade; 75 per cent of this was paid out via dividends and buybacks.

While the top five Indian IT firms had free cash flows of nearly $13bn in the 2023-24, their R&D investment was a pitiful 0.88 per cent of sales. 

6. Finally, what can public policy do to solve some of the scaling challenges? An India Stack for digital payments is unlikely to work for the far more complex area of EHR. Public policy cannot solve market coordination problems (like sharing APIs to allow integration and inter-operability), except in some contexts by defining standards. Public sector driven demand-side channels like Ayushman Bharat can gently force some standards. 

The approach of supporting scaling by procuring for use in public systems runs into the problems of punishing public systems with second quality or inferior products and creating perverse incentives among the startups. Providing a small sample of public hospitals does not address the market scaling challenge arising from network effects, besides also creating procurement problems even if the solution is found effective.

Courts as co-designers of public policy in India

The Supreme Court of India has delivered two highly consequential judgments in the first two months of the year. 

This is in the long list of judgments in the last decade-and-half, some of which have clarified and stabilised the law, and others have introduced deep uncertainties. These judgments have made courts virtually co-designers of policies on critical aspects of the economy, like resource management and taxation, and co-regulators of important sectors. 

In the first judgment, on January 15, 2026, the Supreme Court ruled that the US private equity firm Tiger Global must pay tax in India on its 2018 sale of its 17% stake in e-commerce giant Flipkart to Walmart for $1.6 billion. It overturns a 2024 Delhi High Court decision that allowed Tiger Global to claim tax relief under the old India-Mauritius double-taxation avoidance treaty. The High Court had agreed with Tiger Global’s claim that its gains were shielded from Indian tax because the investment was held through entities that had tax residency status in Mauritius. The government had changed the Indo-Mauritius double-taxation treaty in 2016 through the General Anti-Avoidance Rules (GAAR) which made gains from the sale of Indian shares taxable even under treaties if they were “impermissible avoidance arrangements”. However, it exempted investments made before April 1, 2017. Tiger Global’s investments predate the change. 

Indian tax authorities rejected the claim and argued that the Mauritian firms served as conduits and were used only to avoid taxes, with no real business purpose. The Supreme Court… ruling that tax certificates alone do not guarantee treaty benefits and that the investment structure lacked real commercial substance. It held that foreign investors cannot rely on complex offshore set-ups when those entities don’t carry out genuine business activities of their own. JB Pardiwala, one of the two judges, wrote: “Taxing an income arising out of its own country is an inherent sovereign right. Any dilution of this is a threat to a nation’s long-term interest.”… 

India and Mauritius signed a protocol in 2024 amending their tax treaty to benefit only companies with legetimate businesses and not shell companies set up to avoid tax… India had long tried to attract foreign capital by encouraging investments from companies with structures in countries such as Mauritius, Singapore and the Netherlands, signing treaties to help investors avoid paying taxes twice… Between 2000 and March 2025, Mauritius alone accounted for about $180bn (£133.9bn), nearly a quarter of all foreign direct investment into India, according to official figures.

This effectively means that GAAR’s look-through of treaty structures overrides any treaty claims in the cases of transactions lacking any commercial substance or made solely to avoid taxes. In this backdrop, how does India compare with other jurisdictions in the taxation of gains from share sales?

After the Vodafone case, the government had, in 2012, retrospectively legislated for taxation of offshore share transfers in a foreign company where the underlying shares derive “substantial value” from India. While this is the legal foundation underlying the Tiger Global ruling, it overrides the grandfathering provision in the legislation for prior deals. See below the indirect transfer taxation regimes across countries, which show that indirect transfer taxation is confined to real estate in most developed economies. 

In conclusion, the ruling, which could reshape how foreign investors exit their Indian investments, sets out a tougher interpretation of tax treaties. It allows authorities to deny treaty benefits if offshore investment structures are deemed sham entities with little commercial substance, even when investors hold valid documentation. The judgement gives India wide powers to scrutinise any offshore corporate deal. It also operationalises the Vodafone legislation. 

In the second judgment, on February 13, 2026, in the State Bank of India Vs Union of India, it ruled that telecom spectrum is a natural resource held in public trust and the right to use it does not form part of the insolvency estate of a telecom service provider (TSP). Given that the spectrum (and associated license) is the bedrock for TSP’s business, it forms the basis of TSP’s bankability. MS Sahoo and Raghav Pandey write,

This ruling effectively places the most valuable asset of a TSP beyond the reach of a resolution plan. The likely consequence is the liquidation of stressed TSPs and the fragmentation of the insolvency framework, contrary to legislative design… the ruling rests on a conceptual overextension. The Public Trust Doctrine (PTD) is applied without sufficient regard to the evolution of the modern regulatory state and market economy… The PTD emerged to protect communal access to resources such as air and water, as a check against the privatisation of the commons… In telecommunications, the state has translated the PTD into a detailed statutory and contractual framework of auctions, licences, and contracts. That framework explicitly permits the allocation, trading, and transfer of spectrum-usage rights. 

When the state auctions spectrum, it does not abandon the public trust; it operationalises the use through market mechanisms. A sovereign resource is converted into a regulated, tradeable economic entitlement, juridically embodied in the licence. For the Insolvency and Bankruptcy Code (IBC), 2016, it is this statutory-contractual construct that matters, not the doctrine in abstraction. The judgment does not fully distinguish between sovereign ownership of spectrum and the contractual licence conferring the right to use it. These operate at distinct juridical levels. Spectrum remains vested in the state at all times, while the licence is a statutorily recognised intangible right, acquired for valuable consideration… In accounting and economic terms, the money paid to acquire the licence exits the balance sheet and is replaced by an intangible asset of corresponding value… Banks and financial institutions lend money to TSPs secured against these licences; that security ought not to be diluted by invoking the PTD.

This decision makes India an outlier in the treatment of spectrum in insolvency proceedings. The US, UK, EU, Japan, Brazil, and Mexico treat telecom like other sectors in bankruptcy proceedings. 

Not only does this ruling make India an outlier in telecom spectrum treatment in insolvency proceedings, but it also makes telecom an outlier among other sectors, even in India.

While mining leases, airport, port, and road concessions, and electricity PPAs can be transferred with regulatory approval, the same is now prohibited for telecom spectrum. This is despite telecom having similar features - time-bound lease, competitive allocation, revenue sharing, and regulated transfer - as the others. 

While all countries recognise public ownership of resources, with this ruling, India now diverges from others in the legal test to decide whether a government-granted license or concession is part of the insolvency estate and therefore transferable or usable in resolution. 

The ruling effectively reduces regulated asset values, raises the cost of telecom finance, weakens restructuring scope, and makes telecom concessions riskier than other infrastructure concessions, all this due to an avoidable regulatory interpretation. 

ChatGPT has this compilation of the Court decisions in India that have reshaped economic regimes in their respective sectors. The estimates are unverified and can be significantly off. 

Of these, the cancellations of the coal blocks and 2G licenses, and the spectrum insolvency are major deviations from global norms. Also, on a global comparison, India has a relatively large number of Supreme Court decisions that rewrote regulatory frameworks, applied law retrospectively, and changed business models. 

In countries like the US, UK and EU, courts rarely cancel licenses or concessions, retrospective orders are uncommon, and regulatory regimes are usually shaped by legislatures and agencies. Courts in these countries leave policy design and regulation largely to administrators, legislatures, and independent regulators, and prefer to only interpret statutory frameworks. 

In contrast, Indian courts have taken an absolute view on public trust doctrine (without regard for its economic dimensions), favoured substance-over-form in taxation, and assumed the powers of broad and unconstrained scope in the judicial review of allocation processes. India remains distinctive in the scale and frequency of judicially driven economic restructuring, with courts almost acting as co-designers of policies on public resource management and taxation.

The concern here is not so much with the nature of the decisions per se. After all, some of these decisions are clearly progressive - substance-over-form in taxation, stricter ever-greening standards on patents, homebuyers as financial creditors - and should be adopted more widely globally.

Instead, there are two important concerns. One, at a fundamental level, should courts be making such definitive policy decisions? The argument that courts have stepped in where governments have abdicated cannot be taken as an answer. This is a slippery slope that risks destabilising the constitutional checks and balances. 

The second concern is about the economic uncertainty induced by rulings that upend economic regimes (or rules of the game) based on which business decisions were taken. If investment decisions were taken based on a prevailing interpretation of the regime and the same was widely accepted (governments gave permissions, financial institutions gave loans, rating agencies did not consider them risks, auditors audited statements, tax authorities generally overlooked them, etc.), then a subsequent substantive (logical, ethical, etc.) interpretation should not form the basis for reversing it with retrospective effect. The only exception to this should be if it is established that there was a malafide intent in the decision made by the government entity. 

The economic damages are exacerbated by the consequential decisions forced upon government entities. For example, the cancellation of one license immediately leads to the cancellation of all similarly placed ones. The rejection of a tax avoidance claim immediately triggers tax officials across the country to scout for similarly placed cases and issue notices to them. 

I’m not sure whether the legislative or executive can do anything to address these concerns. Its answer lies in the judicial realm itself, in the form of restraint while courts take such decisions. The Supreme Court could use the opportunity presented by something like a Public Interest Litigation (PIL) or a case to lay down certain judicial principles that should guide judicial rulings on cases with sectoral policy impacts.