TOD is critical for the commercial viability of Indian metros

I have blogged here making the case for doing Transit-Oriented Development (TOD) around mass transit stations in India, here about the challenges with TOD implementation, and here about the low traffic realisation problems facing metro railway systems in India and the ₹1.07 lakh crore (undiscounted) fiscal burden. 

This post will argue that TOD is the critical requirement for the commercial viability of India’s regional and metro railway systems. TOD is a combination of densified, mixed-use development (supply-side), and walkable streets andintegrated feeder services (demand-side) around mass transit stations. 

In other words, regional and metro railway systems in India can become commercially viable only if they can induce ridership, and this shift would require adoption of TOD at both ends of the commute. 

But India’s regional and metro railway systems are encumbered by structural problems compounded by bad design.

For a start, unlike the successful global mass transit systems which precede real estate development and emerge from an integrated land use and transit plan, Indian railway systems follow it. They try to retrofit into the built-up city and suburbs. Second, as I wrote here, Indian regional and metro rail systems unfairly put the burden of capex on the operating entities without endowing them with the real estate that can be monetised to finance capex. Third, the railway construction entity does not control land use decisions, and those making them do not bear the consequences if their decisions erode the financial viability of the mass transit systems. Finally, making matters worse, they are primarily railway construction corporations, for whom urban planning integration and real estate development are afterthoughts. This is unlike in Japanand other East Asian cities where the metro operators are primarily real estate developers who also build and operate railway networks. 

Given the circumstances of India’s mass transit journey, the best that can now be done is to create conditions to induce ridership and improve commercial viability of metro and regional railway systems. And, as aforesaid, TOD is central to this objective. 

The first order requirement is to induce ridership. This can come from two directions. At the intensive margin, ridership can be induced by shifting existing car users to mass transit. At the extensive margin, it would involve locating commuters within walkable distance of the station and expanding geographical access through feeder services, and shortening the transit trip. 

The intensive margin relies on modal shift among people whose destination and origin are fixed, whereas the extensive margin creates new origin-destination pairs of which mass transit is the obvious mode from the outset. An important, but less discussed, mechanism of the extensive margin is that of trip shortening. The regional or metro rail shortens the commute time to distant places (besides also eliminating the hassle of car-based trips) and brings workplaces closer (in time) to homes and thereby expands its potential users. This trip-shortening effect is typically larger than the modal-shift effect, and second only to ridership induced by locating people inside the TOD zone.

However, while convenience (of avoiding road traffic and walkability at both ends) can induce some mode shifts, getting people who would otherwise not prefer living near a station at all to relocate entails going beyond convenience and also making the place itself liveable and financially attractive. Further, the inducement is significantly likely only if there is a workplace within walkable distance at the other end of the commute.

In the case of metros, it is about densified developments (at both home and work sides) around stations with walkable streets, supplemented with feeder services. In the case of regional railway systems, this is the classic downtown-satellite town mode of development. However, in India, since the satellite towns are already well-developed towns and cities, the station vicinity requirements are similar to the metros.

The walkability part is critical, and one completely overlooked. Densification alone without the ease of walkability at both ends will not induce ridership. Walkability and street connectivity are central to expanding the pool of users. If a commuter who lives 800 metres from a regional railway station still needs to find an auto or shared vehicle because the intervening streets are not well connected, unwalkable, and unsafe, the catchment effective radius collapses to perhaps 200 metres, and the ridership projection drops accordingly.

The demand side (walkability) and the supply side (density) are complements, not substitutes. Building one without the other produces either pleasant low-rise neighbourhoods that don’t fill the trains, or towers in traffic that people can’t walk through. Further, the walkability (and associated liveability), in addition to the mass transit commute convenience, are perhaps critical for also incentivising people to trade-off living in a cheaper and larger housing unit distant from the zone with a pricier and smaller one inside the TOD zone. 

Finally, the mixed-use part is also important. For one, it should provide market participants the flexibility to determine the demand responsive and commercially viable nature of development around the nodes. Further, a node that densifies the area around a station but retains single-use zoning (apartments that connect to the city’s existing car-commute grid) will produce less additional ridership than a node that places offices, retail, and housing within walking distance of each other and of the station. The ridership gain from the second type of node comes not only from the peak commute trip but from the off-peak trips that only a genuinely mixed-use, walkable neighbourhood generates. The former will end up creating a mass transit system that is used only in the 3-4 hours of peak commute times. Mixed-use is therefore critical to generating two-way, all-day ridership.

In other words, ridership of a transit system is determined predominantly by the land use in its catchment — how many people live, work, shop, and study within a walkable distance of its stations, and whether those people can get to and from the station without a car. Building the station but leaving the land use unchanged will not change much. It will produce the infrastructure, but underproduce the demand. This has been India’s metro reality. 

All this means that densified, mixed-use, walkable zones around the station, with convenient feeder services, or TOD, is an essential requirement to achieve significant inducement of ridership. 

The case for TOD is likely more relevant and its impacts greater for Indian cities than counterparts in developed countries. For one, the spatial mismatch between job availability (in the city centres) and affordable housing (in the periphery) is very large. A system that closes this gap generates much greater welfare gain, and therefore much greater demand, than the same system operating in a less mismatched city. 

Further, given India’s low car ownership base, the marginal ridership gain from TOD does not come mainly from convincing existing car users to take the train, but from improving trip quality for those already transit-dependent, and from expanding the pool of riders. Finally, India has a low-hanging fruit to harvest in terms of its low FAR baseline, which allows for significant densification and location of potential transit riders. Relaxing FAR therefore would have an unusually high ridership yield per unit of policy change in Indian cities.

The summary from the Indian context is that the inducement from modal shift is likely marginal, and confined to only those whose commute is already partly walkable and partly by car, or whose car commutes are very long. The larger and more structurally significant opportunity is at the extensive margin from locating new workers and from bringing the workplace closer to homes. TOD is therefore the primary demand generation mechanism for a rail transit system. 

Unfortunately, India’s TOD policies have consistently focused on the supply side (FAR uplift), and that too inadequately, while under-specifying the demand side (street design, parking, block structure). Getting both right is the institutional challenge. It is also something that the argument for TOD-for-viability of metros implicitly assumes but rarely makes explicit.

Thoughts on private equity in consumer service businesses

It has always been a matter of debate that there may be a trade-off between building an enduring business and maximising short- to medium-term returns on those business investments. The private equity model of investment, it has often been argued, inclines to the latter. 

I have written here about the incentive incompatibility problem when returns- maximising investors like private equity pursue infrastructure assets with their low but stable returns. More generally, when asset ownership and operation and maintenance are separated, and especially when the former is dispersed, the incentive distortions compound. 

With headline returns capped by regulation, the PE investors must find alternative channels to squeeze more out of their investments. Leverage and asset stripping become the preferred strategies. The British water privatisation is only the most totemic illustration of these trends. 

In this context, I have also blogged here and here, and here about the trend of rising PE investments in the locally operated regular consumer services — healthcare, education, residences, student housing, pet care, salons, mobile trailers, pubs, home repair services, etc. In all these cases, incentive distortions similar to infrastructure have now been widely documented. 

The strategy in all these cases is to target service-based businesses with long-term demand (especially those requiring O&M, and the possibility of subscriptions and service contracts), aggregate or roll up small local providers, and centralise their operations to harvest economies and efficiencies of scale. On top is the purchasing (and bargaining) power with suppliers and customers. And on the bottom-line side, undertake aggressive cost-cutting and asset stripping. 

On the financial side, the strategy is to harvest the scale arbitrage — acquire a large enough regional or sub-regional entity at 6–8x EBITDA, add 10-50 more smaller entities at 5–7x, centralise back-office and procurement, load up debt, and exit the consolidated entity at 10–12x before the consequences of cost optimisation show up on service quality. The primary driver of returns is less operational improvements, but the arbitrage of paying small-company multiples and selling at large-company multiples. This is financial engineering applied to a captive demand. The new buyer, in turn, would be motivated by another business model for the same business. 

There is also the tension between building long-term sustainable business and the pursuit of short-term returns maximisation. In other words, long-horizon value creation trades off against short-horizon, leveraged returns maximisation. It is like the comparison between a business that earns a steady 8% over 30-40 years, and another that targets 15% over 7-8 years in addition to the 2:20 fees. 

In short, the PE model’s problem appears to be that it seeks to maximise returns within a hold period that is too short to bear the consequences of its own cost optimisation. This causes a temporal mismatch — returns are harvested before the deferred maintenance, staff morale declines and attrition, community erosion, and quality degradation. 

Apart from the adverse impact on the quality of service delivery, there are also the negative externalities ranging from a change in the ownership of local businesses (with disruptive impacts on local communities) and increased prices of services, to bankruptcies and closures with attendant job losses. So much so that private equity has become a lightning rod for rising housing prices in the US mid-term elections. 

FT has an article that points to PE firms buying up swimming pools in the UK and offering maintenance services. 

Private equity-backed swimming pool “platforms” have been buying up hundreds of local builders and repair services. The goal is to create national scale in an industry dominated by pint-sized companies. SPS Poolcare and Pool Troopers, two of the largest consolidators, between them bought more than 200 businesses before merging themselves earlier this year. Main Street Capital, a small listed private equity firm, has marked up the value of its equity holding in Cody Pools sevenfold since buying in at the height of the coronavirus pandemic.

The trend might sound surprising to anyone focused on public markets, where many investors ended up deep underwater after a wave of enthusiasm crested in 2021… while buying a pool is about as discretionary as it gets, upkeep on an existing pool is not. There are north of 10mn installed in the US, and they all need regular cleaning and repair. Servicing businesses can therefore offer the sort of subscription-like revenue that buyout firms love. And although construction is in a cyclical slump, there are reasons to expect longer-term growth as climate change makes outdoor pools attractive in more areas.

This mirrors PE investments elsewhere in retail. The FT article writes

Compare the heating, ventilation and air conditioning sector, where there have been more than 1,100 acquisitions since 2020, according to data from Capstone Partners. Mid-market specialists have had success rolling up small companies until they create a group big enough to appeal to bigger buyout specialists, as shown by Blackstone’s $2.5bn purchase of Champions Group earlier this year. Pools are just one riff on a wider home-enhancement theme. Other firms are trying the same playbook with trades like garage doors, pest control and plumbing.

The trajectory outlined above is largely a narrative formed by observations. What does the evidence say?

I have written extensively on infrastructure, consolidated here. See also this recently. The sector with the largest and most conclusive empirical evidence base, and at least less than positive impact of PE is healthcare. In a systematic review in The BMJ, covering 55 studies from 2000-23, Alexander Borsa and others found definitive evidence. Of the 27 studies assessing quality, 21 found harmful impacts; of the 12 studying costs to patients or payers, 9 found increased costs, and none found decreased costs; and of the 8 studying health outcomes, 3 found harmful, 2 beneficial, and 3 neutral impacts. 

This mirrors findings across the segments in health care. Atul Gupta et al, found that PE ownership of nursing homes raised short-term mortality by about 10% or about 20,000 additional deaths over the sample (1,674 PE-owned nursing homes from 2000-17, and covering 4.2 million patients), increased spending by 11%, reduced nurse staffing, and decreased compliance with care standards. 

On dental practices, Kamyar Nasseh et al, found that PE-acquired dental practices raised list prices by 3.3%, and shifted procedure mix from preventive toward higher-reimbursement restorative, specialty, and surgical procedures. Sailesh Konda and Joseph Francis showed that PE-owned dermatology practices saw 4.7–17% more patients per dermatologist, raised prices for routine visits by 3–5%, and employed four advanced practitioners per ten dermatologists (vs three in non-PE practices). 

In the most dramatic evidence of revenue-maximising treatment intensity replacing clinical judgment, PE-managed neonatology practices were associated with 70% higher common NICU days and 54% higher physician spending. Finally, this study by MIT Sloan found that negotiated prices between hospitals and insurers rose 32% after PE investment. It also found that PE firms burden acquired health care organizations with unmanageable debt, stealthily decrease health care competition, often increase costs for patients and payers, can compromise patient care, and can harm health care workers and providers. 

Its findings and recommendations are striking:

Private equity firms’ focus on short-term revenue generation and investor profit also can lead them to strip acquired facilities of their assets, force those entities to raise prices through anticompetitive practices, reduce staffing to dangerously low levels, avoid investment in critical infrastructure, and eliminate vital services—to the detriment of patients, workers, and entire communities… Policymakers must take steps to safeguard the health care system against harmful private equity practices by enhancing regulatory oversight over health care acquisitions, rolling back reporting exemptions on financial transactions in private markets, and changing the incentive structures to limit private equity firms’ interest in engaging in financially risky behaviors that run counter to the public interest.

The empirical evidence in other consumer services is thinner, perhaps only because of lack of studies. There is no peer-reviewed equivalent of the Borsa review for HVAC, pool services, pest control, pubs, pet grooming or salons. This will change in the years ahead as PE ownership in these businesses surge. 

However, in all these cases, there are several anecdotal and journalistic examples of high-profile failures. They are most likely a good representative sample of the general direction of impacts from PE in those market segments. But in the absence of empirical evidence, supporters of PE will continue to argue in their favour. 

This cannot detract from the emphatic finding that the distortionary impacts of PE ownership of businesses are documented wherever the research has been done. 

As with all such debates, the reality is perhaps more nuanced. The evidence from competitive, non-essential service markets like manufacturing, technology, and business services is mixed. For example, Steven J Davis, John Haltiwanger, and others have found that PE-owned manufacturing plants had higher productivity growth. 

I can think of some determinants of a business that increase the likelihood of incentive distortions. One, a service where quality is hard for the buyer to observe at the point of purchase creates perverse incentives. Two, the nature of these kinds of consumer services calls for stewardship and trust, which are most often overwhelmed by market incentives. Three, captive or sticky demand arising from switching costs, essential need, or geographic lock-in, is another factor that encourages market abuse. Four, there is a moral hazard arising from a regulatory backstop or a third-party payer that socialises costs. Finally, a short holding period allows the owner to exit before the consequences of underinvestment show up. 

Healthcare and infrastructure are vulnerable to all these determinants, and others in varying combinations and degrees. The evidence is strongest in healthcare, and the same structural incentives apply wherever the determinants are met. However, the absence of evidence elsewhere reflects the absence of research, not the absence of harm.

All this comes in addition to questions on the financial side too, about the superiority of PE as an investment strategy. I have blogged earlier here and here about Ludovic Phalippou’s extensive research on PE returns (here and here). Across three large datasets for the period 2006-20, PE funds delivered net Multiple of Money (net-of-fee) of ~1.55–1.63x, or about 11% annually, matching public equity indices in the same period. During that time, roughly $230bn in carry accrued to a small number of managers, with the number of PE multibillionaires rising from 3 in 2005 to 22 by 2020. Much of the apparent outperformance, he shows, came from choice of benchmark rather than from the returns themselves.

Weekend reading links

1. China's effect on solar power in a graphic.

At the turn of the millennium, solar panels cost $5-$6 per watt of generation capacity. Today, they sell for around 12 cents per watt, a level described as “offensively cheap” by Dave Jones, co-founder of think-tank Ember. That reduction is entirely the result of an explosion in China’s production capacity, which, according to research firm Wood Mackenzie, stands at roughly 1.36 terawatts, even as Beijing tries to rein in output to combat price deflation.

And rooftop solar generation is causing a loss of good customers across utilities, triggering a 'utility death spiral'. 

Shoprite, Africa’s biggest supermarket group, says the company could “power an entire suburb” with the panels it has been installing on its roofs and trucks in South Africa and Namibia since 2015. But what has been good for Shoprite, which now has around 43MW of peak capacity at its disposal and is looking at battery storage, constitutes a problem for state power utility Eskom. It estimates that rooftop solar panels and batteries, mostly installed by businesses and wealthier customers in response to years of rolling blackouts, were responsible for around 7 per cent of the 11.7 terawatt-hour reduction in its electricity sales for the year ending March, compounding a sharp fall in industrial use exacerbated by high prices. 

The lower revenues come as Eskom needs to fund the upkeep of the grid, which allows households to export electricity during the day, and the power stations that fill the generation gap when it’s dark or cloudy, or when power stored in batteries is insufficient to meet demand. “Everyone who has rooftop solar in any major city, they are still connected to the grid,” says its chief executive, Dan Marokane. “Three weeks ago . . . the whole country had to rely on Eskom generation for three days,” he adds, due to prolonged cloudy weather slashing output from rooftop panels... In Australia, operators of the high-voltage transmission system have grappled with too little demand as consumers draw power from their rooftop panels rather than the grid. Very low demand can make it harder to run certain power plants needed to keep the whole system stable.

2. AI is transforming warfare, and the Russia-Ukraine war is the testing ground for these technologies. 

AI technology has changed how the war in Ukraine is fought. At the start of the conflict it could take 20 minutes to identify a target and launch a strike. Now that process can take less than two minutes, sometimes just seconds... When militaries talk about AI, they are not talking about a single system. They are referring to software being used at different stages of warfare — collecting and processing data, helping commanders identify and prioritise targets and allowing drones to keep operating when communications are jammed. AI-enabled software can process “mass data at a scale, a speed that human staff officers can’t do”, says Anthony King, professor of war studies and director of the University of Exeter’s Strategy and Security Institute. Rather than altering the sharp end of a conflict, King says the primary function of AI has been to improve “situational awareness and intelligence”. The technology can help commanders make battlefield predictions, from the sustainability of a campaign to the rate at which munitions are being depleted.

3. What are the fiscal costs of immigrants?

The UK's comparatively low rates of tax and social insurance contributions for low-paid workers plus a relatively flat state pension mean that people (immigrant or otherwise) who do small amounts of paid work or remain on low incomes contribute little to the exchequer but still benefit from robust state support. In Germany or France’s fiscal systems, someone with the same weak employment and earnings patterns would generate much larger receipts from tax and social insurance due to the flatter tax regime, and would receive a comparatively smaller pension since these are linked more tightly to lifetime earnings.

The result is that in order to be a net fiscal contributor over their lifetime, the average couple arriving in the UK at age 30 needs the primary earner to have a salary at the 55th percentile of the overall earnings distribution, compared to the 45th in France and 28th in Germany. This is according to a new working paper on the fiscal impacts of immigration in different European countries by Usama Polani, a researcher at the Stanford Institute for Economic and Policy Research. Put another way, for immigration to be financially beneficial to the state, the UK needs to attract migrants with much higher pay and rates of employment than its peers.

4. The government seeks to restructure the governance of the storied Indian Statistical Institute (ISI). 

The ISI Bill 2026 seeks to repeal the earlier ISI Act 1959, changing the institute’s status from a registered society to a body corporate. This would replace its 1,800-plus-member general body and 33-member board with an 11-member board, with seven members aligned with the government and four representing ISI... The new board will be leaner by a factor of three: the President of India will be the visitor and will appoint the chairperson as well as four experts of her choice, joined by two representatives from Mospi. The four remaining members will come from ISI... The academic council, which discusses general syllabi, the number of students to be admitted, and new courses, has so far comprised all ISI professors, representatives from all constituencies, and the director as chairperson. The new bill changes that. Under it, the academic council is reduced to a small body, mostly made up of government representatives. The dean of studies is not a part of this council in the proposed bill.

5. Oil prices over the last fifty years.

Oil production is also less concentrated in the Middle East. The US is now the world’s largest producer, ensuring it will not have a supply shortage, and Canada, Brazil, Guyana, Venezuela have all increased production during the Iran crisis. More importantly, a barrel of oil matters less to the global economy than it used to. Oil’s share of global energy demand has fallen below 30 per cent from a peak of 46 per cent half a century ago. At the same time, the amount of energy needed to produce a unit of GDP has fallen by more than a third since 1990. Ben May, head of global macroeconomic research at Oxford Economics, says the inflationary impact of the Iran war has been dulled because “we went into the crisis with favourable oil supply” at a time when overall economic demand was “steady rather than spectacular”. When the war began, by contrast with the 1970s energy shock, the world had strategic reserves, an oil glut and growing supplies of liquefied natural gas, and a sophisticated trading system in which thousands of tankers criss-crossed the globe.

6.  NYT has this investigation of how online gambling giant DraftKings is using algorithms to target gamblers who are likely to lose the most.

So in 2023, DraftKings took customer betting records and built a machine learning model, a form of artificial intelligence that seeks patterns in data, to answer the question: Who was more likely to respond to promotions by gambling — and losing — more?... DraftKings makes money when gamblers lose money. And the model sought to identify those it could get to lose the most. It scored each customer based on their habits: The higher the score, the more money a gambler was likely to lose for each promotion offered... DraftKings has continued to hone its methods, using data science, to target losing gamblers with promotions that encourage more betting, according to six former employees who worked on them. 

At the same time, four other former employees said, DraftKings has stalled or squashed efforts to use similar technology to predict who might develop a gambling problem based on their betting activity... documents show how the model... analyzed dozens of data points for each gambler, including how frequently they played, their daily account balances and how much they typically lost compared with how much they bet. It also incorporated another model that calculated how likely a user was to stop gambling. This betting data may also contain signs that a person is headed for trouble. Yet when employees developed a machine learning model that would have assigned users “risk scores,” the company sidelined it, according to two former employees who worked on that project...

Promotions, which take on forms like a free bet, a “profit boost” or a deposit bonus, play a vital role in DraftKings’s business: The company brought in around $8.7 billion in gross revenue from sports and casino gamblers last year, and gave out about $3 billion in promotions, according to research by Citizens Bank... they were effective because they take advantage of gamblers’ psychology. “I feel I’m getting free money,” he said, “but really it’s dragging me back in.” Several gamblers told The Times that promotions fueled their addictions....

In 2018, when the Supreme Court ruled that states could legalize online sports gambling, it ushered in a new era in which betting has moved beyond casinos and racetracks. Professional sports games are now saturated with celebrity advertisements, encouraging people to wager on their phones. DraftKings and its rival FanDuel dominate this new industry, which has expanded into online casino games and, more recently, prediction markets. DraftKings says it has 11 million customers, compared with five million in 2022.... Silicon Valley firms spent years analyzing every digital interaction to predict what will keep users clicking on advertisements. Now, as companies like DraftKings have made gambling accessible to millions on smartphones, they too have collected an extraordinary wealth of data.

7. The imposition of the 0.4 per cent MDR fee on UPI transactions of more than Rs 2000 has generated intense debate. Janak Raj has a very good article. 

8. Some facts about the SpaceX business model assumptions.

Analysts have the company delivering revenues of over $650 billion, with an operating profit of over $335 billion by 2031. The company itself is projecting revenues of $1 trillion by 2030 (that is 25 per cent of India’s current gross domestic product). The models have the company generating no free cash flow through 2031, with capex of over $1.76 trillion from 2026 to 2031. Total annual capital expenditure for listed Nifty500 companies in India is about $100 billion. Such is the scale and ambition of SpaceX.

9. The cost differential between imported and domestically manufactured solar cells is significant, and assumes importance in light of the restrictions on the use of imported cells for grid-scale solar plants.  

The ALMM List-II mandate requiring domestically manufactured cells for utility-scale projects from 1 June 2026 tightened cell supply and lifted prices... The shift to domestic sourcing can sharply increase costs for projects that were originally bid on the assumption that imported cells would be used... Sudharman Ezhil, director and CEO of Natrinai Ventures (NGE Green Energy), said the cost difference between a domestic-cell plant and a non-DCR plant is currently at least ₹70 lakh to ₹1.2 crore per MW. "On a 50 MW project, that is ₹35–60 crore that was not in anyone's bid model," Ezhil said. According to him, projects bid before mid-2025 assumed imported cells at ₹14–15 per watt. The same module using a domestic cell now costs ₹24–25 per watt, he said. "That alone moves the total project cost by 15–20%," Ezhil said. He said the issue is not limited to cost. "Cell manufacturing is far more complex than module assembly, process control, wafer quality, yield, and India's listed cell capacity is a fraction of its module capacity," he said. Developers are therefore dealing with both higher procurement costs and uncertainty over the long-term performance of newly listed domestic cell lines. Projects awarded at tariffs of around ₹2.50–2.60 per unit were bid when costs were lower. Once a power purchase agreement is signed, developers have limited ability to pass on higher project costs through the agreed tariff... The domestic content requirement (DCR) premium alone could increase utility-scale tariffs by ₹0.25–0.40 per unit, even before the cost of storage is included.

Are we saying another round of defaults, restructurings, and consolidation in the solar industry? 

10. China ramps up gold imports in efforts to diversify its reserves.

China has spent a record sum importing more than 1,000 tonnes of gold this year as the central bank and local investors pour cash into bullion amid rising geopolitical tensions abroad and poor returns on local assets. The world’s second-largest economy spent $158.8bn on gold in the first eight months of the year. That compared with spending of $96.5bn for all of 2025 on 886 tonnes of gold... Chinese investors are increasing gold purchases as part of broader efforts to diversify their assets. Chinese holdings of US Treasuries fell to $618bn in July — the lowest level since August 2008...  

Domestic investment options in China are more limited since the country’s property market began collapsing in 2021. The benchmark CSI 300 index is down 1.8 per cent for the year and is still more than a fifth below its peak in early 2021. Meanwhile, yields on Chinese government bonds are close to record lows.

11. The yen carry trade, which has been an important driver of cross-border capital flows and a major buyer of US Treasuries, apart from contributing to keeping down the value of the yen, may be reversing after nearly three decades. 

The yen carry trade is the term for when hedge funds and others use Japan’s currency to access low-cost financing to make bets in markets across the world. For almost 30 years, investors have borrowed the cheap, stable yen in order to fund higher-yielding investments elsewhere, exploiting differences in interest rates and, in particular, the fact that until this month the central bank benchmark rate had not risen above 1 per cent since 1995. But every so often, those differences threaten to shrink, or the value of the yen shifts with unexpected speed. Investors exit carry trades and dump the acquired assets — fuelling spectacular spasms in global markets from emerging economy debt and Nasdaq stocks to cryptocurrencies and luxury property. For this reason, the health of the global economy is deeply connected to the state of the yen, making the carry trade a proxy — albeit an opaque one — for risk...

Even though its true size is extremely hard to gauge, the current value of the carry trade may far exceed $2tn, making it probably the biggest it has ever been, the world’s regulators heard from the experts in Tokyo. The concern that the cheap yen may be anchoring US Treasuries and may have helped inflate a bubble in AI-related shares, they added, leads the list of worries. Many strategists now say that one of the top risks for the year is the danger that the carry trade unwinds... In previous reckonings of the carry trade, the overseas investments of Japanese companies were rarely considered... According to data from Citi, the total outstanding Japanese stock of foreign direct investment, incorporating equity capital, reinvested earnings and debt capital, reached ¥384tn in 2025, meaning it has increased from 20 per cent of GDP in 2014 to more than half today. Citi estimates that non-financial corporations in Japan now hold more overseas assets than banks, pension funds and insurance companies...

For years, the speculative carry trade and the enormous underlying outflow of corporate Japan’s investment created downward pressure on the yen. Now, analysts are wrestling with the prospect of that being reversed - even if many believe that Japanese households and companies are too conservative to move with speed. The mere fear of such an outcome could start a carry trade unwind in motion as investors rethink how safe their current positions really are... A sharp carry-trade reversal, or a massive but gradual repatriation of Japanese capital taking advantage of decades-high yields on domestic bonds, could prove extremely painful to Bessent and other finance ministers across the world. During its era of rock-bottom yields at home, Japan has been a reliable source of demand to absorb record levels of rich-world sovereign borrowing.

12. Anthropic's spectacular pre-IPO valuation run up.

Founded in 2021, the creator of Claude hadn’t even produced a dollar in revenue until 2023. By August, it was making around $65bn on an annualised basis, although such unofficial numbers should be treated with kid gloves. This growth seems to have come as a surprise to Amodei too. Only 18 months ago, Anthropic expected its revenue in 2027 to be just $12bn. Now, some investors predict a revenue run-rate of $320bn by the end of next year. They are not impartial, of course. But if they are right, then $2tn would represent a valuation of just seven times its 2028 sales. That’s a little less than Microsoft, according to LSEG. SpaceX, meanwhile, trades at 16 times that year’s revenue. Take that as the benchmark — they are both companies with wild aspirations and charismatic leaders — and Anthropic could in future be worth $5tn.

13. Big Tech guarantees are surging and risks spinning out of control. Morgan Stanley analysts estimate $3.1 trillion in guarantees issued by the seven hyperscalers and chipmakers.

Big Tech companies are rapidly expanding their use of guarantees to back debt for AI data centres and chips, issuing up to $300bn in commitments in less than a year while recording little of that exposure on their balance sheets... These arrangements, under which tech companies guarantee a minimum future value for chips or data centres, join a growing set of creative financing structures embraced by Big Tech to accelerate the AI infrastructure boom... They typically backstop debt that is issued not by the tech companies themselves but by special-purpose vehicles that own the infrastructure, allowing the tech groups to lend their financial strength to the deals without needing to fully book the liabilities. The rise of guarantees adds a layer of exposure if Big Tech’s multitrillion-dollar bet on AI does not pay off because of disappointing usage, an oversupply of computing power or the failure to build sustainable business models around the technology... Their introduction to AI financing has unlocked cheaper financing for projects with these guarantees, which typically price at just a 100 to 150 basis-point premium to the guarantor’s own debt... Because the guarantor only covers the gap between the sale price and the guaranteed value, residual value guarantees “get more efficient balance-sheet treatment than a typical payment guarantee.”

Public policy for domestic semiconductor design and components in general

India is spending scarce fiscal resources to support domestic semiconductor chip design startups. These startups and their innovations are most likely to remain stranded unless this support is supplemented with a market access strategy. This is the binding constraint to their scale-up. 

I have articulated this in multiple posts on this blog. This provides a policy framework for public funding of innovation; this looks specifically at catalysing the chip design market; this describes how large domestic corporate groups can be roped into catalysing domestic manufacturing; and this provides a template for the most promising opportunity in smart meter manufacturing. However, given its importance, I think the issue merits one more post. 

India now has a stable of chip design companies, including the nearly 25 startups that were funded under the Design Linked Incentive (DLI) scheme. They have designed chips of various kinds, and at least some are at stages of deployment, if not commercial scaling. 

True, they still need to get their tapeouts to be of good enough quality, build internal capabilities to be ready with version updates, and generally ensure that the chip fabrication and supply chain are standardised and de-risked. While the chip design startups must work on getting these right, this work also requires the visibility of a scaling pathway. 

But such deployment and scaling require Original Equipment Manufacturers (OEMs). But unfortunately, India’s domestic OEM landscape in electronics is barren. And foreign OEMs will not risk disrupting their well-entrenched chip supply chains by experimenting with an Indian chip design firm. There are too many uncertainties associated with such supplier shifts. We therefore have a demand-side binding constraint on the scaling of chip design startups.

So the only option to enable scaling is to create the conditions for OEMs to adopt these chips. There are two strategies in this regard. One is to mandate the use of domestically designed chips, at least some share of it, in those products sold in India. The other strategy is to incentivise the same with some industrial policy support.

I have articulated such a policy in detail here. It is a combination of both strategies. It involves mandating a progressively increasing share of the chips used in their products to be domestically designed, and supporting such product manufacturing with a Production-linked incentive (PLI). 

This approach is easier to adopt effectively in products that are B2G or B2B in nature. Smart meters for electricity consumers, set-top boxes and fixed wireless access devices, surveillance cameras, drones purchased by defence agencies, etc., are examples. The relative simplicity of B2G and B2B markets can be leveraged to formulate market-shaping policies. 

This strategy would be useful for the indigenisation of other components and sub-assemblies too, ranging from compressors to transceivers. Mandate and incentivise the use of a progressively increasing share of components from domestic firms. 

This would emulate the Chinese playbook through which they have come to develop a vast ecosystem of domestic component manufacturers and OEMs. The Chinese smartphone OEMs like Huawei, Xiaomi, Oppo, and Vivo were instrumental in the emergence of domestic chip design firms. India will have to develop both the OEMs and component makers. 

This would require a high level of state capabilities in both designing the policy and calibrating and closely monitoring its implementation. But more critically, it would require a very high degree of inter-departmental coordination among the departments and ministries in the Union government. For example, smart meters would involve collective ownership and close coordination across the Ministries of IT, Power, and Finance, and financing agencies like REC and PFC. This would require a genuine breakdown of silos within the government.

Separating infrastructure construction and O&M, and mitigating the risks

This post revisits a contentious strategy in infrastructure finance: separating the contracting of construction and operation phases (as against life-cycle contracting). 

This blog has long held the view that, given the inherent time and cost overrun risks associated with the construction phase, which cannot be borne by private investors, and the much lower cost of public finance, infrastructure assets should be created with public financing. Once the construction risk is shed, the asset should be transferred to private operators through long-term concessions to fund their operation and maintenance (O&M). 

As early as 2013, we had advocated this two-stage strategy in a co-authored oped: 

In the first stage,a professionally managed special purpose vehicle (SPV) could be established to construct the project using short-term bank loans or through takeout financing by a consortium of banks. This may require the government to provide some form of guarantee or credit enhancement. Once the construction risk is offloaded,short-term loans can be swapped for long-tenor debt. Private participation can be introduced either through long-term concession grants to operate the entire project or parts or it,or by outsourcing certain services. Another strategy would be to capitalise the assets by taking the SPV public.

Let the state bear construction risk cheaply, then sell the de-risked operating asset to patient capital. In other words, use public finance to create infrastructure assets, and then fund their lifecycle using private capital. 

But it can be argued that this creates two problems. One, public financing comes with the risk of compromising on the design and quality of construction. Second, this risk is worsened if construction and O&M are separated. 

Clearly, we have a financing dilemma. The most appropriate risk allocation and lower cost of public finance trade off against the problems arising from the separation of construction and O&M. 

How do we reconcile the dilemma?

If there is one constant that infrastructure construction globally teaches us, it is that cost and time overruns happen over and over again, and are therefore unavoidable. Given the uncertainties involved, most often beyond the control of contractors, it is unfair that this risk be borne by them. Only the contracting agency of government can bear this risk. Further, it is unavoidable that financiers will hedge for those risks and price the capital accordingly, saddling the asset with prohibitive life-cycle costs. It is also virtually impossible to write contracts with such risk allocations that absolve the construction contractors. 

In the circumstances, the only option left is to assume the construction delay risks on the public balance sheet and finance the asset from the budget, and then monetise the asset. 

So if we make a conscious choice for public finance, how to mitigate the risks? How do we reconcile the apparently conflicting requirements of using public finance to construct infrastructure assets and ensuring life cycle quality?

This is a contractual matter, albeit one that is challenging but not insurmountable. The construction contract between the state and the contractor must do the alignment work that ownership bundling would otherwise do. It should specify construction design and quality conditions, validated through independent technical audits at transfer, with financial penalties for any shortfall to be clawed back from performance guarantees. Further, instead of waiting till the end, an independent engineer, appointed jointly but reporting to the grantor, should have enforceable sign-off rights at key construction milestones before disbursements are released. Done carefully, this shifts the incentive problem from ownership to contract design, which is solvable with the right institutional capacity.

All this requires a state with strong enough technical capacity to write demanding output specifications, enforce defect liability, and audit asset condition through the operational period. The bundled PPPs are an admission that since the state is weak and therefore cannot do the above, it must outsource the incentive problem by making the private party bear the whole lifecycle, even with the higher funding cost. 

This fatalism is the wrong long-run equilibrium, also because bundled PPPs, as we have seen, are prone to renegotiations, which themselves require high state capability. Accepting that the state must always bundle to avoid quality risks locks in a model where public borrowing costs are foregone forever, because the state never builds the contract-writing and monitoring capabilities to hold separated delivery accountable.

The better path is to treat the separation as a governance and capacity challenge, invest in that capacity, and capture the financing advantage of public capital as a permanent feature of the model rather than conceding it to the private sector in perpetuity. 

This is the standard approach followed historically in countries like Germany, Japan, and Singapore. The canonical example is Germany’s Autobahn model, followed by the National Highways Authority of India (NHAI), where the federal government finances motorways cheaply through public borrowing, while construction is let to private contractors under demanding output specifications with long liability tails. It is just as applicable to many infrastructure segments, such as urban infrastructure.

Weekend reading links

1. A fascinating article by Ed Conway documents how a cod fish caught off the Scottish coast travels 30,000 nautical miles to Rotterdam, Shandong, China (for manual deboning and skinning), and back to the UK, where it is breaded and sold as fish fingers.

2. AI solves the Navier Stokes equation breakdown.

Scientists plug in numbers to the NS equations to derive accurate answers for the behaviour of moving fluids. But the NS equations can sometimes return apparent nonsense, predicting things like infinite velocity and zero volume (called a “singularity”)... Can it be proved that the NS mathematics don’t allow infinite velocity? Or alternatively, can it be proved that this can happen in some theoretical cases, weird as that may seem? Mathematicians refer to this possibility as a “breakdown” in the “existence and smoothness” of NS. .. NS is one of the seven famous Millennium Problems that were listed in 2000. The Clay Institute offers a prize of $1million for a proof that, one way or another, addresses four specific conditions for NS. 

On Tuesday, September 8, OpenAI announced that it had found a solution to the NS problem by using a new AI model after 88 hours of intense computation involving 10,000 AI agents and 17 hours of verification. The agents exchanged 3 million messages and used 130 billion output tokens, which amounts to costs of well over $10 million. OpenAI claims its AI resolved two out of the four statements in the proof demanded by the Millennium Prize and that, indeed, infinite velocity is theoretically possible under certain conditions.

3. Brilliant description of the Dauphin discount, applied when a founder-owner hands over the reins to his entitled but not similarly competent children. 

4. There may have been an inflation regime shift in the US in the last few years. Consumer prices overall have risen more than 30 percent since the beginning of 2019. That’s two and a half times as much as they went up from 2012 to 2019.

Prices have risen sharply across the board.

5. US inflation clearly owes to the Iran misadventure. 

6. Benn Steil has one more reason for the rising US bond yields. 

The interest rate demanded by investors to absorb this debt appears not to be a straight linear function of its growth. Instead, more debt seems to accelerate the rise in the rate demanded. The reason lies in who buys the debt. Nineteen years ago, 76 per cent of US Treasury bonds were held by price-insensitive investors, such as central banks, who bought them more or less reflexively according to their stable reserve-management policies. Today, they hold only 43 per cent. The majority is now held by price-sensitive investors, such as households and investment funds, which demand greater returns as government debt grows and inflation erodes their purchasing power...

The twin problems of surging Treasury supply and stagnant foreign official demand will be exacerbated further still if new Federal Reserve chair Kevin Warsh ploughs forward with his stated ambition of reducing the central bank’s security holdings. The last episode of Fed balance-sheet reduction saw the share of Treasuries held by price-sensitive investors soar by 17 percentage points over 2022 to 2025, while the so-called term premium — the extra compensation demanded by investors to hold long-term debt — rose by 1.1 percentage points. Given the continued increase in price-sensitive investor dominance, further Fed balance-sheet reduction could see yet sharper rises in the price of long-term US debt.

7. VC wealth multiplication.

Founders Fund, for example, turned a roughly $600mn investment in Musk’s rocket, satellite and AI group into a stake worth more than $50bn at the company’s initial public offering, according to PitchBook estimates.

8. More on the collateral benefits of Aliko Dangote's business activities (HT: Adam Tooze).

Africa’s richest man, Aliko Dangote, wants his drivers, cooks and security guards to own a share of the continent’s largest refinery. At a signing ceremony in Lagos, Nigeria, on Monday, 7 September, Dangote Petroleum Refinery and Petrochemicals (DPRP) free establishment zone launched the formal process for what is expected to be Africa’s largest public share sale, aiming to raise ₦2.15trn ($1.4bn). The offer comprises 4.1 billion ordinary shares at ₦525 per share, with a minimum subscription of 10 shares. It opens on 14 September and closes on 13 October. “There is no segregation of who can own the share. We want every human being living on the continent to be part of this action,” says Dangote, president and chief executive of Dangote Industries Limited. “This is why we have called it the IPO for the people.” 

He said the offer, which opens on 14 September and closes on 13 October, is designed to allow “drivers, cooks, servers, our managers, everybody” to own a stake in the refinery. The IPO is targeting about 10 million retail investors, according to FirstCap Limited, one of the parties to the transaction. “This transaction is not just about the size of the offer; it’s about the significance and how it’s going to shape retail investment in Nigeria’s capital markets,” said Ukandu Ukandu, MD/CEO of FirstCap, describing the deal as digitally driven and built for mass participation. “Family members, security guards, drivers, colleagues, schoolmates, community members are all encouraged to participate.” Dangote described the raise as less of a financing exercise than a wealth-distribution one, calling the ₦2trn target “too small” and “a meagre amount” relative to the group’s needs. He said the proceeds are earmarked for the refinery’s expansion.

9. Tiruppur facts of the week.

Its knitwear exports rose from $3.3 billion in 2020-21 to $5.3 billion in 2024-25 (TEA, 2026). The cluster accounts for about 68 per cent of India’s knitwear exports and supports the livelihoods of more than a million workers, around 70 per cent of them being women. Within roughly 20 km, yarn, knitting, dyeing, printing, stitching, finishing, packaging and dispatch are woven into one production ecosystem. Nearly 20,000 units operate across different stages, from knitting, dyeing and printing to garmenting and ancillary activities. This is the ecosystem effect where firms specialise, workers specialise, and thousands of jobs are created around a common market.

10. India's affordable housing market facts.

The preference for midsize housing priced between ₹45 lakh and ₹90 lakh fell to 21 per cent in H1 2026, from as high as 35 per cent in H1 2021. Similarly, preference for housing priced below ₹45 lakh declined from 27 per cent in H1 2021 to 18 per cent in H1 2026. “This is also mirrored in supply, with affordable housing’s share of new launches across the top seven cities declining from 26 per cent in H1 2021 to just 8 per cent in H1 2026,” says Anuj Puri, chairman of Anarock.

11. KP Krishnan makes very important points while questioning the RBI's recent FCNR (B) scheme which attracted $136 billion.

The capital flow of $136 billion is brought in by providing currency hedging to commercial banks at no cost. A forward guarantee on the exchange rate creates a contingent liability for the state. If macroeconomic fundamentals guide the exchange rate to ₹110 per dollar in three years, the RBI will pay ₹15 per dollar. This is a cost of approximately ₹2 trillion... Such tricks have been done before. India Development Bonds 1991 raised an estimated 0.6 per cent of gross domestic product. Resurgent India Bonds of 1998 raised 1 per cent, and India Millennium Deposits 2000 raised 1.2 per cent of GDP. The 2013 FCNR(B) exercise raised 1.4 per cent of GDP. The 2026 FCNR(B) raised 3.5 per cent of GDP. The scale, the opacity, the complexity have gone up. ... The daily inflow and outflow across India’s borders is $11 billion. The global daily trading volume on the rupee is $140 billion. Throwing just $2 billion a day at the problem for 250 days will burn through $500 billion.

We should instead take one step back and ask the foundational question: All this drama is in return for what? Recent research by Hande et al. (2026, https://bit.ly/4dlXP5d) indicates that the natural annualised volatility of the USD/INR exchange rate, without intervention, is approximately 7.5 per cent a year. The apparatus of intervention generates a realised volatility of roughly 5 per cent a year. We suffer fiscal risk, distorted monetary policy, instability, and constraints on financial development, which impact the people through inferior economic growth. In return, we get a 2.5 percentage point reduction in currency volatility, which benefits a small set of business users.

12. This is a staggering statistic on IPOs 

Anthropic’s investors are expecting the company to reach a valuation of $2tn when it goes public in the coming weeks. Add in SpaceX, which began trading at $2tn after its IPO in June, and OpenAI, which is considering raising money privately at $1.2tn ahead of a public listing next year, and these companies alone could be worth well north of $5tn. Now compare that with the entire history of IPOs from 1980 to 2025. The 3,365 tech companies that went public in that period were worth a combined $4.1tn when they started trading, according to data compiled by Jay Ritter, emeritus professor at the University of Florida’s Warrington College of Business.

13. The return of conglomerates, but with overlapping national strategic goals.

In China, Huawei has gone from an empire with two business lines to a local champion with 70 per cent of sales at home in five divisions. Chinese car companies own their supply chains — BYD operates ships. To defuse sovereignty fears abroad, they are willing to forfeit control, using licensing and joint ventures. India’s Reliance and Tata have re-embraced nation building across industries, from cola to air defence. Even in America, Amazon and Alphabet have become conglomerates that own parts of their supply chains, such as chip design. SpaceX’s strategy is to be a techno-national champion with up to nine divisions, from asteroid mining to tourism. In the pursuit of state-endorsed AI, OpenAI has gone further than Amodei, offering Uncle Sam an equity stake. US pharmaceutical companies want new drugs invented in China, but geopolitics makes takeovers impossible. Instead, there is a surge in licensing. Bristol Myers Squibb and Pfizer have made bets worth up to $26bn. In Europe, some firms are diversifying to plug strategic holes. The parent of Lidl, a supermarket, does data centres. Renault is expanding into military drones. Governments are trying to dilute US tech companies’ control over their local subsidiaries.

14. Gig work platforms are a form of unemployment insurance, and are now running the risk of being disrupted by the likes of autonomous vehicles and robo-delivery agents. 

Gig platforms have indeed become a safety net for people who have fallen out of — or struggled to access — the formal labour market. A report by the World Bank in 2023 estimated there were between 154mn and 435mn online gig workers globally, representing between 4.4 and 12.5 per cent of the global labour force. Among the advantages of this development, according to the report’s authors, are that it “helps manage income shock” and “serves as unemployment insurance”.

China is perhaps the best example of a country in which gig work has served as a labour market shock absorber. Amid a prolonged construction downturn and the increasing automation of manufacturing work, the number of people who work as food delivery or ridesharing drivers in China increased by 10mn in two years to reach 53mn in 2025, according to estimates by the China New Employment Forms Research Center, a Beijing think-tank. 

15. A very good oped by Boddu Srujana raises the important point about what UPI has done for its users, especially those in the informal sector whose services are now captured in the formal net. She writes,

This system has made India’s informal workers extraordinarily visible to banks, lenders, the state and to the platforms that manage their labour, without making them correspondingly secure... A worker who has been absorbed into a real-time national payments system is traceable and taxable, still has no pension, no enforceable minimum earning, and no institutional means of bargaining... The JAM trinity — Jan Dhan, Aadhaar, mobile — was to deliver subsidies with less leakage. UPI extended that logic to everyday commerce by enabling a vegetable vendor not to worry about carrying cash, with payments settled in seconds rather than requiring a trip to a bank that takes hours. But this is not the same as securing income and financial inclusion. As operationalised in India, it has functioned as a substitute for the much harder political project of labour formalisation... 

A worker’s UPI trail has become the raw material for an entire private lending and “alternative credit scoring” industry, sitting atop a public rail, useful to someone who lacked collateral, but built with little meaningful consent. The same logic now governs platform work. UPI-enabled instant settlement is what allowed gig and delivery platforms to scale... it has done so by making these workers legally almost unclassifiable — employee or contractor — and by subjecting them to algorithmic discipline such as ratings and incentive structures, without any corresponding algorithmic right to contest a rating or know why an account was switched off.

16. Finally, some FDI facts about Indian states.

Nearly 82 per cent of all FDI equity inflows into India went to just six states or Union territories — Maharashtra, Delhi, Karnataka, Gujarat, Tamil Nadu, and Andhra Pradesh (including Telangana) — from January 2000 to March 2026... Among sectors, services had the highest share at 16.37 per cent in gross FDI equity inflows from January 2000 to December 2025, followed by computer software and hardware (15.62 per cent) and trading (6.55 per cent). Financial services accounted for the largest share within the services sector.

And several questions about the value of EoDB and Business Reform Action Plan (BRAP) rankings in determining FDI flows. 

States such as Maharashtra, Delhi, Karnataka and Tamil Nadu fared poorly in the Ease of Doing Business (EoDB) rankings released by the government in different years over the past decade. For instance, Delhi and Tamil Nadu featured outside the top 10 in 2015, 2016, 2017 and 2019, while Maharashtra and Karnataka’s best performance in these four years was the eighth rank... On the other hand, states like Madhya Pradesh, Chhattisgarh, Jharkhand, Rajasthan and Telangana had relatively better rankings but considerably lower FDI inflows to boot... None of the top FDI receiving states was in the top two categories in the latest ranking of 2024. Experts say EoDB and BRAP rankings are not the prime reason behind FDI inflows. Instead inflows reflect states’ infrastructure, access to ports, prosperity and presence of industrial networks.