A framework for the application of AI in public systems

I have not blogged about AI in development. One reason is that, notwithstanding several claims, I have not come across promising examples that have been successfully applied at a reasonable scale within public systems in the Indian context. 

I have blogged here, urging caution on the expectations of the impact of AI on development in countries like India. The central argument is that the binding constraint on AI impact in the public sector is rarely technical capability. It is the gap between what AI can do and what institutions are prepared to validate, adopt, and integrate, a gap shaped by the validation cost, regulatory frameworks, incumbent system stickiness, and the political economy of transition.

As an analytical framework for the application of AI in development, I can think of three lenses: the layer where the AI intervention is proposed, the nature of the problem to be addressed, and the constraints on deployment and adoption. 

In the first lens, AI can provide structured information to feed into decision-making; or directly enable decision-making by synthesising information into ranked options or risk scores, thereby augmenting human judgment; or deliver specified outputs in the form of drafts, transaction alerts, assessments, etc. 

The second lens can distinguish between complicated/complex and wicked problems. The former has many variables but is tractable with better data and modelling (crop yield prediction, supply chain optimisation, credit risk scoring, traffic signal timing, etc.), whereas the latter involves contested goals, adaptive actors, and systemic feedback (improve learning outcomes or skills, institutional reform, manage urban growth, etc.). While there are limits to AI, especially in the latter, complicated sub-problems within the former are amenable to AI. 

The third lens covers the gap between technical feasibility and deployed impact. This gap is shaped by validation cost (regulatory standards must be met), the incumbent system stickiness (when the AI solution demonstrates value that helps overcome the systemic inertia), trust and legitimacy (system or society must accept the efficacy and reliability), and political economy (overcome the entrenched vested interests). 

The three lenses are analytical dimensions, or orthogonal axes on which any individual intervention can be located. Then we have the horizontal or vertical use cases, which are a typological cut across the universe of interventions. It sorts use cases by their organisational footprint (does this sit in every department or only one) rather than analysing the nature of any single use case. 

The graphic below applies the three lenses to each item in an illustrative catalogue of horizontal/vertical use cases. 

As a prudent strategy, it may be useful to move first in low-friction domains - where there are no regulatory incumbents, where information asymmetry is high, and where the beneficiary is a motivated adopter - and use the evidence and trust built there to lower the political cost of adoption in high-friction domains. In this reading, as is the trend in the private sector, it will be some time, if ever, before vertical sectoral use cases of AI become efficacious and reliable enough to be adopted at scale. 

Based on the above, what are the most valuable and widest spanning productivity and efficiency increasing uses of LLMs by governments in developing countries (with low state capabilities)? 

I can think of two in particular - application in the generation of quasi-judicial and adjudication orders of all kinds, and in the recording of M-Book in engineering works. The breadth of their use, the stakes involved, the low current baseline, and the manageability of their deployment gaps make them areas with potentially transformative impacts. 

The first one ranges from charge sheets to orders in service matters and disciplinary cases, to regulatory and licensing decisions, assessment and adjudication by tax, land, and other authorities exercising statutory powers, including court and tribunals. 

Each of these has a standard structure - factual matrix, applicable legal provisions, consideration of arguments, findings, and operative order - that AI can populate with high-quality given structured inputs. The official provides the facts and the decision, and AI generates the legally coherent reasoning and formal language.

The value of this comes from the low baseline of generally poor quality of orders issued that suffer both from basic procedural/hygiene deficiencies and more substantive application of judgment. These deficiencies and lapses, especially of the former kind, immediately invite disputes. In fact, such poorly crafted orders become the starting point for a long series of administrative processes that manifest in disputes and harassment, appeals and litigations, that clog administrative bandwidth, lock up scarce capital, impair balance sheets, and generally waste effort and resources. 

These kinds of orders also contribute a very large share of the litigation against the government that clogs the judiciary. The major reason why governments lose such cases is the quality of orders in terms of poor drafting and adherence to basic administrative compliance and procedures. Poorly drafted orders are often set aside, not because the underlying decision was wrong, but because the reasoning was inadequately articulated. AI can have a significant impact in this area in both improving the quality of orders and even forcing systems to comply with procedural requirements.

The volume of such orders is enormous. A mid-level income tax officer may need to issue hundreds of assessment orders annually; a district collector may handle thousands of revenue proceedings. The drafting burden is a critical bottleneck in disposal rates. Further, AI-generated drafts can be trained to flag inconsistencies, missing procedural steps (e.g., failure to give opportunity to be heard), and citation of superseded legal provisions, thereby acting as a compliance check before the order is issued.

An AI application that generates the first draft of the order, for the adjudicating officer to revise and issue, can have a transformative cascading impact down the chain. The first draft can be hard-coded to serve as a forcing function to ensure compliance with a checklist of processes; the order itself could be validated, and its logic could also help with tightening the substantive exercise of judgment itself. 

On the face of it, this is a low-hanging fruit. All it requires is a library of orders of all kinds bundled together, and a template for different kinds of orders. An enterprise or team subscription to Claude or OpenAI, that has a Zero Data Retention (ZDR) configuration for their API (thereby ensuring customer data is not retained and used for training the algorithm), and which allows access to Claude through Amazon Bedrock or Azure OpenAI within a private Virtual Private Cloud, can address any privacy and security concerns on sharing internal data. 

This would be a quick deployment and superior to any indigenous or sovereign LLM model development (which should continue and could perhaps be deployed in parallel to learn and get refined). 

The latter work can be taken up in a mission mode by the National Informatics Centre or a public think tank (say, the National Centre for Good Governance, NCGG) by deploying a team of competent experts. It can consolidate the library of orders and develop standardised LLMs for a few high-volume and value use cases - disciplinary cases, GST and income tax adjudication orders, and orders on a few categories of land claims settlements. They could work with a few state governments, the CBIC (state indirect tax departments), and the CBDT.

The other high-value and promising application can be in the recording of the Measurement Book (or M-Book) for engineering works - roads, buildings, canals, drains, electricity lines, etc. It records the quantity of work done, certified by a junior engineer, against which payment is authorised. It is also one of the most fraud-prone documents in public works administration, susceptible to over-measurement, fictitious entries, and post-facto alteration. Besides, it is also a procedurally burdensome activity. Its importance arises from the fact that it is the basis on which contract payments are made. 

For certain categories of works, AI-assisted quantity and quality measurement from photographs and drone-based progress tracking and volume estimation can be very reliable. Smartphones equipped with computer vision applications can estimate dimensions and quantities from site photographs (lengths of pipes laid, areas of surface plastered, volumes of earthwork excavated). Computer vision models can be trained to assess the quality of construction work from photographs by detecting visible defects in concrete, checking alignment of masonry, and identifying sub-standard finishing. For large civil works (earthwork, embankments, reservoirs, large buildings), drone imagery processed through photogrammetry software can generate accurate volumetric estimates and three-dimensional progress models.

At the very least, AI-assisted photographs and drone outputs can be used to validate M-Book entries, surface quality concerns and other discrepancies. Once M-Book entries are digitised, AI can generate contractor bills from certified measurements, cross-check against contract rates, identify arithmetic errors, flag unusual patterns (e.g., a sharp increase in claimed quantities near bill submission deadlines), and route completed bills for authorisation. Apart from improving the quality and increasing accuracy of payments, the time savings and reductions in delays in payments will be significant. 

The examples of order-drafting and M-book opportunities are likely to be particularly appealing for officials because they also significantly reduce their workload and drudgery. They are deployable at medium friction, directly legible to senior officials, and capable of generating rapid, measurable impact that builds institutional confidence in AI tools. 

Another area of promise is the categorisation or triaging of cases in several public domains. 

The Ken has an article on the application of AI by courts in India. A promising application of AI is in the categorisation and triaging of cases.

Half of India’s backlog—25 million cases—could disappear quickly if AI were used not to decide cases, but to triage them. The opportunity lies in the huge volume of matters that are effectively already dead on arrival: cases filed beyond limitation, petitions missing mandatory disclosures, matters rendered moot by changes in law, or disputes where precedent has already settled the question, leaving only a narrow technical point to close. Courts identify these only in occasional manual clean-ups—slow, inconsistent, and impossible to scale. 

An AI system, by contrast, could scan filings in bulk, flag statutory defects, detect outdated or defective pleadings, classify time-barred matters, and surface cases appropriate for summary disposal. These could then be bundled and placed before judges for quick orders, clearing the undergrowth so courts can focus on disputes that actually require judicial time.

Like with court cases, triaging of outpatient (OP) cases in primary health centres (PHC), community health centres (CHC), district hospitals, and medical colleges is an area where AI can play a potentially significant role. 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. 

In all these hospitals, OP cases come to doctors with limited or no triaging. An AI-based triaging application where the symptoms are entered at the OP-registration, nurse and doctor-level, can significantly improve work conditions, increase hospital productivity, and enhance the quality of treatment. 

Similarly, citizen grievances or consumer complaints received in any office, especially public-facing ones, can be triaged for routing them to the right desks/officials, escalating to senior officials, analysing repeat complaints, and so on. 

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.

In general, the judiciary’s case load management is an area where, in theory, AI applications can have a transformative impact, especially in categorising cases, transcription of witness statements, and the preparation of orders. The Ken article has a good description of an AI application in a courtroom in Kerala.

A witness spoke; an AI-powered tool listened. A clean, searchable transcript appeared in real time—punctuated, structured, permanent. Stenographers were not clambering to catch up, litigants were not begging for readable copies. This split-screen view—one courtroom running on memory, the other on machine comprehension—is not a metaphor. It is India’s judiciary in 2025: a system where 19th-century workflows and modern AI systems operate side by side, neither quite replacing the other. Earlier in 2025, the Kerala High Court issued an office memorandum, making AI-assisted live transcription mandatory across all district courts… 

“Today, the AI-transcribed witness testimony is uploaded as soon as the proceedings are completed. Earlier, the witness testimony would be handwritten by the judge, and a party’s lawyer would apply for a readable copy and seek adjournment on that basis,” said Joseph Rajesh, the IT Registrar of the Kerala High Court. “All this time delay is now cut down to nil.”… A court order that once took four or five hours to type can now be generated in under an hour. A witness deposition that once required handwritten dictation, shorthand transcription, and final review can be captured in a single digital stream. 

The article also has this graphical exploration of the various possible use cases for AI within the judiciary.

There can also be daunting practical challenges to its adoption

Consider defect detection—the clerical step where filings are checked for completeness. In Delhi’s Tis Hazari courts, the rules are so complex that Nyaay’s engineers joked: If we can crack defect detection here, we can crack it anywhere. The joke holds true. AI had to be customised court by court. And transcription? In many urban courts, there are stenographers, but in district courts, they are scarce. AI fills a vacuum, but only if the court has electricity, microphones, and a judge willing to trust the output.

So how can these AI applications be deployed?

For all these use cases, the development of robust AI applications that can be deployed across institutional levels nationwide is an industrial engineering endeavour. It requires diligent, long-drawn, and high-quality problem-solving, including pilots (or beta testing), iterative adaptation and refinement. It cannot be achieved by an experiment undertaken by a district or state-level entity (or even as ad-hoc officer-driven initiatives within central government departments), as is often the case today. 

The Ken article cited above points out the problems of not having one entity in charge of the creation of these national public goods.

The downside is that every state is effectively running its own lab experiment: non-standard, improvised, and dependent on the priorities (or disinterest) of whoever happens to be leading the High Court. Some innovate aggressively. Some slow-walk. Some judges independently approach AI companies, but without formal approval, nothing can enter the courtroom workflow… High Courts and district courts continue testing tools piecemeal. Language models multiply while vendors proliferate and workflows diverge… India could end up with 25 distinct judicial AI ecosystems, none interoperable, standardised, or scalable nationally… 

India’s AI push is happening largely outside the Supreme Court’s flagship e-Courts project—the initiative meant to modernise the judiciary. Phase III of the programme still focuses on the basics: creating digitally readable records, enabling e-filing, and automating service of summons. A February Press Information Bureau note pegs the Phase III budget at Rs 7,200 crore, but only a little over Rs 50 crore is reserved for “integration of AI and blockchain technologies across High Courts”. In effect, the national blueprint is still laying the plumbing while the states are already experimenting with smart faucets…

If India manages to fuse Kerala’s mandate with Karnataka’s case clustering, Delhi’s defect detection, the Supreme Court’s governance frameworks, and the entrepreneurial urgency of tools like Adalat and Nyaay, the judiciary could undergo a systemic leap.

It would require a dedicated team at the national level (government department, Supreme Court, CBIC/CBDT, etc.) engaging single-mindedly on the endeavour - formulating the problem, collecting and cleaning data, developing models for different levels of government, undertaking pilots with close oversight, iterating with tight feedback loops, documenting processes, and scaling solutions gradually.

The models required for lower courts, High Courts, and the Supreme Court would vary and must therefore be customised, just as those required for hospitals of different kinds, tax and land adjudicating and appellate officials at different levels, and disciplinary authorities for different kinds of functional entities. In each case, there would be a need to do rigorous pilots to be able to refine and finalise robust enough solutions that can be deployed at scale.

Weekend reading links

1. PE firms sitting on $4 trillion of unsold assets, on investments largely made between 2020 and 2022 when rates were slashed to zero, are finding creative ways to offload them. Sample this

Blackstone is marketing a so-called collateralised fund obligation that will bundle more than $2bn of stakes in leveraged buyout funds into bonds to sell to investors and insurers, according to people familiar with the matter. The deal would provide an infusion of cash to investors in a Blackstone Strategic Partners fund, the firm’s unit that invests in other private equity groups’ funds. It is unclear if Blackstone will ultimately go ahead with the securitisation or seek to sell the stakes in a secondary transaction, one person briefed on the matter said... The vehicles, which are sliced and diced to give investors exposure to different levels of risk and return, have boomed. Issuance of CFOs soared to a record of $25.9bn last year from a modest $4.8bn in 2021, according to credit rating agency KBRA.

2. This is a brilliant articulation of the problem with articulating something purely in terms of absolute numbers and aggregates.

In his novel Hard Times, Charles Dickens described a girl called Sissy, who was having a terrible time in her lessons. Her schoolmaster told her to imagine that her schoolroom was a nation in possession of “fifty millions of money”. Wouldn’t that mean it was a prosperous and thriving state? “I said I didn’t know,” she relayed afterwards to a friend. “I thought I couldn’t know whether it was a prosperous nation or not, and whether I was in a thriving state or not, unless I knew who had got the money, and whether any of it was mine. But that had nothing to do with it. It was not in the figures at all.”

3. Interesting story about how old companies are reinventing themselves to profit from the AI-boom.

AI servers must be more tightly linked together, increasing the need for advanced cabling and optics. Shares in Corning, the 175-year-old inventor of Pyrex glass that also supplies screens for Apple’s iPhones, have increased by more than 270 per cent in the past year after it signed deals with Meta and Nvidia to supply optical fibre cabling to AI data centres. The vast amounts of electricity needed for AI training are also fuelling demand for specialised power management, high-voltage electronics and cooling technologies. This has led to big interest in traditional suppliers of electrical equipment, typically deployed in residential and industrial projects. Eaton, an Ohio-based power management company, received 240 per cent more data centre orders in Q1 this year...

 

French electrical equipment maker Legrand has doubled its revenues this decade with half of the growth coming from data centres, which now make up more than a quarter of its turnover. Air conditioning and liquid cooling — using water to stop chips from overheating — are in demand too. Shares in AC maker Comfort Systems USA have shot up 260 per cent over the past year, while Schneider Electric bought a stake in data centre liquid cooling specialist Motivair for $850mn last year. Utilities are rushing to supply AI companies with power — including Spain’s Iberdrola, a leading supplier of power contracts to tech groups in Europe, according to Pexapark data, and Entergy in the US, whose share price hit a record high after a $10bn deal with Meta... Several generator and engine companies have also pivoted to supplying data centres, including Caterpillar, Boeing supplier Howmet Aerospace, Finnish ship engine maker Wärtsilä and Baker Hughes, which formerly focused on oilfield services.

4. China demographics facts of the week.

This year’s cohort of gaokao-takers were mostly born in 2008, a year of 16.1m births. By 2025 births had more than halved, to just 7.9m. The demographic cliff is already visible in nurseries, which saw pupil numbers plummet from 46m to 32m between 2022 and 2025. Numbers in primary schools have also started to thin. Inevitably, over time, secondary schools and then colleges will follow.

And technology adoption

A survey of 322,000 students last year by the China National Academy of Educational Sciences, a state-affiliated think-tank, found that 85.6% of them had already tried using AI to complete their homework. On popular apps such as Zuoyebang (“Homework Help”) and Yuanfudao (“Ape Tutoring”), pupils snap photos of questions and ai walks them through the solutions. (Teachers are using similar technologies to help mark homework.)

5. Public moods on the role of government in the UK - 70% support nationalising energy and 82% water.

Rail subsidies have been rising, £12bn in operational support in 2024-25, up in real terms from £2bn in 2000-01. 

6. Indian economy facts of the week.

The World Trade Organization data show that for non-agricultural goods, the share of tariff lines in the category 10-15 per cent increased sharply from 1.4 per cent to 33.5 per cent between 2014 and 2024 and for the tariff category 15-25 per cent from 1.7 per cent to 14.9 per cent, while the share of tariff lines in the category 0-10 per cent fell steeply from 90.5 per cent to 42.5 per cent over the same period.

7. From Thomas Astbridge's book on the Black Death 

In its most intense phase, from 1347 to 1353, the Black Death killed more than 100mn people, or about half the population in the areas infected, Asbridge estimates. This makes it more lethal than two other great pandemics — the 6th-century Plague of Justinian in the Mediterranean and the pestilence that swept across Asia from the mid-19th century until the aftermath of the second world war — and far worse than Covid-19 in our times... Asbridge demonstrates that the Black Death was probably more devastating in cities such as Cairo and Damascus than in, say, Constantinople or Florence. In Cairo, a metropolis of 500,000 people, almost 10 times larger than London’s population, perhaps 250,000 died, Asbridge suggests.

8. On the role of luck in football tournaments.

According to one study of historical matches, the chances of the team with the worse record winning was 45 per cent in football, compared with just 36 per cent in America’s National Football League. (Yes, this pep talk has statistics. Bite me.) The knockout structure raises the role of chance, as just one dodgy penalty can crash a team out of the competition. According to numbers crunched by James Tozer of Prospect, a sports analytics company, betting odds gave the top four teams in the most recent Premier League a combined 89 per cent chance of winning (after adjusting for bookies’ ability to take advantage of fans’ optimism that their team would win). In the World Cup an upset is more likely, as that figure is only 48 per cent.

9. Aldi effect, as the discount grocery retailer seeks to expand aggressively, as it envisages 4000 stores at an investment of $9 billion in a US market where consumers are facing higher prices due to persistent inflation.

Credit card data analysed by the bank found that when an Aldi store opened, it shaved an average of one percentage point off annual sales from competitors within a 10-mile radius... Aldi prospered in postwar Germany under brothers Karl and Theo Albrecht before a disagreement led to a split in the 1960s. One offshoot, Aldi Süd, oversees Aldi’s US business after opening the first store in Iowa in 1976. The other, Aldi Nord, owns the quirky US grocer Trader Joe’s. The discounter’s stores are austere places with only about six staff on duty. They are designed for maximum efficiency: groceries are shelved without leaving their cardboard shipping trays and oversized bar codes are printed on packing so checkout operators can scan at pace. Customers must deposit a coin to obtain a shopping trolley, which is refunded if they return it. Operating cost savings fund the chain’s low prices... Aldi’s compact stores, which stock only about 2,000 product lines, are often located near competitors such as Walmart, whose large-format stores carry about 120,000 products, including low-priced groceries.

10. Andhra Pradesh shrimp production facts.

India exports approximately 8 lakh tonnes of shrimp a year, with Andhra Pradesh accounting for over 60 percent of production. The state accounts for 80 percent of the country’s shrimp exports and 34 percent of marine exports, valued at around Rs 21,246 crore annually. The state has 2.5 lakh aqua farmer families, of which 2 lakh are small and medium farmers. Another 30 lakh people depend on allied sectors. According to the Union Ministry of Commerce and Industry, India exported a record 17,81,602 MT of seafood worth US$ 7.38 billion (Rs 60,523.89 crore) in 2023-24, of which frozen shrimp alone accounted for 92 percent — a significant share from Andhra Pradesh.

11. India's PPP pioneers

GVK’s 216-megawatt (Mw) Jegurupadu plant became an early proof-of-concept under a power purchase agreement. IL&FS built a 12-km toll road between Rau and Pithampur in Madhya Pradesh, marking India’s first private toll concession.

12. This is a true success story for the Indian economy.

Between 2020 and 2026 the number of Indian retail investors rose from around 40m to 130m.

The quartet of global imbalances

President Donald Trump’s single-minded pursuit of rebalancing the US economy using tariffs is unlikely to yield much without addressing fundamental distortions that have crept into the US economy and financial markets. Trade surpluses are a mere symptom of deeper malaises. I had blogged about the global twin structural imbalances, arguing that an American economy skewed towards consumption and a Chinese one skewed away from consumption are two sides of the same coin. 

Helene Rey (see also this report to G7) has a very nice summary of global imbalances, where she locates it within the dynamics of saving and investment, and questions the focus on tariffs.

A country that saves more than it invests lends abroad and runs a current account surplus; one that invests more than it saves borrows and runs a current account deficit. It is the collective saving and investment decisions of a country’s households, companies and government that drive imbalances. There is now some agreement — crystallising in the G7 discussions — that the sources of imbalances are linked to unbalanced growth models and mostly made at home: chronically weak consumption in China, feeble productive investment in Europe and outsized fiscal deficits in the US. Tariffs are not an effective mechanism to change any of those and issuing an international currency is no justification to run current account deficits. 

She also proposes the solutions in terms of complementary actions. 

China rebalancing towards consumption, Europe lifting productive investment and America repairing its public finances are not three grudging favours. They are parts of a single, mutually reinforcing policy. One country’s exports are another’s imports; one country’s capital outflow is another’s inflow. When all three move at once, each adjustment cushions the others: the deficit country that consolidates finds external demand waiting as surplus countries spend more at home, and the surplus country that stimulates demand finds a market at home rather than a protectionist wall… The IMF’s scenario of simultaneous rebalancing raises global output by around 0.8 per cent and narrows medium-term imbalances by half a percentage point of world GDP.

I think this analysis misses a critical fourth leg of the imbalance, financialisation. It cannot be seen as merely a symptom of US borrowing. As evidence, there are two natural experiments. The last two episodes of US fiscal deficit reduction (in the late nineties and mid-2000s) were accompanied by financial market bubbles (the dot-com bubble and housing and mortgage market bubble). 

Underlining this, Ricardo Caballero, Emmanuel Farhi, and Pierre Olivier Gourinchas have shown that if the financial markets do not work well, the economy might accommodate investments that deliver a rate of return that is below the growth rate of the economy. In this situation, both stock market bubbles and government debt can play the useful role of displacing inefficient investments. 

Foreign savings flowing into the US must find a US asset to absorb them - either Treasury debt (the fiscal deficit channel) or private financial assets (the financialisation channel). These two are substitutes, not complements. Squeezing fiscal deficits without restraining private credit pushes the same global savings glut into asset bubbles. So US fiscal consolidation alone is not sufficient, and must be paired with financial-sector reform that prevents the likes of private credit from filling the void.

I asked Claude to generate a graphic that describes the global imbalances quartet. 

So, how to address these imbalances?

The rebalancing would require global diplomacy and coordination to mobilise support from all key stakeholders. This looks onerous in a deeply polarised world, of rising tensions between the West and China and the unpredictable and whimsical nature of the Trump Presidency. 

In fact, among the four adjustments required, interventions pertaining to the much-derided Euro area appear to be the most promising and likely to materialise. In fact, on investments, the train has already started. 

It is possible that a deep crisis on the economic front, a very likely near to medium-term possibility in either case, may force both China and the US to rebalance towards consumption and consolidation, respectively. However, there are daunting political economy challenges to be overcome in both countries, especially the US. 

It is the fourth leg that might prove the most challenging. The financial markets are where the power of entrenched interests is so strong that it might require a counter-revolution to upend the order and regulate financial markets more tightly. This will also require global coordination and collective action, and not mere reforms at the US front. 

Countering China's weaponisation of its manufacturing dominance

China’s weaponisation of its manufacturing dominance, most famously through its control of rare earth magnets production, is generally accompanied by a narrative of resignation that its trading partners must live with this reality till they develop alternative supply chains. It is widely perceived that China has a definitive upper hand, and all other countries, including the US, must play catch-up. 

This begs a few questions. Isn’t China susceptible to imported products and services that are essential to its economy? Aren’t there rare earth equivalents that the US and Europe can restrict access to China, thereby bringing a bargaining equivalence between the two sides? What are those rare-earth equivalent dependencies for China today? What is the economic leverage that the West has over China that can be exercised in response to the rare earths restrictions? 

This post will examine this question in greater detail.

In a 2025 G-7 summit, the European Commission President Ursula von der Leyen aptly described China’s industrial policies as creating a pattern of “dominance, dependency, and blackmail”.

Having built up dominance and dependency, the blackmail is now intensifying. The hide your strength and bide your timephase is past. In the recent past, China has come up with several measures to leverage its dominance. 

China’s widely known weaponisation of its manufacturing dominance has been underpinned by a series of regulations that impose restrictions and penalties. Pre-empting efforts by Western multinationals to diversify away from China, in April this year, to “prevent security risks in industrial and supply chains”, the State Council issued Regulations on Industrial and Supply Chain Security to investigate and punish foreign firms that stop using Chinese suppliers in response to political pressure from their governments. This is a summary.

Under the new rules, regulators can question employees and examine corporate records during investigations. The regulations also allow the authorities to bar companies and individuals from leaving China if they are suspected of moving supply chains elsewhere under foreign pressure… The State Council, China’s cabinet, justified the measures as necessary to protect the country’s economic stability and national security… China’s global network of ports and port-management software gave Chinese officials detailed insight into multinationals’ supply chains, allowing them to detect when companies shift to suppliers elsewhere.

In February, it amended the state secrets law by broadening the scope of the type of information that would be considered a national security risk. It includes a new legal concept called “work secrets”, defined as information that is not an official state secret, but that “will cause certain adverse effects if leaked”. This broad sweep allows for interpretation as convenient for the government, and makes foreign companies and their employees further vulnerable. 

The restrictions are not confined to foreign companies. In early June, the State Council announced rules requiring national security screening for Chinese companies seeking to invest overseas. 

The rules also give the authorities new powers to scrutinize Chinese companies seeking opportunities abroad, subjecting them to national security reviews that place investments into one of three categories: encouraged, restricted or prohibited. Part of the motivation for this, lawyers say, is to keep money, talent and intellectual property in fields where China has a competitive edge from leaving the country… The measures restrict the movement of certain talent in sectors deemed sensitive, though Beijing has not defined which sectors qualify. They also give officials broader authority to review the movement of capital, including the power to force investors to sell shares or halt investments if national security concerns arise. The rules also lay the legal groundwork for regulators to bar foreign entities from investing or operating in China, including expelling them from the country, in retaliation for actions taken by their governments against Chinese investments.

In this context, it is surprising that even as China increases its bellicosity both in trade and in its foreign policy, especially the breadth and frequency of military actions in the Taiwan Straits, the response from the West has been remarkably muted. Doubtless, the dysfunctional nature of the Trump Presidency has been a major contributor. But the lack of proportionality of response from the US holds, even including the Biden Presidency. 

So what are the chokepoints and vulnerabilities that China faces from the West?

Ironically, China’s biggest vulnerability is in the very industry that it utterly dominates. China dominates the downstreamof electronics (assembly, packaging, volume manufacturing) and the upstream of raw materials (rare earths, gallium, refining), but it is deeply dependent on imports for the midstream - the tools and ultra-pure materials that actually make advanced chips. Underlining this reality, China imported $385 billion of integrated circuits in 2024, more than the $325 billion it spent on crude oil, making chips its biggest single import. 

China's vulnerabilities cluster at the highest-precision, most knowledge-intensive nodes, the "tools that make the tools." Photoresist is the cleanest analogue to rare earths: a narrow, chemically exotic input where Japan holds a near-monopoly and a cut-off would, in one analyst's framing, leave Chinese manufacturing with "no rice to cook with". China could meet only about 5% of its own demand for the KrF resists used in 110–180nm chips, and high-end localization was under 5% in 2022. In simple terms, semiconductor equipment is the broadest lever by value, and machine tools are the most pervasive across general manufacturing. More than 60% of the indigenous passenger jet C919’s components, including engines and flight controls, are imported.

The graphic below plots the levers (or chokepoints) that the West have over China in terms of how hard it bites and how long it would take China to replace them. The colour represents the usability of the lever, without unacceptable self-harm; the dashed arrows show the levers China is actively closing.

The red bubbles (EUV, photoresist, EDA) are near-monopolies a single country can switch off. The amber ones (equipment, foundry, CNC tools, bearings, jet engines, instruments) are oligopolies that only work as leverage if allies coordinate, which is exactly why the US has spent two years building the Netherlands-Japan-Germany-Taiwan coalition rather than acting alone. 

The West’s strongest cards (AI chips, EDA software, jet engines) are precisely the ones with a shelf life, because Beijing is pouring state money into domestic substitutes. The genuine bargaining equivalence lives in the top-right quadrant - frontier technology that bites hard and takes a decade-plus to replace (EUV lithography, leading-edge foundry). 

It is to be noted that the arrows point left because it points to China travelling in the direction of self-sufficiency. It highlights that while real and binding today, there is a declining shelf-life or option value with these chokepoints.

However, having said this, any escalation risks retaliation. In 2025, the US suspended engine and EDA exports, China tightened rare earths, and within weeks, both sides walked it back into a one-year truce because each could hurt the other badly. It also points to the value of a bargaining strategy where the West gradually introduce restrictions on multiple such products where China depends on imports. The restrictions should be phased in carefully and subtly by plugging the procedural and process links that China exploits to its advantage. 

In this backdrop, Ely Ratner and Nick Danby have a very good essay in Foreign Affairs that outlines the broad contours of a plan to identify and squeeze China’s vulnerabilities and do unto it what it is clinically doing by weaponising its strengths. The most obvious one is, as discussed above, to tighten the restrictions on access to the semiconductor chip supply chain.

China continues to leverage chip-smuggling networks, overseas data centers, and model distillation, a technique that exploits access to frontier AI models to replicate their capabilities. New policy measures should target the channels China uses to acquire restricted chips and supporting architecture, including shell companies and unlisted subsidiaries, as well as cloud-based access to U.S. computing power and servicing arrangements that keep older semiconductor manufacturing equipment operational. Equally urgent is synchronizing U.S. export restrictions with those of the Netherlands and Japan, whose companies—ASML and Tokyo Electron—control critical chokepoints in the advanced semiconductor supply chain. Although both governments began strengthening their own policies in 2023, their controls on equipment sales, servicing, and subcomponent exports to Chinese fabrication plants and toolmakers fall short of U.S. restrictions. Washington should press The Hague and Tokyo to close these gaps. If diplomacy fails, it should consider invoking the Foreign Direct Product Rule, which extends the extraterritorial reach of U.S. export controls to restrict products made with U.S. software or technology.

The authors also write that China’s huge export volume and $1.2 trillion trade surplus can be as much a liability as it is a strength, especially at a time when the economy is weakening and struggling for anchors of growth.

It should push back against China’s export surge by bringing advanced economies facing deindustrialization together with developing countries whose own manufacturing aspirations are being displaced. This coalition could then coordinate trade measures to protect their industries, including steel, shipbuilding, batteries, and drones. Alongside tariffs, the United States could pursue high-standard trade agreements that institutionalize requirements for subsidies, state-owned enterprises, and forced technology transfers that China cannot meet. Like-minded partners could also create an anticircumvention regime by strengthening rules of origin, sharing customs data, and imposing penalties on goods routed through third countries to avoid trade restrictions. They could further impose outbound investment screening to prevent companies or individuals in the United States and allied countries from financing Chinese capabilities that the controls seek to limit.

Notwithstanding its large reserves, for a country which imports three-fourths of its crude oil with 90% delivered through vulnerable sea routes, China is extremely vulnerable to energy security. 

Below the threshold of a full blockade, the U.S. Treasury Department, through the Office of Foreign Assets Control, can use maritime sanctions to dissuade shipping companies, insurers, brokers, and banks from supporting prohibited shipments. Pressure on insurance, port access, and flag registration would raise costs and create uncertainty for China-bound tankers without requiring direct military action. The Pentagon should nevertheless demonstrate its ability to disrupt or interdict China’s seaborne energy imports by exercising U.S. naval control over key chokepoints along energy trade routes.

Commodity imports are another chokepoint

China imports roughly 80 percent of its iron ore, a foundation of its steel industry, predominantly from Australia. And most of its copper and lithium inputs, which are critical to battery and defense manufacturing, come from Australia, Chile, the Democratic Republic of the Congo, and Peru. As with oil, these dependencies offer additional pressure points that can be leveraged to strengthen deterrence and compound China’s challenges across multiple sectors simultaneously. If Australia were prepared to restrict exports of iron and lithium ore, and the United States and its partners had a plan to tighten access to copper and cobalt, they would send a message to China that its industrial base could be easily disrupted and its defense production capacity degraded if circumstances warranted.

Finally, the US dollar’s dominance is the nuclear option available.

Were Washington to restrict China’s dollar access—moving from sanctions on banks supporting PLA activities to broad limits on dollar transactions in advanced technology and military manufacturing—it could impose severe costs on Beijing, disrupting Chinese financial markets and potentially triggering wider economic instability… Washington must prepare for this scenario by first communicating unambiguously that only severely destabilizing acts would trigger consequences of this magnitude: for example, large-scale cyberattacks on critical U.S. infrastructure, Chinese export restrictions that seriously imperil the U.S. economy, or an armed attack against U.S. allies and partners.

The Cold War and trade tensions between the West and China are here to stay for the foreseeable future and are most likely to be ratcheted up over time. It is therefore important that others can mobilise sufficient bargaining levers with China. All the aforesaid are likely to be very effective in restricting the Chinese economy if deployed in a coordinated manner. This would require the mobilisation of a global alliance, something the US-led West did with great effectiveness during the Cold War with the Soviet Union. 

Now, with the hostility and dysfunctionality of the Trump administration, any such cohesive and credible global alliance looks very unlikely. In its absence, whatever restrictions are imposed by the US and EU independently are merely band-aid solutions, and likely to get circumvented in various ways against an antagonist who is disciplined, plays the long game, and does painstaking groundwork to accumulate its strengths and overcome restrictions. Trump 2.0 is, therefore, perhaps the best thing that an embattled Chinese government could have been gifted by its opponents. 

Examining the gravitational pull of index investing

The blockbuster IPO of SpaceX has drawn attention to the role of the index investing market segment, which originated fifty years ago. The scrutiny will only intensify in the days ahead as both Anthropic and OpenAI are listed. 

Despite only about 4 per cent of SpaceX’s shares being listed, a high level of demand was built in, also because some index investors will soon be required to add the company to their portfolios. Nasdaq amended its rules (as also CRSP and FTSE Russell) to allow SpaceX to enter the Nasdaq-100 index via fast-track, thereby providing it free liquidity through a captive market. Morningstar follows 6,006 US-registered mutual funds and 5,100 ETFs, covered by 3,203 separate benchmark indices against which $41.1tn of assets are managed. 

The current entry requirements for indices are very liberal for large companies.

However, what distinguishes SpaceX is its very low float (just 4.25% compared to above 90% for the other big companies) and the fact that it does not yet make any profit. Toby Nangle and Co at FT Alphaville tabulated that active fund managers (who want to ignore Musk) and passive portfolio managers (who have no choice) must purchase $14.2bn of SpaceX stock in the first three weeks of trading to avoid going short. 

We think that managers who have absolutely no interest in going either long or short Elon will need to collectively buy $8.5bn of SpaceX stock on the 19th June, a further $1bn on 26th June and a final $4.7bn on July 3. And so we’re talking a cumulative, de facto mandated $14.2bn of mutual fund and ETF orders by July 4 to avoid having to take a view on Elon. That number would’ve been around $11bn higher if the S&P 500 index committee had leaned a different way. But it’s $13.2bn bigger than the $1bn it would’ve been if index committees had sat tight on their existing fast-track methodologies. And this is before we count institutional assets like pension funds, insurers and foundations, as well as every single foreign owner whose benchmarking habits we currently lack information.

So this is a big deal. Clearly, the market has been brazenly manipulated by changing the rules to create the stage for SpaceX (and the other two mega IPOs). 

I took the help of Claude to generate a few graphics to understand the scale of index investing, its dynamics and distortions, and possible reforms. 

The conventional measures dramatically understate index investing because they count only labelled index mutual funds and ETFs. However, using the indexation definition, passive ownership of global equity mutual funds and ETFs is 50% and 60% for US equities. In fact, a stunning three-fourths of the equity market exposure of the big three US asset managers is through ETFs.

SpaceX’s shares could have an index exposure of 43% in a year when it joins the S&P 500, and that too on a volume which would be a fraction of that for the Mag7 firms. It is estimated that S&P 500 funds would need to absorb 19% of SpaceX's public float upon inclusion, with the Russell 1000 and Nasdaq-100 funds absorbing another 24%. A tiny supply of $22-27 bn would meet a mandatory demand of 43%. 

The research on the impact of index inclusion on stock price offers striking findings. Gabaix and Koijen find investing $1 in the stock market increases the market’s aggregate value by about $5. Since marginal holders of equity (index funds, pension funds, insurance companies) operate under mandates that fix their equity allocations within narrow bands, when aggregate demand shifts, few participants can absorb the change, and prices must move substantially to clear the market. 

Haddad-Huebner-Loualiche point to a Mathew Effect in indiex investing. They find that a $3.6trn market cap company is only about five times as liquid as a $100bn company, despite being 36 times the size. The largest stocks are disproportionately impacted by each dollar flowing into passive funds, causing the largest stocks to outperform and stock market concentration to rise. Passive investing has reduced market efficiency by over one-third.

Index investing brings the benefits of access to a low fee, diversified, and tax efficient asset pool to the retail investors. In the words of the legendary investor Jack Bogle, index investing is the equivalent of not looking for the needle but instead buying the haystack. 

However, it distorts price discovery, creates inelastic demand, amplifies concentration in stocks and asset managers, and erodes governance. All of these distortions are greater for the mega-cap stocks. 

Further, index-investing comes from asymmetric flow elasticity. While passive funds are mechanically symmetric (the 1:5 multiplication), the rest of the system (margin lenders, derivatives dealers, momentum funds, retail behavioural agents) is not. When flows reverse, the multiplier still applies, but additional positive feedback channels switch on, creating an overshoot below fundamental value before mean-reversion can bring prices back. The resultant instability can tip markets into prolonged and deeper downturns. This is an illustrative example

Suppose 2 years post-IPO, growth disappointments and rate-policy reversal trigger a ~20% net outflow from Nasdaq-100 trackers ($280 bn × 20% ≈ $56 bn). Applying the firm-level multiplier (~2 for stock-specific flows, higher for concentrated names): mechanical impact on SpaceX share price could be ~30–45%, on top of any fundamental revaluation. The same flow proportionally affects all Mag 7. Index investing does not just amplify upside — it removes the price-discoverers who normally cushion downside.

Below is a listing of some channels that amplify the downside. 

All things taken together, index investing must reconcile the conflicting requirements of democratising investing (through passive funds) and limiting market instability. The answer lies in the empirical fact - the more capital becomes index-tracking, the less price discovery the market performs, and the more the system is exposed to flow-driven valuation and procyclical unwinding. Clearly what is optimal for individual investor is not optimal for market function. In the circumstances, this can be an illustrative framework to reconcile the conflicting requirements. 

The problem with index investing is that of internalisation of its negative externalities. Passive investors capture the upside of cap-weighted indexation (low fees, diversification, momentum) without bearing the price-discovery cost, whereas active investors bear the cost of analysis but cannot compete on fees. 

This free-riding can be resolved only if passive investors internalise the marginal cost they impose on the system without eliminating the substantial benefits they deliver to retail investors. Like with carbon emissions, this can be done through fees, governance obligations, or structural caps. 

To dive a bit deeper, there are perhaps four levers for reforming index investing - index methodology (rules of index construction), voting and governance obligations, fund-level liquidity thresholds, and limiting structural concentration in the passive investment ecosystem. A combination of all of them is a formidable regulatory toolkit. 

Each lever will have its set of reforms. It may be required to prioritise a limited set reforms that have the highest-impact-per-cost. The simplest and highest value reform could be mandatory free float minimums (say 10-15%), a twelve-month public trading history before index inclusion, and a 4-5% cap on any single stock in an index. Second, bring index providers under SEC regulation as “investment advisers” subject to fiduciary duty and disclosure requirements. 

Third, for passive funds with more than 1% holding, return governance to the ETF or index fund holders by allowing them to vote their proportionate share directly via the fund, thereby reducing the influence of the Big Three asset managers. Fourth, cap the passive fund ownership of US banks at 10% (the regulatory "controlling interest" threshold) and impose mandatory liquidity stress tests for funds with more than $5 bn AUM in cases of extreme outflows and market dislocation. 

Fifth, the Department of Labour should tighten defined-contribution retirement-plan portfolio diversity rules, thereby providing a structural counterweight to cap-weighted default portfolios. Such plans must offer either equal-weight or capped-weight default or cap cap-weighted defaults at e.g. 80% of the plan portfolio. This would reduce forced concentration in 401(k) default portfolios. Finally, there could be regulatory roadmap for orderly winding-down of overweight positions during the life-cycle or retirement transition. A coordinated guidance, including some rule-based gradual decumulation, reduces fire-sale risk. 

The proposals above seek to reconcile the diversification and fee benefits of indexation while also internalising its costs. Once index providers face fiduciary duty, once Big Three votes are partially passed through, once retirement defaults include non-cap-weighted options, once liquidity is stress-tested, and there is some transition guidance, the same products can continue to serve retail savers while delivering far less of the distortion that has been documented.