#243 Hope as a Strategy

Analysing the Interim Budget, Fintech Troubles Yet Again, and the Folly of Buying GPUs-on-cloud with Taxpayer Money

India Policy Watch #1: Getting The Budget

Policy Issues Relevant to India

— RSJ

The interim Union Budget 2025 was presented this week, and, as was expected, it was a tame affair. Elections are less than a hundred days away, and there’s no real need for this government to make a populist gesture in the economic sphere to draw votes. Therefore, all that was of interest was how the key metrics panned out and whether there were any inferences to draw from them about the performance of this government on economic management. I mean, one can get caught up in all those outlays and allocation numbers while discussing budget and, possibly, miss the big picture. Broadly, I drew four conclusions from the speech and the numbers.

Firstly, as I had written in the forecast edition at the start of the year, this government likes to take fiscal consolidation seriously. The big positive surprise was that the union fiscal deficit for the year came in at 5.8%, a full 10 basis points below the best estimate of 5.9 per cent. What’s more surprising was that the projected deficit for FY25 was 5.1 percent which is significantly below the 5.5 per cent expectation. The obvious upshot of this is that the FRBM (fiscal responsibility and budget management) target of 4.5 per cent by FY26 promised by the government seems to be in sight. In absolute terms, what this means is that the deficit has remained at the same level for four years now. Also, the FY25 nominal GDP growth estimate of 10.5 per cent (with a 4 per cent GDP deflator) is somewhat conservative given what we have seen in this year. Even the estimate for the current quarter on revenues and expenditures is conservative. The trend that the current FM had begun a few years back of not using aggressive assumptions to ‘manage’ metrics has solidified now. This is good, and I hope this will make it difficult for any future incumbent to go back to the good ‘ol days of playing with numbers. The continued prudence in managing the finances, a good contrast from other large economies at this moment, is a strong signal to the market and investors. After all, only a couple of months back, there was a minor kerfuffle when the IMF raised concerns about India’s elevated public debt levels. So, this will help with the sovereign yields and also with the image of the government with rating agencies who are looking for stronger commitment to the path of fiscal consolidation. 

Secondly, it is useful to parse the numbers a bit on the revenues side. Income tax as a percentage of GDP is now at 3.5 per cent, which is almost double where we were about a decade back. While this is way lower than the 8-10 per cent benchmark that most developed countries have, the continuous rising trend on this metric is a good indicator of the quality of revenue. Income taxes are progressive (the poor get taxed less, the rich more), and the tax buoyancy seen in the last few years suggests better compliance because of greater digitisation of the tax payment system and the widening of the tax base. Also, the aggregate corporate tax collections have been growing consistently for a couple of years after seeing through the impact of tax cuts and COVID-19. GST collections, which were rising somewhat disproportionately compared to GDP a few years back, are now growing in line with GDP growth. The collections as a percentage of GDP is settling at about 6.7 per cent which seems reasonable as a trend to project into the future. The one area that hasn’t worked out again is the expected revenues from divestment. This needs better execution, and it should be an area of focus after the elections. The government still runs way too many business enterprises where greater value can be unlocked by bringing in private capital and expertise. On the flip side, the government holdings in listed companies have increased in value because of a strong equity market performance. This provides a window for raising revenues for a capex push, if necessary, after the elections. 

Thirdly, on the expenditure side, there is a problem. This has grown faster than the rate at which tax as a percentage of GDP has grown. Some of this is because of a sustained capex increase, which has gone up to about 3.3 per cent of GDP. I cannot grudge this because almost 62 per cent of capex is going to roads, railways and defence. This is good quality infrastructure spend whose multiplier effects will come through over the years. The problem is still in the other expenses where the size of the Indian state continues to grow. The debt-to-GDP ratio is still high, and the interest payment burden at 3.7 percent to GDP has to be brought down. The improvement in the quality of revenues should give the government confidence to focus on the primary deficit and bring it down through a focus on transmission efficiencies and a focused reduction in subsidies.

Lastly, the broader message I took away from the speech and the subsequent discussions is that as much as this government would like to project an image of an India that deals with the world on its own terms, it remains sensitive to its image of being fiscally conservative and reform-oriented in the eyes of its global stakeholders. This is a pragmatic approach because we shouldn’t start believing in our own hype. There is still a long way to go in terms of how our economy is seen beyond just its size. In the absence of a real political opposition that can hold the government accountable, it is somewhat reassuring to see that the pressure of being seen as responsible by global fund managers and rating agencies keeps it on the straight and narrow. Small solace for those who like democratic tug and push of debates. But solace nevertheless. 

---

Addendum

— Pranay Kotasthane

Two other things of note about the budget.

One, the projected collection under cesses and surcharges is set to increase further. This isn’t good.

First, a quick recap of these two concepts. A cess is tied to an earmarked purpose. The money raised through a tax is held separately to fund a project for which the cess is meant. Surcharges are merely taxes on taxes. They are not raised for an earmarked purpose.

Both are evil because, through some legal chicanery, the money raised by the Union government through cesses and surcharges doesn’t need to be shared with other levels of government (except the GST Compensation Cess). Essentially, cesses and surcharges are anti-federal.

Aap chronology samajhiye. Vertical devolution refers to the division of taxes collected by the Union government between itself and all states taken together. The denominator the Finance Commission uses for vertical devolution is the divisible pool of resources. It’s not the same as the gross tax collections of the Union government. So, when you hear that the 15th Finance Commission (FC) recommended that the Union devolve 41% of its funds to the states, the denominator being used is the divisible pool, not the gross tax collected.

Now, here’s the relationship between the two. The divisible pool of resources = Gross Tax Collected - Cost of Tax Collection - Union Cesses and Surcharges. Just as the 14th FC increased states' share from 32% to 42% of the divisible pool, the Union government subverted this recommendation by levying new cesses and surcharges. So, while the Union government nominally accepted the FC recommendations, its subversion meant that the states didn’t see their pot of funds grow, while the citizens were burdened with additional cesses and surcharges.

In this year’s budget, the money that will be carried off from the divisible pool amounts to almost 10 per cent of the Gross Tax Collections. That’s a total of 3.9 lakh crore, a tad more than the entire expenditure of the Tamil Nadu state government, for instance, in FY24.

Despite this pilferage, state governments posing to champion federalism keep squabbling over the horizontal devolution formula, i.e., the formula that decides the division of the states’ share amongst all state's themselves. If the pie is smaller, there’s no point in arguing which state has the bigger piece.

That’s why I’m hoping one state government will constitutionally challenge this proliferation of cesses and surcharges in the Supreme Court as violating federalism, a core element of the Basic Structure of the Constitution.

The 15th FC had commissioned an excellent study explaining the legal and constitutional problems with cesses and surcharges. However, since the Union government appoints the FC, it’s unlikely to challenge the imposition of these distortionary taxes. Nevertheless, I hope the newly constituted 16th FC proposes ways to subsume cesses and surcharges in the divisible pool. In any case, the abolition of cesses and surcharges should be a major point around which opposition states should mobilise. Instead, they will fall back to the low-level equilibrium of gaming the horizontal devolution formula.

Two, the government plans to incentivise private sector R&D investment through a corpus of Rs 1 lakh crore, which will be used to provide long-term loans at low interest rates to companies. At face value, the focus on private R&D is great. As we have discussed in these pages, India’s private sector in-house R&D spending is pathetic (0.25% of GDP) compared to the world average (1.4% of GDP). The government R&D expenditure normalised for India’s income level is not that bad actually.

Thus, pushing the private sector to spend more on R&D is necessary to sustain India’s economic growth. The recent increases in the human power capacity of India’s patent offices suggest that India is also beginning to address other segments of the R&D supply chain.

Nevertheless, good ideas can only get us thus far. We often say that “good policy; bad implementation” is a myth. Any good policy takes implementation into account. As for this announcement, there are no implementation details yet. Last year, we had good things to say about the National Research Foundation Bill, another important development in this field. But the NRF is yet to see the light of the day.

Arun Shourie once characterised the slow pace of reforms using this memorable line: platon ki awaaz toh aati hai, par khaana nahi aa raha (one can hear plates clattering, but there is no food to be had). Let’s hope that in R&D spending, the government is determined to prove this characterisation wrong.

---

India Policy Watch #2: Paytm; Die Another Day?

Policy issues relevant to India

— RSJ

The big news in corporate India this week was that the banking regulator pretty much shut down the Paytm payment bank, a subsidiary of one of India’s largest fintechs that was once a poster boy of digital India. Here’s the Economic Times reporting on it:

“In an order issued on January 31, RBI asked Paytm Payments Bank to stop all basic payment services through various platforms and technology railroads, including the Unified Payments Interface (UPI), Immediate Payment Service (IMPS), Aadhaar-Enabled Payment System (AEPS) as well as bill payment transactions, with effect from February 29.

RBI also found KYC checks for hundreds of thousands of customers were missing, and some of the accounts were either owned by individuals with past issues with enforcement agencies or had abnormal balances, amounting to crores of rupees in some cases. The central bank has flagged multiple instances of a single permanent account number being used to open more than 1,000 accounts.

Security agencies are looking at the possibility of the entity being used as a front to launder money.

In March 2022, the regulator had barred the Noida-based company from onboarding new customers and directed it to appoint an external auditor. In 2022, an RBI inspection found that many of the compliance reports that the payments bank was submitting were not in line with requirements and, in some cases, there were outright false claims. The banking regulator also allegedly found violations of the Prevention of Money Laundering Act, which led to the bank being barred from accepting fresh deposits, said the person.”

Regulatory overreach?

Maybe so but the sequence of events suggest there were sufficient meetings and warnings given to the entity to mend its ways. I have written about fintechs and disruption in financial services in India in a few previous editions. There are three points to highlight in this context.

The real disruption in financial services in India has been in payments through UPI. The credit for it should go to the state - the UPA 2, which conceived of and implemented Aadhar, and the subsequent NDA governments that have used it as a foundation to build digital public infrastructure that anyone can ride on easily. Every fintech has used it and built journeys, products and services that are largely imitation of similar apps in the West or China. Paytm, which had big Chinese fintechs and private equity funds as its anchor investors, largely picked up stuff directly from there including BNPL, embedded finance with insurance, travel and mall built in and attempts to become a superapp. Once the UPI infrastructure came in, it levelled the playing field for everyone and increased competition further. And with MDR driven to zero, the payments business was good for scale, but you could make almost no money out of it. In a way, the state disrupted the fintechs in this model.

As the madness of easy money that chased random vanity metrics like GMV, number of users, et al. abated, these fintechs had to find ways to show a path to profitability to continue to draw external funding. So, they entered into lending. Now, the core principle of lending business is not how easy you make it to give away money to likely borrowers. Instead, it is about how you build a model so that you get your money back from your borrowers. It is all about managing risk - in choosing your borrowers, assessing their creditworthiness, giving them loans that they can afford to pay back and running a tight collections unit. However, if your incentive is to show how quickly you have built up a Rs. 1000 crore book or a Rs. 10,000 crore book or you have hit a run rate of Rs. 100 crores of monthly disbursement, then you have got the whole model wrong.

But that’s what has been happening in India in the past year or so as these fintechs with willing banks or NBFCs supporting them with capital, have gone and created what seems like an asset bubble in the small ticket unsecured personal loan market. That they have done so by claiming they are disrupting traditional banks with more intuitive customer journeys, frictionless processes and intelligent use of data needs to be taken with huge loads of salt. You don’t really need any technology if you want to give away loans without following good KYC processes, as it is alleged in the news report. You can just put a banner outside your office that you are willing to give loans without any serious documentation and you will soon have people traveling miles to avail of your wonderful service. You don’t need any tech or app for it. You will draw exactly the kind of customers that everyone else who uses real metrics for the lending business has rejected. You can rejoice at how well your business is growing, but you are merely riding a tiger. 

Now, you might say. that if they are riding a tiger,; the market will figure that out and punish them. Why should the regulator get involved with a sledgehammer? The problem is you would like to believe there’s information symmetry between a listed company and the market. That is everything the company knows; the market will eventually know because of the disclosure norms set out by the market regulator. Well, there are two problems with that. It is correct in theory, but in practice, as we have seen over the years, the management team will always hold back something either because it believes it can manage the problem or because it doesn’t want to be transparent. So, there’s always quite a bit that the management team knows that can remain suppressed till it becomes so large that it blows up. Also the other issue is time. The information between the market and the listed entity becomes symmetrical, but it can take time. Unfortunately, there’s very little information asymmetry with the regulator, like an RBI, who in India can ask for every minute detail of the business and the associated risks. And once the regulator has that information, what can it do? It can wait till the market punishes the entity. It has to act. 

It will be interesting to see how the parent entity of Paytm payment bank comes out of this. Investor confidence is shot, its reputation is down, and no amount of spin can take away from the concerns any bank, NBFC, merchant, or customer will have in dealing with it. It will be a long haul ahead. In a way, with what’s happened to Byju’s, Paytm, PharmEasy and a whole bunch of other startups that have found their valuation collapse on the back of a flawed business model, it is time for the Indian startup ecosystem to check their premises on what constitutes disruption and a successful business built on it. 

---

PolicyWTF: AI as a PSU

This section looks at egregious public policies. Policies that make you go: WTF, Did that really happen?

— Pranay Kotasthane

No, I’m not talking about Air India’s re-nationalisation. Check this report from the Indian Express:

“The Centre’s ambitious artificial intelligence (AI) Mission may soon head for Cabinet approval and could have an outlay of more than Rs 10,000 crore, Union Minister of State for Electronics and IT Rajeev Chandrasekhar told reporters Tuesday.

As part of the programme, the government wants to develop its own ‘sovereign AI’, build computational capacity in the country, and offer compute-as-a-service to India’s startups…

The Indian Express had earlier reported the capacity building will be done both within the government and through a public-private partnership model, highlighting New Delhi’s intention to reap dividends of the impending AI boom which it envisions will be a crucial economic driver.

In total, the country is looking to build a compute capacity of anywhere between 10,000 GPUs (graphic processing units) and 30,000 GPUs under the PPP model, and an additional 1,000-2,000 GPUs through the PSU Centre for Development of Advanced Computing (C-DAC), Chandrasekhar had earlier told this paper.”

The stated policy aim is to create a GPU computing facility in India that start-ups can access over the cloud at a lower cost—a noble goal. But is a government-sponsored cloud GPU the right way to go ahead? I doubt it.

There's no market failure at the level of cloud service providers. Even retail cloud service providers such as DigitalOcean and Vultr already offer GPUs on the cloud. And yet, the costs of using these GPUs are really high. That’s because of an underlying problem—the current Nvidia GPU chip production rate cannot catch up with the demand. Investing Rs 10000 crore in a cloud service provider (CSP) will not solve this GPU shortage problem. India will still be in the same long queue, with wait times over two years.

By then, both these two things would have happened. First, there will already be many more GPU-on-cloud-for-hire players in the market, and the costs of renting one will further go down. Secondly, we might see better architectures replacing GPUs for handling cutting-edge AI loads, such as Google’s proprietary challenger (TPUs), new customised chips (ASICs), or reconfigurable hardware (FPGAs).

So India will be locked in with some costly GPUs three years later in the best case, while the world could've moved on to other architectures.

The real problem is with the market power of Nvidia in GPUs. So, if the government does have Rs 10,000 crore to allocate for AI infrastructure, it should consider financing companies researching to identify next-gen architectures for AI loads. Not just on AI chips but also on the AI software and firmware toolchain that drives the adoption of a particular hardware. This isn’t easy, but the knowledge spillovers from such a project will be significant as it plays into India’s comparative advantage in chip design and software.

Building a government-backed GPU cluster will not give India any strategic advantage. Leave it to the market players. Don’t fall into the PSU mindset while thinking about this AI.

---

HomeWork

Reading and listening recommendations on public policy matters

1. [Podcast] Over at Puliyabaazi, listen to the world’s foremost scholar of the Pakistani military explain that Fauj’s military empire. It’s a treat, don’t miss.

2. [Win] Meanwhile, the government has reduced import tariffs on mobile phone components. As we often repeat here, there are no competitive exports without imports in electronics manufacturing.

---