India Policy Watch: The Long Trek To SummitInsights on current policy issues in India—RSJThe India AI summit concluded yesterday with the government machinery making the usual claims associated with such events: PM Modi’s vision for responsible AI, Viksit Bharat by 2047, investment pledges totalling up to $250bn (some $20bn in deep tech ventures alone, apparently) and the usual self-congratulatory stuff about India leading the global south, etc. During the week, the AI summit gave us all the masala that we have come to expect from a Bollywood blockbuster. Song and dance, comedy (more on it later), action, dialoguebaazi, tragedy and hope. The only element missing was farce which was duly supplied on the last day by youth Congress activists who demonstrated shirtless inside the venue. Cutting through the hyperbole and the usual self-serving corporate speak, of which there was plenty, the summit was helpful in getting a first-hand understanding of the state of play of AI in India and how global companies are viewing the India opportunity. The idea of an annual AI summit is sound, and it is good that India has taken an early start in occupying this space. India has the market and an open tech ecosystem to draw in all the key AI players. The prevailing thesis that India will benefit from downstream AI services that will be used by various industries can only be tested and augmented by proving it through a platform of this kind. If the likes of TCS and Infosys believe that the foundation models will take time to be adopted in large enterprises because of legacy platforms and data inconsistencies, and this gap offers tremendous engineering services opportunities for players like them, then a platform like an annual AI summit is the best place to prove it to the world. The same argument applies to AI native startups who are building on foundational models to solve real-world problems in healthcare, education, governance or financial inclusion. Any solution using AI at scale for India will have its application in similar markets (‘global south’) that might have problems of access and talent. Again, a globally branded AI event will help these startups find newer markets faster. The way I think of this is like one of those large consumer expos in China where companies from all over the world converge and find there is always a Chinese company that exists that can offer you a world class product (everything from cosmetics to home furnishing to machine tools or suitcases) and paraphernalia (packaging, shipping, billing etc) at unbelievably low costs. This is how hundreds of D2C brands in India have taken off in the past 3-4 years. There is some design innovation or branding insight from the Indian promoter to get the product going, but everything else gets done from China (in most cases). In a similar vein, I can imagine a future India AI summit where companies from world over come to find there is always an Indian vendor who has used AI already to build a world-class product or a service proposition that they can just pick up from the summit and deploy in their business or their country without having to reinvent the wheel themselves. Despite the many glitches in this edition, this summit is a step in the right direction, and there should be greater investment in time and effort to make it bigger next year with a fixed calendar that everyone knows well in advance. There is a general investor sentiment about India being among the basket of countries seen as “AI losers”. The best way to address this is through an event of this kind that could provide a systemic response on economic gains to be made through AI adoption and being positioned as champions in helping others adopt AI. I spent a fair amount of time reading through the interviews and content of panel discussions that featured in the extensive media coverage of the event. As luck would have it, I could meet a few of the participants outside of the event during the week to get a first-hand sense of their views on AI and India’s position in it. There are five conclusions I drew from these about the summit specifically and the AI space more broadly in India. One, the government should hand over the organisation of the summit to an industry body and restrict itself to the role of enabler and promoter of the event globally. Almost everything that was cringe about the event was on account of the government running the event. There is already the excellent example of the Global Fintech Fest (GFF), managed by NPCI, which takes place annually in Mumbai. It has gotten bigger, better and smoother every year with global players, local fintechs and governments from around the world participating in an area (fintech esp payment and lending) where India is seen as a world-class innovator. In fact, other nations are now copying the GFF playbook (Singapore is the most recent) to showcase their fintech capabilities. The desire for optics over substance, cronyism in selecting who will set up pavilions, lack of quality control at every step and absolute thoughtlessness about the venue facilities and logistics can easily be avoided if we just follow the GFF example. Two, like almost everyone else, I was also impressed with a lot of homegrown AI innovations showcased at the summit, including, of course, Sarvam.ai. Ignore the hyperbole of Sarvam doing with 40 researchers what DeepSeek could do with 5000 (there is really no comparison) and the jingoism around a sovereign foundation model, which has no meaning for where India is at this stage in AI evolution. Or even the impressive real-time local language translation capability that the platform demonstrated. The thing that was noteworthy to me was the clarity among its founders about building the core model using local data, algorithms and engineering for enterprise use cases and B2B applications and not going beyond it for now. It isn’t trying to be a free chatbot and compete with global LLMs, as many in India are hoping it to do. This approach of bounded sovereignty that helps Indian enterprises to not be dependent on global LLMs while benefitting from products built on it that are customised for local markets will have resonance in many other emerging economies. Three, there are other infrastructure constraints that will come in the way, which will require long-term policy support and enablement. The announcements about significant investments in data centres by global and domestic players need to be tempered with an understanding of what it takes to run such large data centres. Reliable, cheap, and clean energy is already in short supply in India. The future power requirements to support AI ambitions and commitment must be mapped, and more specific plans to address them have to be put in place. The private power and energy sector is increasingly an oligopoly, and that won’t help when the demand shoots through the roof. Similarly, talent will become a bottleneck soon if the long term vision of dominating AI downstream services isn’t articulated and communicated to the university system. The usual assumption that in a free market, the demand and price signal for this talent will eventually solve for supply might not work or work with a lag. I have visited the AI labs (or equivalent “futuretech” labs) in at least three different modern, private universities of variable repute in the past four months. Every single one of them had that robodog that has global fame now, thanks to the shenanigans of Galgotias University. But beyond students playing with such toys (drones, robodogs, etc.), there is no real and rigorous AI training happening that inspires confidence. Surprisingly, in a good way, each of these universities had some good quality applied engineering labs for students to use their hands in creating a tangible product solving a real-world problem. In any case, there is a need to address such constraints more proactively to support all those ambitious talk and investment pledges. Four, there is a need to build a wider consensus among the political class on the risks and opportunities of AI, job losses and creation challenges and a common willingness to support this vision through future government changes and election cycles. I’d put it at the same level as defence or foreign policy in having consensus across the political spectrum on what we should defend, where we should invest and how we should promote this core capability. I suspect this government won’t have much interest in doing so on its own, but those advising it should suggest the importance of this. Lastly, if India wants to position itself as the champion of AI services (legacy modernisation, automation, agentic platform building, etc.), it has to embrace LLMs from Chinese tech giants too. After all, it is a two-horse race on the upstream capabilities, and India should be prepared to work off any of these models regardless of their provenance. In a similar vein, India should view the Chinese markets as a big opportunity for its system integrators (IT services giants) to innovate on AI services. This means India should actively court China as a market and be open to the use and learning of Chinese LLMs within India. Whatever sovereign or data risks we might think we have with Chinese LLMs are equally true for the U.S. models in the Trump (and post-Trump) world. India needs to be indifferent to these. On balance, there were way more positives emerging from the India AI summit. While my view on medium-term risks for India's IT services model was further strengthened from what I saw or heard at that event, I was also encouraged to see the alternative opportunities that might partly replace them. These will only get stronger if we continue the momentum in creating the market for ourselves. Global Policy Watch #1: Tariff TantrumsGlobal issues relevant to India—Pranay KotasthaneWhat a week it has been for the US reciprocal tariff regime! The Supreme Court upheld the cases filed by two small firms and struck down the Liberation Day Tariffs. The case conclusively argued that tariffs are a form of tax on Americans. And no taxation can happen without the approval of the representatives of the people, i.e., the legislature. That’s because it’s the body where representation is most granular. The American President represents every American citizen, which paradoxically means the President represents no one in particular. This is a weaker form of representation for tax purposes, because taxation creates winners and losers at a local level. Hence, the mandate is that there can be no taxes without deliberation and approval of the legislature. Further, the complainants were able to prove that, while the President has the power to reduce imports, it does not give him the power to impose taxes, i.e., tariffs, to achieve this goal. This is a fantastic case of US institutions fighting back against whimsical executive power. But it doesn’t mean tariffs will wither away soon. As we discussed here, Trump is convinced about the utility of tariffs and idolises the “tariff king”, President William McKinley. Most of all, tariffs are the quickest way to create the spectacle of country leaders kowtowing to the American emperor, something the President clearly enjoys. Thus, this reversal is good but unlikely to reduce global uncertainty. There are still sufficient constitutional mechanisms available to the executive to impose tariffs. Sure enough, Trump soon imposed a 10 per cent global tariff by invoking Section 122 of the Trade Act, a provision that allows the president to impose duties of up to 15 per cent for 150 days to address "large and serious" balance-of-payments issues. A day after floating this balloon, the administration further increased the tariffs to the 15 per cent upper limit. Then there are two other slower means to impose tariffs: Section 301 and Section 232. Under Section 301, the US Congress has granted the US Trade Representative (USTR) authority to investigate and respond to unfair trade practices by other countries. But it mandates a formal investigation process. This is the mechanism used for the China tariffs starting in 2018. Under Section 232, the President can impose tariffs on specific items if the Secretary of Commerce determines that imports of such goods threaten national security. This measure again requires the US Congress to determine the national security impact. In other words, while the sword has been blunted by the recent case, it still remains with the wielder. What it has definitely achieved is to pause this mercurial tariff imposition through social media. Future changes will be far slower and more deliberate. It also means that wildly different tariff rates against all countries are unlikely to sustain—it will either be a constant ‘fifteen is the new zero’ rate on all countries; on a handful of countries proven to be using unfair trade means; or on dominant exporters of goods important for US national security. Not much yet, but given the stakes and recent trends, the world will accept this gleefully. Global Policy Watch #2: Emotional TaxesGlobal issues relevant to India—Pranay KotasthaneWe have long argued that Robin Hood Taxes don’t work. We have also explained what the calls repeatedly made by well-meaning NGOs for taxing the wealthy get wrong in the Indian context. Nevertheless, these arguments keep coming back and are steadily gaining political currency in the US. The Economist has an excellent lead story arguing that even robbing the richest entrepreneurs in the world wouldn’t cause a big dent on inequality. Here are a few striking numbers:
This argument applies more so in the Indian context, where the rich are far fewer. Wealth Tax, Inheritance Tax, Gift Tax—India has tried it all before. The only long-term solution is economic growth. HomeWork
|
