By Dr Ashok Bardhan, Economist and Romi Mahajan, CEO ExoFusion
In late January, a fleet- footed Chinese startup sent shockwaves through the technology community. The company, DeepSeek, created an AI language query model, as powerful, effective, and useful as those of storied competitors such as OpenAI.
In response to a query, you can see the model think aloud, exploring and refining its thought process, asking questions of itself in a continuous linked chain-of-thought procedure. Of course, it is understood that there is really no conscious thinking or reasoning going on, but mimicry can be very convincing for some purposes, and many of the results can be vetted further for accuracy and fit. And they did all this with far less money, less energy resources, older generation chips and no foreign talent (Even if the cost was higher than they claimed, it is still many orders of magnitude less than Big Tech has spent.) Many AI experts in the US had deluded themselves into thinking that was impossible. Just a few days earlier Trump inaugurated a $500 bill dollar AI infrastructure initiative. The DeepSeek development upends the spending scams and funny-money funding going on in the AI sphere; a deliberate attempt to create a money-based moat around a castle in the air, to mix metaphors.
As the DeepSeek news and its implications dawned on the financial classes, more than a trillion dollars of paper wealth was wiped off the valuations of the tech companies the next trading session- the greatest single one-day loss in history, and President Trump himself was compelled to say that it is a wake-up call for America. But, it’s also a wake up-call, or rather, an inspiration for others, including the country which has all the necessary advantages to be very competitive in this field – India.
Think about all the attributes that DeepSeek harnessed and match them up against the traditional strengths of India and its now mature software sector:
1. It used a traditional fallback option of the developing world in the absence of resources, cutting edge tools and devices: a mode of thinking, a problem-solving approach and bootstrapped innovation that stresses a “make do with what you have”, or what we in India call “jugaad.”
2. Its methodological approach and algorithms, to put it simply and without going into technical details for now, involve sparse, simplified, smaller models using fewer parameters that do not require huge data centers with their gigantic energy and water guzzling requirements. The whole idea was to effectively use brain power, modified algorithms, software and engineering when hamstrung by older, legacy chips and hardware. Outrageous financial resources, vast fallow land for gigantic data centers, diversion of energy from other uses… None of these was needed. These resources are in short supply in India as well, and our scientists, engineers are proficient at doing more with less.
3. DeepSeek staff do not have fancy degrees from expensive varsities. Practically all of the creative team members are homegrown experts. The vast reserves of software talent and engineering graduates in India, although they may not be in the same league in terms of numbers and degree levels as in China or the US, provides the single largest vital resource for developing AI related models, algorithms, and software. Moreover, a great deal of human talent is needed to curate, fact-check and ensure accuracy- another win for India. With an abundance of an extremely agile and adaptive workforce, what we have is an embarrassment of talent, if not of riches.
4. The most intriguing aspect of DeepSeek CEO’s hiring philosophy is that their engineering teams also included language, literature, and humanities graduates to help refine the company’s AI. through their distinctive creative inputs. It is ironic that Deepseek’s large language models are also optimized for English, and not just Chinese! One would think that India, with perhaps the largest pool of English speakers as well as expertise in, and institutional compatibility with, English language dependent skill sets in a wide range of subject matters, would be able to reap significant competitive benefits with its own language related AI models!
5. There are also opportunities arising out of economies of scope and innovation spillovers with other software enabled services sectors where India is competitive – back-office services, data analytics, design and R&D, legal/accounting/insurance/finance and other already well-established offshored beachheads in India. These services have moved up the value chain in the last two decades and provide a ripe milieu of innovation for AI.
The Indian technology sector has comfortably coasted along as an offshore appendage, dependent mostly on demand from the United States, with, of course, notable exceptions of some dramatically successful startups/companies catering to the domestic market and some striking successes in innovative products and services. But it is time to break this mold. The critical mass of resources, assets, competitive advantages have been there for a while, and the indigenous tech sector has the raw capability to be one of the lead players in this field; as stressed above, what is needed are not huge financial resources, but a collective commitment to snap out of this dependent daze, easy offshored pickings, and crisis of confidence.
These large, summarized response language models and platforms are in fact a low hanging fruit. There are many others that the Indian tech sector can concentrate on – for example, domain-specific AI application models for healthcare, energy, supply chains, agriculture, logistics, finance, and integrated, multi-source data analytics among others, as well as many other applications that can be of broader benefit to the economy. In many of these, the Indian competitive edge remains strong, including in some of the next breakthrough stages of the LLMs themselves – whether causal reasoning and “understanding” models; focused, personalized human-AI collaboration platforms; and integrated solutions combining different types of AI together with LLMs; many of these developments would require similar skill sets and resources.
There is no need to cede the AI game to Western Big Tech or to China’s burgeoning AI sector. There is much to be gained from India assuming a leadership position in this space.
