The second day of the Global IndiaAI Summit 2024 witnessed an insightful session called “IndiaAI: From Seed to Scale—Empowering India’s Startup Ecosystem.” The session shed light on the vibrantly evolving startup ecosystem in India, emphasising startup financing.  

Ms. Chandra R. Srikanth, Deputy Executive Editor of Moneycontrol, moderated the session. Mr. Amitabh Kant, G20 Sherpa, Government of India, delivered the keynote address. Other panellists included Shri. Jayesh Ranjan, Special Cheif Secretary, IT&C and Industries and Commerce Department, Government of Telangana; Shri. R Ramanathan, First Former Mission Director, Atal Innovation Mission, NITI Aayog; Mr Ranjan Anandan, MD, Peak XV Partners and Surge; Mr. Mayuresh Raut, Co-Founder and Managing Partner, Seafund; Mr. Abhinav Aggarwal, Founder, Fluid AI and Mr. Abhishek Upperwal, Founder, Soket AI and Mr. Abhishek Singh, Additional Secretary, MeitY. 

In his keynote, Amitabh Kant shared India’s current state of AI: “The Government of India has committed a substantial five-year budget of 10372 crore for the IndiaAI mission. This visionary initiative is designed to fuel AI innovation, which strengthens public-private partnership.” He further claimed India’s remarkable AI journey over the past years.  

India’s AI investment 

Abhishek Singh commented on the AI investment of India, saying, “The AI investment shows the government’s intent about the way we want the AI ecosystem and deeptech ecosystem to go”. “We have almost 5,000 crore earmarked for offering more than 10,000 GPUs that are needed to back the creation of compute capacity under the mission", he added. More investments are expected from the investors, which will be added to the income pool for the AI startups. India’s core strengths are its talent, workforce, and humungous datasets. He stated that although the government was now floating tenders to get the chips, direct purchase would not be the strategy.

The plan is not to purchase chipsets and establish an infrastructure that is owned and operated by the government. Although the government will partially subsidize the cost of access to that compute, we will still prefer private actors to make the investments. In order for people who require computing power—those who are developing, training, inferencing, conducting research, or working on algorithms—to have access to it for a lot less money than they are now paying, he clarified.

“All we need is to invest in the research ecosystem, which would ensure that the researchers can build cutting-edge solutions”, Abhishek Singh claimed.  

Ranjan Anandan claimed that within Peak XV, they had over 25 AI investments in the last year. Ranjan highlighted the importance of having more AI researchers in India by bringing them back from other countries. India possesses the largest number of STEM graduates. “Since we have raw talent, we should make sure we invest in them”, says Ranjan. 

According to Mayuresh Raut, the window of opportunity is not that great, though AI has been in the industry for so long. The window of opportunity that we once had is decreasing. “India has a lot of tailwinds that we need to grab with both hands,” he added.    

AI strategies 

Jayesh Ranjan explained how the Government of Telangana has been approaching AI for years. The government has supported more than 150 startups and deployed 52 AI use cases. Additionally, Jayesh declared, “If you’re a homegrown startup and have a product or solution that is relevant to the government, I’ll guarantee that I’ll procure it literally.” The startups will be offered their first work order without any bids or tenders.  

R Ramanathan, about the Atal innovation mission, said they had two primary purposes: empowering the young generation to acquire problem-solving and innovations. The ultimate aim of the mission is to create a nation of job creators, not job seekers. Since our current educational system does not enable access to these technologies, the Atal innovation mission was formed to enable students from the school level to access technologies through tinkering labs. He also emphasised the importance of industry partnership. “Without industry partnership, we don’t understand the demand,” he said. 

Supporting the startups  

Regarding training LLMs, Abhishek Upperwal says there has been tremendous support from the government for startups. Since they are one of the beneficiaries of the C-DAC, Abhishek says the computing availability at C-DAC and affordability are less. The systems are expensive, and unless you are highly funded, you won’t be able to access such systems. He further said that he is glad to hear people discussing research aspects when building foundation models, which is of utmost importance. “If we have to be at the frontier, we have to essentially invest in research and break the barrier”, Abhishek noted. 

According to Abhinav Aggarwal, the challenges for Indian models, like the localisation of Indian use cases, where the Indic LLMs become agentic. The model thinks through more than 6-7 reasoning steps before giving the answer. The reasoning for English models is phenomenal, while with Indic models, they are still getting to the level where they are just good at basic prompting. “We worry about how they will reason in the future because most models are becoming reasoning-based. Our approach to solving this as for now is by putting the translation model in the middle”, he added. “AI is nothing, that’s when it is everything”, Abhinav commented. According to him, when we put AI in the background and try to solve a problem, and AI is one of the tools that enables us to solve that problem, startups will succeed far better.