Vishnu Vardhan, the founder and CEO of SML and Vizzhy, emphasised on the huge investment required to build and scale complex AI models. In an exclusive interview with AIM, the creator of Hanooman disclosed his plans of raising $200-300 million for the same.
“That’s the kind of money we need to launch this [Hanooman] kind of a product. We’ve already spent tens of million dollars, but that won’t work,” he said.
Vardhan rued that most Indian investors are “not ready to spend on research and deep tech startups”. He noted that Indian investors are willing to commit to minuscule amounts of investments for AI research vs the large amount of 100s of crores required.
Investment for India AI
Recently, the Indian government approved $1.25 billion investments for AI projects, that includes developing LLMs and computing infrastructure. The money would also be used to fund AI startups and accelerate AI applications in the public sector.
While this may be a recent announcement, the reality of Indian investors is starkly different in the current AI landscape. It is largely believed that many Indian VCs don’t have a thesis on deep-tech investments.
If you look at some of the prominent investments in Indian AI startups over the past few months, you’ll find that only a handful of companies have received significant funding. India’s AI ecosystem has raised $700 million in the last three years, where over $500 million in funding was raised by 24 generative AI startups in 2023.
Ola’s Krutrim, touted as India’s own AI company, received $50 million for a valuation of $1 billion, making the company India’s first AI unicorn. However, major funding came in from Matrix Partners India, one of Ola’s primary investors.
Sarvam AI, a Bengaluru-based AI startup, which is training and building Indic-LLMs, received $41 million funded by VC firms Peak XV and Khosla Ventures. The company is also receiving backing from big tech biggie Microsoft.
Indian businessman and former board member of Infosys, Mohandas Pai, highlighted the abundance of talent but not funding when it comes to AI investments.
Source: X
Founder and investor Gaurav Sharma pointed out that with a $4 trillion economy, India should be ‘bold’ and allocate a $30-40 billion sovereign AI investment plan for the next 2-3 years. He even emphasises on how India needs a comprehensive AI policy like Japan.
While India struggles in raising AI investments to build complex models of GPT-4 or a GPT-5 prowess, US cruises through the challenge.
When Google DeepMind chief Demis Hassabis was asked about the OpenAI- Microsoft investment of $100 billion to build a supercomputer, he simply replied with, “We don’t talk about our specific numbers, but I think we’re investing more than that over time”.
Hassabis even spoke about how Alphabet Inc has better computing power than its competitors, including Microsoft, and named it one of the reasons for them to tie up with Google. “We knew that, in order to get to AGI, we would need a lot of compute,” he said.
Fathoming a $100 billion investment or even more may seem incomprehensible. However, the reality showcases the massive scale of investments required to build powerful AI models.
US Encourages Investments
Last year, when OpenAI chief Sam Altman visited India, he was criticised for saying that a model similar to ChatGPT cannot be built with $10 million. While many took to challenging that notion, one year later, the truth remains that huge investments are a must. This is something that the big-tech companies in the US have remained unaffected by.
According to the Stanford AI Index Report 2024, the training costs for GPT-4 touched $78M, with Google Gemini Ultra reaching $191.4M. Interestingly, in 2017, Google’s Transformer model cost $930 for training.
The report continues to also show how there is a direct correlation between the training cost and computational requirements.
Source: AI Index Report 2024
As per the report, the funding for generative AI increased eightfold from 2022 to $25.2 billion with prominent AI players such as OpenAI, Anthropic, Hugging Face and Inflection receiving significant fundings.
Emerging startups in the US also received huge fundings. The AI-powered answer engine, Perplexity received a massive backing from NVIDIA, Jeff Bezos, and others. The company received $70M in the last round, taking its valuation to $520M.
Additionally, the US government also supports funding. In 2023, the government allocated a total of $1.8 billion for AI research and development. For 2024, $1.9 billion has been requested.
While it may seem like there is no dearth of investments for AI companies in the US, a stark difference in the pattern of AI investments in India remains.
The post India will Need at least $200-300 Mn to Build GPT-5-level AI Model appeared first on Analytics India Magazine.
