Indias Quest For A World Class AI Engine
Indias AI Aspirations and the Global Race
India, with its vast pool of over 5 million IT professionals and an increasing educational emphasis on artificial intelligence (AI), appears perfectly poised for what is shaping up to be a global AI technological race. However, despite the US establishing a benchmark with ChatGPT in 2023 and China rapidly catching up with its potent DeepSeek chatbot, India is still working towards developing a comparable Large Language Model (LLM) capable of human-like communication.
This isn't due to a shortage of ambition. According to data from market intelligence firm Tracxn, the Indian AI sector includes 7,114 startups that have collectively raised $23 billion (€20.15 billion) in equity funding so far. Last year, Prime Minister Narendra Modi's Cabinet approved the IndiaAI Mission initiative with a budget of nearly $1.21 billion, aiming to "undertake the development and deployment of indigenous Large Multimodal Models (LMMs) and domain-specific foundational models in critical sectors."
DeepSeek Cheap AI model shakes up tech industry
This week, IndiaAI Mission CEO Abhishek Singh said Indian startups need to think beyond their home turf to compete and succeed against global giants.
"They will have to ultimately compete with the best in the world," Singh said at the Accel AI Summit in Bengaluru. "Initial levels of support may come from the government, but that will not sustain them in the long run."
"They will have to have a global vision in mind when they are training models," Singh said.
AI Development Needs Industry Government and Academia
Representatives of the National Association of Software and Service Companies (NASSCOM), the voice of India's $283 billion tech industry, point out that building a globally recognized AI model is a complex, resource-intensive process.
"The argument is not whether India can catch up, but whether we can move fast enough and define an AI identity on our own," Satyaki Maitra, senior manager communications of NASSCOM, told DW.
Last week, IndiaAI Mission announced the addition of 15,916 Graphics Processing Units (GPUs), which are essential for AI research due to their capacity to perform calculations in parallel. The latest boost will bring the total national AI computing capacity to 34,333 GPUs through public-private partnerships.
Chipmaker NVIDIA profits up despite trade war challenges
Startups such as Gan AI, Gnan AI, SarvamAI and Soket AI, supported by the IndiaAI Mission, are building foundational models tailored to India, while firms like Sarvam AI, Fractal and CoRover AI are focusing on AI innovation.
"However, AI success cannot be achieved through isolated innovation," said Maitra. "It requires cohesive collaboration between government, industry, and academia to build the full value chain, from computer and data governance to model training and real-world deployment."
What is Holding Indian AI Back
Pawan Duggal, the country's foremost cybersecurity expert, told DW that India is likely to face a shortage of high-end AI hardware, limited access to advanced GPUs, and insufficient cloud computing resources, which are essential for training large-scale AI models.
"There is also a significant investment shortfall when compared to global peers. While venture capital investment in Indian AI startups has increased, it remains a fraction of what is seen in the US or China," said Duggal.
"The US invested $2.34 trillion and China $832 billion in ventures and startups from 2014 to 2023, while India invested $145 billion in the same period," he added.
Can AI blackmail us Facts vs fear
Duggal believes that India is already moving towards creating its own AI model, but it has yet to address significant challenges including infrastructure, funding, talent, data, and regulation.
Indias Talent Pool and Data Hurdles
Another issue facing Indian engineers is the diversity of languages in India, with English being just one of 22 official languages in the world's most populous country. Furthermore, official languages make up only a tiny fraction of over 1,600 languages spoken within its borders.
"The only use-case of an 'Indian' LLM is if it works in our various languages, which is difficult as of now given that there is a lack of quality data for LLMs to train on in most Indian languages," Yash Shah of Momentum 91, a leading custom software development company, told DW.
"For an LLM in English, there are other companies and countries which are far ahead of us and would continue to be that way," said Shah.
Godfather of AI urges tougher government rules
However, Utpal Vaishnav of Upsquare Technologies, a global technology holding company, says the real obstacles are actually "risk-shy investors, patchy data rules, and tight GPU supply."
"Brains are plenty in India. GPUs are on the way and our multilingual data is waiting to be shaped. Give this talent, patient capital and clear problems and a compact, world-class LLM can be launched in two, three years," Vaishnav told DW.
