After generative artificial intelligence (GenAI) took the tech world by storm, reality has set in. Can you monetise GenAI? And if so, how exactly?
There are plenty of GenAI enthusiasts, however, who believe the technology will transform the way we work and live. Indian companies are gearing up to meet the challenge. According to an Intel-IDC report, spending on AI, still small in India, is expected to treble to $5 billion (Rs 42,000 crore) by 2027. By then at least three semiconductor and chip making plants will be up and running in India.
The Ola group whose electric vehicle (EV) firm Ola Electric Mobility had a successful IPO in August 2024 is bullish about AI’s prospects in India. Ola CEO Bhavish Aggarwal says his AI startup Krutrim will design, build and launch its first AI chip by 2026.
As Aggarwal told Business Standard, “India has a unique opportunity to become a global superpower in technologies of the future such as AI. India being the source of 20 per cent of the world’s data, and home to one of the largest tech talent pools, is fully equipped to lead the AI wave.”
According to Aggarwal, over 25,000 developers have used Krutrim Cloud so far. More than 10 million people have used Krutrim’s Chat app. Around 250 billion API (application programming interface) calls were made on Krutrim and over one trillion tokens generated since the startup’s launch in early 2024.
India has made a late start in AI technology. While Nvidia and other chipmakers are building chips of 3nm (nanometers), India’s first home-grown chips will be a more modest 28nm. But such chips are widely used in the automotive sector where ultra-thin chips are not necessary.
Most experts believe India should focus on chip design and building an assembly ecosystem for semiconductors rather than go straightaway to fabs (fabrication plants for semiconductor devices). These are capital intensive and can cost up to $4 billion (Rs 33,200 crore) to build.
The world’s largest semiconductor fab is TSMC (Taiwan Semiconductor Manufacturing Company). It was established in 1987 and has steadily built up its global leadership. It commands 62 per cent of the world’s chip market. Its annual revenue is over $75 billion (Rs 6.30 lakh crore).
TSMC’s advanced 3nm and 5nm chips account for 46 per cent of its total wafer production. Competition though is emerging from South Korea and China. Samsung Foundry is the world’s second largest semiconductor manufacturer. However, it has only one-sixth TSMC’s annual revenue but is innovating rapidly with even thinner 2nm chips.
China’s SMIC (Semiconductor Manufacturing International Corporation), the world’s third largest chipmaker, commands just 5.7 per cent of the global market, underlining TMSC’s domination. Only one Western chipmaker, Global Foundries, based in California’s Santa Clara, is among the world’s top 10 semiconductor manufacturers. The other nine are Taiwanese (four), Chinese (three), Korean (one) and Israeli (one).
While India’s private sector plays catch-up in a technology that’s critical to AI-powered smartphones, EVs, microelectronics and data centres, companies like the Hiranandani group’s Yotta Infrastructure are building an ecosystem with GPU (graphic processor units)-based computing capacity for use in advanced visual-backed AI devices.
*Tool-kit
AI is a tool that can be put to a variety of uses. For example, as Saakar Yadav, founder of Lexlegis.ai, told The Times of India: “We are first building a world-class language model, which will accelerate research and expedite justice by transforming weeks of work into 30 seconds or less, and then we intend to use AI to cover the entire legal workflow for all stakeholders.”
India’s legal system is clogged with over 50 million cases. Every year several million cases are added to the backlog. Clearance levels are low. AI can speed up case resolution by cutting through voluminous paperwork.
AI can also transform Indian education and healthcare. Vivek Raghavan, co-founder of Sarvam AI, says: “We believe India will interact with GenAI through voice in our own languages. Our goal is to help a billion Indians use GenAI to do something useful in their lives. Just as UPI made digital payments easier in India than anywhere else, we want to make people’s lives simpler through GenAI. These technologies have the potential to revolutionise access to quality education and healthcare for every Indian. That is our vision and motivation.”
Raghavan’s Sarvam AI co-founder Pratyush Kumar adds: “We have built state-of-the-art models for detecting the language, recognising the speech, speech synthesis, generating the text, calling functions, for guardrails. We have put together these models on a platform that is super-efficient, spent a lot of engineering hours building good abstractions, low latency, omni-channel pipelines.”
India is thus adopting a two-track policy in AI. First, use GenAI in sectors where its application will generate tangible benefits – from electronics to the judiciary, education and healthcare. Second, build a semiconductor ecosystem led by design and assembly infrastructure, followed by fabs with 28nm chipmaking capacity, gradually coming down to 5nm and 3nm wafers which are used for ultra-sophisticated devices, including microelectronics and EVs.
Elon Musk who co-founded OpenAI with Sam Altman and others but left over differences on strategy is moving forward with his own AI innovations. He recently released Grok-2 and Grok-2 mini, two large language models (LLMs). Musk says Grok-2 has performance levels competitive with advanced models such as GPT-4o from Altman’s OpenAI.
India missed the first AI bus. It will take years to catch up with levels already attained in Taiwan, China, Korea and parts of the West. But India’s large pool of software engineers and tech entrepreneurs are an asset few other countries possess.
In the archetypal race between the tortoise and the hare, it’s who finishes first, not who started first, that counts.