High compute costs, value-conscious users to challenge Indian consumer AI startups: Lightspeed's Harsha Kumar

India's consumer artificial intelligence-focused companies in segments such as education, content and healthcare will have to navigate increased compute costs, value consciousness from customers and changing retention metrics if they want to scale up and monetize, according to a top executive at global venture capital firm Lightspeed India. “We are sort of back in a way to 12 years ago, you know, when it was hard to monetize India,” said Harsha Kumar, a partner at the firm. “We're going to go through that journey again. Founders will have to be smart about what they build and how they build and how they monetize.” Venture investors have been looking more closely at the consumer AI segment and began allocating capital to it last year. Most are betting that the next set of large companies will emerge across health, education and entertainment. But as Indian companies build AI apps for consumers, computing costs will be an inhibitor in a business that is already expensive to scale. Compute costs are expenses incurred on the processing power, hardware and the energy required every time a consumer interacts with AI. Demand for AI hardware such as graphics processing units exceeds supply and is driving cloud rental and hardware lease rates higher. “Compute pricing keeps founders from building more mass-market solutions. Even now, founders have to build more vertical companies so they can actually have margins,” said Kumar. “It'll be a challenge for a few years at least.” Indian startup founders are willing to wait a few months for cheaper open weight AI models to emerge from China instead of paying the high fees to US companies for the use of models such as OpenAI’s ChatGPT and Anthropic’s Claude. Retention, metrics While Indian consumers are now more willing to pay for online services, especially in areas deemed important for growth like education, retention continues to be a challenge. One way to solve this, Kumar said, is to focus more on providing outcomes. “I don't think there's such a thing as an addictive product, at least for India. We're value-conscious people. We like value. If there's no results, people will not stick,” she said. For founders building AI products, the way they track metrics also has to change. Previous indicators such as monthly active users, churn rate and day-N retention become less relevant, especially if users have high engagement with specific parts of a product. “How do you measure quality of conversations happening on a platform, how much unique data are you collecting with every chat, what is the value you can add? Do you have a metric for your data depth? Are you measuring chat retention specifically? How are you measuring cost of servicing/token spend to retain each user? I don't see enough people measuring that today,” said Kumar. There has been some fundraising on the consumer AI front this year. In education, Arivihan raised $10 million from Accel and Prosus. Mint reported last week that AI-based language learning company Supernova was in talks to raise $15-20 million from Vertex Ventures and Verlinvest. But Kumar sees education as a segment that will require innovation beyond using voice AI or AI-generated videos. “There are ways to design products in education that augment critical thinking. We're not seeing too much of that just yet,” she said. Bullish on content Content is another area the firm is bullish on. Lightspeed's portfolio company, Pocket FM, an audio entertainment platform, is back in the market to raise $100-120 million at a valuation of $1.5-2 billion, Mint reported in April. Its rival, Kuku FM, filed for a $360 million initial public offering that would value it at about $1.8 billion. The company received regulatory approval last month. Companies are increasingly choosing to build their own products around using AI to generate content and to help produce user-generated content to drive retention. Pocket FM released Sherpa, Kuku FM has Kuku Studio and Stellaris Venture Partners-backed Dashverse released Frameo last September. This week, EloElo Group, a maker of short dramas, announced its agent studio called Dolphin. “AI helps make content a commodity. But really, it's about the plot and the story. That's where the uniqueness is, that's where the creative talent is needed and that is where scarcity is,” said Kumar, referring to companies building their own content-generation platforms. “If you solve for that, you get infinite supply, then production is an AI problem that is rapidly getting solved.” Preventive healthcare is an area of healthcare that is expected to grow. While some investors have focused on creating infrastructure businesses around it, as in the case of BioPeak, Cent Health, Nura and now Ultrahuman Labs, others like Lightspeed have focused on tech-enabled businesses like Superliving. The firm wrote the startup's $7 million Series A cheque just six months after Kae Capital provided $2 million in seed money. “The curative side is open, and we'd be keen on doing something in that space, but it'll be a more human-in-the-loop approach,” Kumar said. “Wearables with AI are interesting as well. The beauty would be when you have products that measure several things about you and then draw inferences about where your health is headed. That may be the future of preventive care.”