ET WLF 2026: AI driving shift in value creation

As artificial intelligence makes software development faster and cheaper, value creation is expected to shift from software applications to computing infrastructure, silicon chips, power and physical systems, business heads said at The Economic Times World Leaders Forum.

Shailendra Singh, managing director of venture capital firm Peak XV Partners, told ET's Samidha Sharma that software had historically been difficult and expensive to build, which allowed application companies to capture much of the value. AI could change that equation by making software easier to write, rewrite and replatform.

"Once you flip that on its head and software is fungible and can be rewritten, replatformed, etc., very quickly... the value starts to shift to the infrastructure layer, to silicon, to the people who are going and going to run compute," Singh said. That shift, he said, would increase the importance of the physical infrastructure needed to run AI.


"What do you need to substitute human intelligence for machine intelligence? You need tonnes of compute and tonnes of electricity at a cheap cost," he said.

India could benefit from this trend because the cost of power and data-centre construction and operation is relatively attractive, Singh said.

"There is a chance that we may see a monster build-out of data-centric capacity in India in the next 5 or 10 years," he said.

Singh said venture capital firms have become "somewhat marginal" in the AI investment cycle given the scale of capital required, and large enterprises should increasingly fund AI companies alongside traditional investors.

Pointing to HCLTech's investment in Indian foundational model startup Sarvam AI, and global bets such as Amazon's backing of Anthropic, he said corporate balance sheets can deploy capital at a scale most venture funds cannot match. The shift in software economics is also changing how AI companies build businesses. Krish Ramineni, founder and chief executive of AI meeting assistant startup Fireflies.ai, said companies may initially depend on existing models but can differentiate themselves as they build enterprise-specific capabilities around them. "Any company that starts off, even if it was a wrapper or something that's making an LLM call, over time, it hardens, it solidifies, you build a lot of enterprise business logic into it," Ramineni said.

The CEO said the economics of AI would also change as the cost of intelligence falls. "When it becomes 10x cheaper, you're going to use it 100x more," he said.

This expansion is particularly relevant in India, where companies building AI products have to deal with multiple languages and a wide range of users. Tanay Kothari, co-founder and CEO of voice AI startup Wispr, said India should build products around its own requirements rather than replicate what is being developed in the US.

"India comes with its own set of unique challenges, which is dozens of languages. You need to be extremely cost efficient ...the kinds of problems you need to solve are just fundamentally different," Kothari said, pointing to the need to support dozens of languages while keeping products affordable.

Wispr is looking at experiences that could cost users Rs 1-2, instead of products that are priced for the purchasing power of developed markets, he said.

Kothari also sees voice as a way to bring more people into technology, particularly users who are not comfortable typing on phones or computers.

"It's basic infrastructure," Kothari said, describing voice technology as an enabler for people to use technology rather than simply another AI application.

Krishna Rangasayee, founder and CEO of edge inference chip startup SiMa.ai, said processing AI on devices could become more important because of privacy, security, cost and latency considerations. "What we are seeing is really a good groundswell for robotics, industrial automation and automotive," Rangasayee said, adding that India is already seeing increased activity in drones, robotics, industrial automation and automotive applications.

The shift towards physical AI also has implications for national security and manufacturing. "National sovereignty and ownership of protecting borders is going to become the most critical topic on AI," Rangasayee said.

He said he expects more AI processing to happen outside centralised cloud systems as governments and enterprises become more concerned about where sensitive information is stored and processed.