Krutrim — India's first generative AI unicorn — is done pretending it can compete with the frontier model companies.

The company founded by Bhavish Aggarwal (Ola, Ola Electric) raised $50 million at a $1 billion valuation in January 2024, backed by SoftBank. The pitch was India-built AI models for India-scale problems. The reality: building frontier AI models is expensive in ways that make the business model math hard.

The Pivot

Krutrim is shifting to cloud services. The late-2025 overhaul already paused chip design efforts — the company tried to build custom silicon for AI workloads and abandoned it. Now the pivot is away from AI model development entirely, toward providing cloud infrastructure and services.

The Revenue Reality

Krutrim reports ₹3 billion (~$31.5 million) in FY26 revenue with first annual profit and >10% margins. That's a real business with real margins.

But the asterisk: approximately 90% of that revenue comes from group companies — Ola, Ola Electric, and related entities. That's not market validation. That's self-dealing within an affiliated business group.

The $31.5M revenue figure looks impressive until you realize it's largely intra-group transfers. The actual market for Krutrim's services outside the Aggarwal ecosystem is essentially unproven.

Why the Model Business Failed

The economics of building large-scale AI models are brutal:

  • Training costs run into hundreds of millions of dollars for frontier models
  • Ongoing inference costs scale with usage
  • The gap between frontier model capability and second-tier models keeps widening
  • Enterprise buyers have shown they prefer established players over newcomers

India-built models face an additional challenge: the domestic enterprise market that would pay for India-specific AI is still developing. The companies that would be the natural customers for India-first AI are still in growth phases, not enterprise-software spending phases.

What This Tells You About India's AI Landscape

Krutrim's pivot is a useful data point, not a verdict on India's AI potential. The country has genuine strengths: large engineering talent pool, growing enterprise market, government AI initiatives.

But the story of "India builds its own ChatGPT competitor" was always going to run into the same wall: the capital requirements for frontier AI are on a different scale than what most Indian VC ecosystems can support. Krutrim had $50M — OpenAI has raised billions. The math was always going to favor the capital-rich player.

The cloud services pivot is sensible: use the infrastructure you've built to provide services where you have a real addressable market, rather than chasing a model capability race you can't win.

For Indian tech founders: Krutrim's experience is a lesson in the difference between being a "first" in a category and being competitive in a category. First mover in India's GenAI space is a trophy. Sustainable business is the goal.

Sources: TechCrunch