If you've been following India's AI story, you know Krutrim — Bhavish Aggarwal's ambitious attempt to build an entire AI stack in India. This week, a Mint report put a fresh number on how deeply Krutrim is now woven into the Ola group: Ola Electric's payments to Krutrim SI Designs nearly quadrupled to ₹100 crore in FY26. On the surface it's a corporate disclosure. Dig a little, and it's one of the clearest windows we have into India's sovereign AI cloud experiment — who's paying for it, and whether it's working.
01.What the numbers actually say
According to Ola Electric's annual related-party disclosures, the ₹100 crore breaks into two parts: ₹54 crore for services purchased from Krutrim (up from ₹26 crore the previous year) and a first-of-its-kind ₹46 crore advance for services ordered but not yet invoiced as of March 31. And here's the context that makes it interesting — this jump happened in a year when Ola Electric was cutting costs everywhere. Total expenses fell from ₹6,253 crore in FY25 to ₹3,245 crore in FY26, even as revenue from operations dropped from ₹4,514 crore to ₹2,253 crore, with a pre-tax loss of ₹1,829 crore. So the Krutrim bill went up roughly 4x while everything else shrank.
02.Why the bill jumped: zero external cloud
The company's explanation is straightforward. "Ola Electric has fully migrated its cloud and AI infrastructure to Krutrim, and expenses to external cloud vendors now stand at zero," a spokesperson said in an email reply to Mint. In plain terms: the money that used to go to the likes of Microsoft Azure now flows to Krutrim instead. Bhavish Aggarwal confirmed this direction back in May, when he posted on X that the Ola group had completely ended its dependence on foreign cloud infrastructure. If you're wondering what a cloud migration looks like at group scale — this is it. For smaller businesses doing the same math on a much tinier budget, having a framework for weighing AI and cloud costs before you commit is invaluable; the cost optimization chapter in our GPT-6 Astra — Zero to Hero Guide exists for exactly this kind of "should I migrate or not" decision.
03.Krutrim's quiet reinvention
Here's the part I find most interesting. Krutrim launched in January 2024 with grand plans — its own LLM trained on Indian data (it understands roughly 22 Indian languages and generates text in 10), its own AI chips, even a consumer assistant called Kruti. It raised $50 million from Matrix Partners at a $1 billion valuation, becoming India's first AI unicorn. But over the past year, the story changed. Funding dried up, senior leaders left, work on models and chips was paused, and Kruti was pulled from app stores. What emerged instead is an "AI-first cloud" — GPU compute and infrastructure for Indian teams, claiming 500+ teams on its platform. And by the company's own account, it's now profitable: FY26 revenue of around ₹300 crore with a first annual profit and a PAT margin above 10%. Honestly, what surprised me most is that a sovereign AI infrastructure bet turned profitable this early — even if it's profitable as a cloud company, not as a frontier model lab.
04.The governance question
Now the uncomfortable part. Krutrim is not an arms-length vendor. Aggarwal is chairman and managing director of Ola Electric, and he controls Krutrim through his family office — Krutrim AI Designs LLP holds nearly 90% of the operating entity. Two directors, Krishnamurthy Venugopala Tenneti and Arun Sarin, sit on both boards. Ola Electric's six-member board has three independent directors; Krutrim's five-member board has none. Related-party transactions aren't illegal, and Ola Electric maintains these were done on arm's-length terms with the required approvals. But governance experts quoted in the Mint report have flagged the ₹46 crore advance and the quadrupling bill as things listed-company shareholders should keep an eye on. Fair point. The counterpoint: if the cloud migration genuinely works, the spend is an investment rather than a leak.
05.The bigger bet: an Indian AI cloud
Zoom out and this is really a story about sovereign AI infrastructure — the same theme running through NTA moving NEET translations onto its own air-gapped AI and, at the funding level, India's AI-only VC funds closing bigger and bigger cheques. Krutrim's pitch is that Indian companies shouldn't have to rent their compute from American hyperscalers forever. It's a genuinely hard business — the competition is Azure, Oracle and AWS — but if an Ola-scale group can run entirely on Indian AI infrastructure and Krutrim can hold its first profit, the "India builds its own AI backbone" argument gets its first real data point.
Key Takeaways
- Ola Electric's payments to Krutrim nearly quadrupled to ₹100 crore in FY26 — ₹54 crore for services plus a ₹46 crore advance
- All of Ola Electric's cloud and AI infrastructure now runs on Krutrim; external cloud vendor spend is zero
- Krutrim reported FY26 revenue of ~₹300 crore and its first annual profit, with a PAT margin above 10%
- The related-party structure — Aggarwal's family office owns ~90% of the operating entity — has governance experts asking questions
- Krutrim has pivoted from LLMs and chips to an "AI-first cloud" for Indian teams
Frequently Asked Questions
Why did Ola Electric's payments to Krutrim jump?
Ola Electric's payments to Bhavish Aggarwal's AI firm Krutrim nearly quadrupled to ₹100 crore in FY26, because Ola Electric has moved its entire cloud and AI infrastructure to Krutrim — external cloud spend is now zero.
How is Krutrim performing?
Krutrim reported FY26 revenue of about ₹300 crore and its first annual profit, with a PAT margin above 10%. The related-party angle has governance experts asking questions.
What This Means for You
If you're an engineer or a cloud professional in India, Krutrim's AI-cloud pivot is worth watching — a homegrown hyperscaler alternative means new jobs, new pricing pressure, and a new "sovereign infrastructure" line on your resume. If you run a business, the Ola case study is a useful reminder that cloud and AI spend is now a board-level number, not an IT line item. And if you're a student trying to make sense of where India's AI economy is actually making money today — it's not in chatbots, it's in the boring, profitable layer underneath: compute, migration and infrastructure. To work in that layer, you need solid fundamentals — prompting, agents, cost optimization, safety — and our GPT-6 Astra — Zero to Hero Guide covers all of it in one practical package.