India’s First GenAI Unicorn Walks Back Model Ambitions, Shifts Focus to AI Cloud Services

Krutrim, once billed as India’s leading generative-AI startup, said Tuesday it is moving away from building large-scale foundational models and toward offering AI cloud services to enterprises. The Bengaluru company — which paused its chip design work and reallocated capital and staff after a late-2025 restructuring — framed the change as a practical response to the rising costs of developing massive AI systems.

The pivot follows more than a year since Krutrim released its Krutrim-2 base model and a lengthy spell of limited public news: the firm’s last update on social channels was in December, and it was largely absent from major industry gatherings in recent months. The startup has also trimmed its workforce in repeated rounds, with Indian media reporting cuts that exceed 200 roles, and it pulled its consumer-facing AI assistant from app stores in April.

Krutrim reported about ₹3 billion (roughly $31.5 million) in revenue for the 2026 financial year, saying that represented roughly a three-times increase over the prior year and produced its first annual net profit with margins above one-tenth. The company declined to break down how much of that income came from outside customers versus revenue tied to Ola Group businesses; earlier coverage suggested roughly nine-tenths of FY25 sales were intra-group.

Executives said demand is rising for the startup’s AI cloud offerings, with the firm claiming more than two dozen enterprise clients spanning telecom, financial services and healthcare, and that most of its GPU capacity is already booked for external workloads. Krutrim had previously attracted investor attention, raising $50 million at a $1 billion valuation in early 2024 as founders pitched a homegrown alternative to overseas models.

Analysts say the move makes commercial sense but urged caution on headline profitability. “Shifting toward infrastructure can shore up revenues quicker than chasing a competitive base model, but claimed profits need independent verification,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. He added that for many Indian startups, selling compute and managed AI services may be the more realistic near-term route even if ambitions to develop proprietary models remain.

Krutrim did not provide detailed answers about its customer roster, the exact split of revenue sources, or the recent organizational changes when contacted. The company’s shift underscores a broader reassessment across the local AI sector as founders weigh the high fixed costs of model training and chip design against steadier business-from-cloud opportunities.