Nvidia backs Indian inference startup Simplismart

Dallas Venture Capital led the all-primary round, with Nvidia, Accel, Shastra VC and Micromax also putting in money. Including the new cash, the four-year-old, Bengaluru-based company is now valued at about $100 million.

Simplismart builds software that sits between AI models and whatever infrastructure runs them, handling the tuning and deployment work — speed, cost, request volume — that companies would otherwise have to manage on their own. The pitch has traction: video-generation platform InVideo says the tool trimmed its GPU bill by more than half (56%), and customers now include video startups HeyGen and Higgsfield, enterprise-AI firm Ema, and Indian cloud company Yotta, which runs its serverless inference offering on Simplismart’s stack. Cloud providers, too, can use the software to stand up their own GPU-backed inference services.

CEO and co-founder Amritanshu Jain told TechCrunch that annualized revenue has multiplied 14 times over the past six months into the eight figures — a number he calculated from current customer usage, not signed contracts, and one that strips out money simply passed through to infrastructure providers; he wouldn’t give the precise figure. More than half of that revenue is American, he said, even though Simplismart has more customers in India than in the U.S. Nvidia’s involvement goes beyond the check: its engineers are collaborating with Simplismart on integrating Nvidia’s Dynamo inference software, and the chipmaker has introduced the startup to prospective customers and partners. The arrangement reads as a paradox — the industry’s dominant GPU supplier funding a company that helps customers need fewer GPUs — but Jain argues efficiency gains make Nvidia’s chips more attractive overall and expand AI adoption, which should ultimately benefit Nvidia too.

Jain, previously an engineer at Oracle, and co-founder Devansh Ghatak, previously at Google, founded Simplismart in 2022 after meeting as students at BITS Pilani; the company has since grown to 58 employees. It competes for inference business against Nvidia’s own offerings, the major hyperscalers — Amazon Web Services, Google Cloud and Microsoft Azure — and a crowded field of newer startups, but Jain says Simplismart wins customers with strict service-level and cost requirements by tailoring a balance rather than offering preset fast-or-cheap tiers. The fresh capital will fund U.S. expansion, hiring, and further development of the inference software. The round comes after The Economic Times reported Simplismart was negotiating a larger raise — $40 million to $50 million at a $300 million to $400 million valuation; Jain declined to address that report directly but said the Series B had closed and would not be extended.

Why it matters

That the industry’s dominant chip supplier is funding software that lets some customers use fewer of its GPUs looks contradictory, but it reflects Nvidia’s bet that cheaper, more efficient inference grows total AI demand rather than shrinking its own market.