Nvidia Joins $14M Round For Indian Startup That Cuts GPU Needs For AI Inference

Nvidia and a group of investors have put $14 million into Bengaluru-based Simplismart, valuing the four-year-old company at roughly $100 million after the new cash. Dallas Venture Capital led the Series B, which included Accel, Shastra VC and Micromax; Nvidia participated as a strategic backer. The financing aims to help Simplismart broaden its reach in the U.S., hire more engineers and accelerate development of software that aims to make running AI models less GPU-intensive.

The startup sells middleware that sits between machine-learning models and the machines that run them, adjusting deployment and resource use to match business targets for latency, throughput and cost. Companies are pouring money into accelerators, yet putting models into production often requires painstaking tuning. Simplismart’s tools automate many of those settings so teams don’t have to hand-configure trade-offs for each workload.

Customers can use Simplismart’s platform directly or install the software within their own cloud or datacenter environments. The product set includes model fine-tuning utilities, automatic scaling that reacts to demand, and performance-monitoring dashboards. The company also licenses its stack to cloud operators that want to offer managed inference services on GPU infrastructure.

Simplismart cites several clients in media and enterprise AI. One customer in video generation reduced GPU serving bills by a little more than half after adopting the software, the company said. Other partners include HeyGen, Higgsfield and enterprise startup Ema, and Indian cloud provider Yotta has tapped the technology for the serverless inference option in its Shakti Studio offering. Simplismart added that its annualized revenue run rate has climbed about 14 times in the past half year to an eight-figure level; that figure reflects current usage, not contracted recurring income, and excludes amounts passed through third-party infrastructure vendors. More than one-half of its sales come from customers in the United States, despite a larger headcount of buyers in India.

Nvidia’s involvement extends beyond the checkbook. Simplismart said its engineers are working with Nvidia solution architects on integrations tied to the chipmaker’s Dynamo inference software, and that Nvidia has introduced the startup to potential clients and partners. Some observers see a contradiction in a major GPU supplier funding software that can reduce hardware needs. But industry analysts say making inference cheaper and more predictable can expand overall demand for accelerators by opening up use cases that were previously too costly.

“Tools that let customers dial in performance and cost help both buyers and hardware makers,” said Anjali Rao, an infrastructure analyst based in Bangalore. “If running models becomes less painful and more economical, more teams will deploy them, and that supports the whole market for specialized processors.”

Simplismart was founded in 2022 by former Oracle engineer Amritanshu Jain and ex-Google engineer Devansh Ghatak, who met at BITS Pilani. The startup employs 58 people. It faces stiff competition from vendor offerings — including from Nvidia itself — and from major cloud providers such as Amazon, Google and Microsoft, as well as a growing list of niche startups focused on inference efficiency. A report in an Indian business daily had suggested larger fundraising plans and a much higher valuation; Jain declined to discuss that story and told reporters the recently closed Series B will not be extended.