Simplismart raises $14M Series B to run AI models on fewer GPUs

Nvidia sells AI chips. It just funded a startup whose job is helping customers use fewer of them. TechCrunch reports that Nvidia joined a $14 million Series B round in Simplismart. Dallas Venture Capital led the round, with Accel, Shastra VC and Micromax. The deal values the four-year-old Indian company at about $100 million.

What the software does

Simplismart sits between a model and the hardware it runs on. Its software adjusts how a model is deployed. The mix of response time, throughput and cost can then fit the job. The company says a team can get more work out of its existing GPU stack.

Co-founder and CEO Amritanshu Jain told TechCrunch that most buyers do not need the fastest inference or the cheapest. They need a speed and a price that fit the job. Simplismart's platform runs in its own cloud or inside a customer's cloud or data centre. It includes fine-tuning tools, autoscaling and monitoring.

The customers and the growth claim

Simplismart says it cut InVideo's GPU serving costs by 56 percent. HeyGen and Higgsfield, both AI video companies, are also named customers. So is Ema, an enterprise AI startup. Indian cloud provider Yotta runs the serverless AI inference inside its Shakti Studio platform on Simplismart's software.

The company told TechCrunch its annualised revenue run rate has grown 14-fold over the past six months. The sum is in the eight figures. Jain did not give the exact figure. The number is based on recent customer usage, not on contracted recurring revenue. It also leaves out revenue passed through to infrastructure providers. More than half of revenue now comes from the United States, Jain said, even though more customers are in India.

Why Nvidia is in the round

Nvidia is the biggest seller of AI chips. Backing software that helps buyers use fewer chips looks strange on paper. Jain's answer to TechCrunch was simple. Cheaper inference means more AI applications reach production. More AI applications then need more chips to run. The two companies have been working together on technical integrations, including around Nvidia's Dynamo inference software. Nvidia has also made introductions to potential customers.

Simplismart was founded in 2022 by Jain and Devansh Ghatak. Jain was an engineer at Oracle, Ghatak at Google. They met at BITS Pilani. The startup now has 58 people. It plans to use the new money to grow its U.S. team.

Deployment is where the profit now sits in AI products. Picking the model is close to a commodity. Turning a model into a feature with a cost and a latency a business can accept is harder. That is where paid vendors are now selling. If a team cannot explain its own cost per request today, competitors are already paying less for the same work.

Source

Nvidia backs Indian startup helping companies run AI models with fewer GPUs, TechCrunch.