Nvidia Backs Indian Startup Simplismart to Run AI Models With Fewer GPUs

Nvidia participated in a $14 million Series B round in Bengaluru-based Simplismart, a startup building software that helps companies run AI models more efficiently on existing GPU infrastructure. The round was led by Dallas Venture Capital, with Nvidia, Accel, Shastra VC, and Micromax also participating. The deal valued the four-year-old company at approximately $100 million, including the new capital.

What Simplismart Actually Does

Simplismart builds a software layer that sits between AI models and the compute infrastructure hosting them. Rather than requiring engineering teams to manually tune deployment parameters — latency, throughput, memory utilization — Simplismart's software automates that tuning work, adjusting how each model is served according to the customer's specific requirements. The approach works both in the cloud and on on-premises infrastructure.

The company has three business lines. For AI startups and enterprises, it offers inference optimization as a service. For larger organizations, it provides a unified control panel for managing multiple production models. And thirdly, it allows cloud providers to build and run their own AI inference services on GPU infrastructure using Simplismart's software — turning the company from a tool for AI firms into a white-label engine for data center operators.

The most cited case is the video-generation platform InVideo, which according to Simplismart cut its GPU bill by 56% after adopting the platform. Other notable customers include AI startups Ema and HeyGen.

Why Nvidia Invested

Nvidia's participation is strategic, not purely financial. At face value the bet looks paradoxical — the world's dominant AI chip supplier backing software that lets companies use fewer GPUs. But the logic is consistent with the direction Nvidia has already taken across its other initiatives. The company has historically expanded from a hardware vendor into a platform provider: CUDA, cuDNN, TensorRT, and now the NIM microservices ecosystem. Simplismart fits that pattern by creating an abstraction layer that makes inference infrastructure more accessible, which in turn expands the total volume of inference run on Nvidia hardware.

"Making inference more efficient ultimately expands usage," explained Amritanshu Jain, co-founder of Simplismart, citing the same reasoning behind Nvidia's investments in Sarvam AI and partnerships with local cloud and infrastructure providers in India.

The growth figures impress even by round standards. Simplismart reported a 14-fold increase in its annualized revenue run rate over six months. More than half of its revenue comes from the United States, despite the company having more customers in India — a pattern common among Indian infrastructure startups that attract higher per-customer spending from American clients than from their larger Indian user base.

The Macro Shift: Inference Over Training

The bet on Simplismart reflects a structural shift in the AI market. During the first years of the generative boom, attention and capital concentrated on training ever-larger models — a multi-billion-dollar game of chips and data centers. But as models reach satisfactory capabilities, the cost curve shifts to inference: running these models in production, where the volume of requests far exceeds the volume of training.

This pattern has already been observed with other startups. d-Matrix, for example, developed inference-dedicated chips rather than training chips. Perplexity announced plans to use Nvidia's new Vera CPUs for AI agent inference tasks. Modal, a startup that rents compute power to AI developers, began using Vera chips specifically to accelerate autonomous agent inference.

The difference is that Simplismart does not attempt to replace the chip — it tries to squeeze more performance from the existing chip. It's the same principle that drove Nvidia's TensorRT and the open-source vLLM, but applied to a complete commercial product with control panels, performance metrics, and cloud provider integration.

What Comes Next

Simplismart plans to use the round's capital to expand its presence in the United States, grow its team, and invest in further inference software development. Jain confirmed that the United States remains the key market, given that more than half of revenue already comes from American clients.

The $100 million valuation appears conservative compared to earlier conversations: The Economic Times had reported that Simplismart was in talks to raise $40 million to $50 million at a valuation of $300 million to $400 million. Jain dismissed that possibility, saying the round was closed and would not be extended.

The question remains whether Simplismart can scale sufficiently to justify a multi-digit exit in a market increasingly contested — where both inference-dedicated chips and software solutions compete for the same spend on GPU infrastructure. Nvidia's bet suggests it believes in the answer.

Sources: TechCrunch, Dealroom, Progressive Robot

✓ Independent sources cross-checked and verified before publishing