Simplismart Draws Nvidia Backing as Inference Optimization Becomes Priority

Nvidia has invested $14 million in Bengaluru-based Simplismart, a startup developing software to optimize AI inference, as enterprise focus shifts from raw model capability to production efficiency.

AI & Culture Desk
Updated 3d ago 2 min read
Simplismart Draws Nvidia Backing as Inference Optimization Becomes Priority

Nvidia has joined a $14 million Series B funding round for Simplismart, a four-year-old startup based in Bengaluru. The investment, which values the company at approximately $100 million, includes participation from Dallas Venture Capital, which led the round, alongside Accel, Shastra VC, and Micromax. The move underscores a growing industry pivot toward serving harness engineering—the architectural layer that manages memory, scheduling, and inference optimization. As organizations transition from proof-of-concept experiments to high-volume production, the primary engineering challenge has shifted from simply acquiring compute to maximizing the efficiency of existing hardware infrastructure.

Simplismart’s platform functions as an intermediary layer between AI models and underlying compute infrastructure. According to CEO Amritanshu Jain, the software automates the configuration of how models are deployed, allowing companies to balance latency, throughput, and cost based on specific application requirements. Rather than relying on a one-size-fits-all approach, the system adjusts parameters such as request routing and hardware utilization. The company reports that its technology has enabled clients, including video-generation platform InVideo, to reduce GPU serving costs by 56%. Other enterprise users include HeyGen, Higgsfield, Ema, and the Indian cloud provider Yotta.

The investment presents a strategic nuance for Nvidia, the dominant supplier of AI hardware. While Simplismart’s software helps companies run workloads with fewer GPUs, Jain argues that this efficiency actually benefits Nvidia by lowering the barrier to entry for enterprise AI adoption. The two companies are already collaborating on technical integrations, specifically involving Simplismart’s Dynamo inference software. Nvidia’s solutions architects are working directly with Simplismart’s engineering team, and the hardware giant is actively facilitating introductions to potential customers and partners, according to Jain.

Simplismart’s commercial traction highlights the intensifying demand for deployment-focused software. The company reported a 14-fold increase in its annualized revenue run rate over the past six months, reaching an eight-figure sum. Despite being headquartered in India, more than half of the startup’s revenue originates from U.S.-based technology companies. This growth reflects a broader market trend where foundation model capabilities are becoming increasingly commoditized, forcing enterprises to compete on the economics of their production backends.

The necessity of such optimization tools is rooted in the physical constraints of production AI. Successful deployment requires identifying specific bottlenecks—such as GPU memory pressure or KV cache growth—rather than simply adding more compute nodes. Because current hardware and model architectures evolve rapidly, enterprises are increasingly prioritizing orchestration tools that can adapt to new requirements without necessitating a complete architectural overhaul. Simplismart, founded by former Oracle and Google engineers, now employs 58 people and plans to use the new capital to expand its U.S. presence and further develop its inference stack.

Sources

  1. 01 Nvidia backs Indian startup helping companies run AI models with fewer GPUs — TechCrunch
  2. 02 Designing an AI serving harness built to last — MIT Tech Review