Key Takeaways

  • Simplismart’s Series B values the Bengaluru startup at about $100 million and brings Nvidia into its investor group.
  • The startup says its platform helped InVideo reduce GPU serving costs by 56% by tuning model deployment and resource use.
  • Growing inference demand is shifting enterprise attention from acquiring GPUs to improving utilization, throughput, latency, and total cost.

Simplismart has raised $14 million in an all-primary Series B round led by Dallas Venture Capital, with Nvidia, Accel, Shastra VC, and Micromax participating. The financing values the four-year-old startup at about $100 million, including the new capital.

The deal puts fresh attention on a less glamorous but increasingly important part of the artificial intelligence stack: inference. Training a model captures headlines, but serving that model reliably to thousands or millions of users creates a recurring infrastructure bill. Small changes in latency, batching, autoscaling, and GPU utilization can materially alter the economics.

Simplismart positions its software between AI models and the infrastructure on which they run. Customers can use the platform in Simplismart’s environment or deploy it across existing cloud and on-premises systems. Its capabilities include model deployment, fine-tuning, automatic scaling, monitoring, and optimization of response time, throughput, and compute consumption.

“Nobody actually needs the fastest or the cheapest inference,” co-founder and CEO Amritanshu Jain told TechCrunch. The practical requirement, he argued, is a combination of performance and cost suited to a particular application.

That distinction matters. A real-time conversational service may prioritize low latency, while a video-generation workload might tolerate longer response times in exchange for higher throughput or lower cost. Enterprise deployments can also carry service-level commitments that make a one-size-fits-all configuration unappealing.

Simplismart says AI video-generation platform InVideo reduced its GPU serving costs by 56% after adopting the platform. Its customers also include HeyGen, Higgsfield, and enterprise AI startup Ema. Indian cloud provider Yotta uses Simplismart technology for the serverless inference service within its Shakti Studio platform.

Simplismart reports its annualized revenue run rate increased 14-fold in six months and reached an eight-figure sum. That calculation reflects recent customer usage rather than contracted recurring revenue, and it excludes revenue passed through to infrastructure providers. More than half of revenue comes from the U.S., even though Simplismart has more customers in India.

Buying more accelerators is only one response to capacity pressure. Better scheduling, quantization, request batching, model routing, and autoscaling can increase the work completed by an existing GPU fleet. The amount of spare capacity hiding inside poorly tuned inference deployments varies by workload, driving Simplismart's optimization strategy.

Nvidia’s participation may look counterintuitive because Simplismart can help customers run certain workloads with fewer GPUs. Yet greater efficiency can expand the number of economically viable AI applications, potentially increasing overall demand for Nvidia hardware. Nvidia is also working with Simplismart on technical integrations involving Dynamo, its inference-serving software, while providing introductions to prospective customers and partners.

Competition will be intense. Nvidia, Amazon Web Services, Google Cloud, and Microsoft Azure are all developing inference infrastructure and software. The broader category also includes specialist products focused on model serving, quantization, autoscaling, observability, and GPU utilization. Kubernetes commonly provides the orchestration layer, while OpenTelemetry can give operators vendor-neutral visibility across distributed inference services.

Demand provides room for multiple approaches. Mordor Intelligence projects the global AI-inference GPU market will expand from $11.89 billion in 2025 to $57.29 billion by 2031, representing a 30.97% compound annual growth rate. Meanwhile, Reuters reported in 2025 that India’s Deep Tech Alliance had secured more than $850 million in capital commitments across AI, semiconductors, robotics, and space.

Founded in 2022 by former Oracle engineer Jain and former Google engineer Devansh Ghatak, Simplismart now employs 58 people. The founders met at BITS Pilani and built the business from their experience deploying machine-learning systems.

The new capital will support U.S. expansion, hiring, and further product development. Jain also rejected the suggestion that the round would be extended after The Economic Times reported separate fundraising discussions. For enterprise buyers, the bigger test now is whether Simplismart can reproduce its reported cost gains across more model types while preserving the service levels production applications require.