Nvidia Backs Simplismart's $14 Million Series B, a Bet That AI Needs Fewer GPUs, Not More

As the AI boom strains global computing capacity, Nvidia has made an unusual bet: it invested in software designed to help companies run AI models with fewer GPUs. Simplismart, a Bengaluru-based startup founded in 2022, raised a $14 million Series B led by Dallas Venture Capital, with Nvidia, Accel, Shastra VC and Micromax joining the all-primary round. The deal values the four-year-old company at about $100 million including the new capital, according to TechCrunch.

The apparent paradox is the story. Nvidia dominates AI chip sales, yet it is now funding a company whose product optimizes inference workloads so some tasks need less of that hardware. Simplismart's co-founder and CEO Amritanshu Jain sees no contradiction: making inference cheaper and more reliable, he argues, ultimately makes AI applications — and the hardware they run on — more attractive. Nvidia has gone further than a check, collaborating with Simplismart engineers on technical integrations around its Dynamo inference software and introducing the startup to prospective customers.

Simplismart's software sits between AI models and the computing infrastructure they run on. Different workloads demand different trade-offs between response speed, throughput and cost, and the platform automates the tuning that companies otherwise do by hand — how models are deployed, scaled and monitored across cloud and on-premises infrastructure. Cloud providers can also use it to build their own inference services on GPU hardware. Jain's pitch is that customers rarely need the fastest or the cheapest inference in isolation; they need a workable balance, which is where he says Simplismart wins against rivals offering preset fast-or-cheap modes — on strict service-level agreements and total cost of ownership.

The traction numbers come from the company and have not been independently audited. Simplismart says its annualized revenue run rate grew 14 times over the six months before the announcement to an eight-figure sum, a usage-based figure that excludes revenue passed through to infrastructure providers. It says its software helped AI video platform InVideo cut GPU serving costs by 56 percent. Other customers include video companies HeyGen and Higgsfield, enterprise AI startup Ema, and Indian cloud provider Yotta, which uses the technology for serverless inference in its Shakti Studio platform. More than half of revenue comes from the United States, Jain said, even though most customers are in India.

Founded by former Oracle engineer Jain and ex-Google engineer Devansh Ghatak, who met at BITS Pilani, the company now employs 58 people and plans to use the funding to expand in the US market and deepen its inference software. The competitive landscape is crowded: Nvidia itself, the major hyperscalers, and a growing pack of inference-optimization startups all sell adjacent tooling. Jain declined to comment on a report that the company had discussed a larger round at a higher valuation, saying the Series B was closed and Simplismart was not actively fundraising.

The round lands at a telling moment. GPU financing has become the AI economy's biggest bottleneck, with debt markets pulling back from neoclouds while hyperscalers commit hundreds of billions. In that environment, software that squeezes more work out of chips companies already own is not a nice-to-have — it is the arbitrage between $40,000 GPUs and the bills they generate. Nvidia investing in that arbitrage says as much about the market's direction as any roadmap slide.