Nvidia has invested in Indian startup SimplySmart, which aims to cut companies' artificial intelligence model operating costs and improve GPU efficiency. As demand for AI inference grows, the move is seen as a bet on software technology that uses expensive GPUs more efficiently.

On Oct. 27, local time, tech media outlet TechCrunch reported SimplySmart raised $14 million in a Series B round. Dallas Venture Capital led the round, with Nvidia, Accel, Shastra VC and Micromax participating. After the investment, which was carried out through a share issuance, the company's valuation was estimated at about $100 million. Founded in 2022, SimplySmart is headquartered in Bengaluru, India.

SimplySmart develops inference software needed to run AI models in live service environments. For companies to operate AI services, they need not only to secure expensive GPUs but also to balance response speed, throughput and costs. That is because required performance differs by service, making it necessary to fine-tune model deployment and how computing resources are used.

The company provides software that optimises operations between AI models and computing infrastructure. It boosts utilisation of existing GPUs by adjusting response time, the number of requests it can handle and the amount of computing resources used to meet customer needs. It also supports features including model fine-tuning, auto-scaling in response to demand changes and performance monitoring, and it can be applied to existing cloud and on-premises environments. It also has functions to help cloud providers build and operate AI inference services based on their own GPU infrastructure.

Amritanshu Jain (아므리탄슈 자인), SimplySmart's chief executive, stressed that each company has different needs for AI inference performance and cost levels. Rather than offering the fastest or cheapest service unconditionally, he said the company's competitiveness lies in finding the best operating approach that fits each customer's service level agreement and total cost of ownership conditions.

Customers have also seen cost-cutting effects. AI video generation platform InVideo said it cut GPU serving costs by 56 percent after adopting SimplySmart's software. The company is also working with HeyGen, Higgsfield and Emma. Indian cloud provider Yotta applied SimplySmart's technology to the serverless AI inference service offered on its 'Shakti Studio' platform.

Revenue is also growing quickly. Jain said the company's annualised revenue run-rate increased 14-fold over the past six months, reaching an eight-digit dollar level. He said the figure is not recurring revenue confirmed by contracts, but an estimate based on recent customer usage. It also excludes revenue that accrues to infrastructure providers, and the company did not disclose an exact amount.

More than half of revenue comes from the U.S. market, but the number of customers is higher in India. SimplySmart plans to use the funds to expand in the United States, hire more staff and upgrade its inference software.

Cooperation with Nvidia also carries meaning beyond the investment. Jain said the two sides are working on technical integration centred on Nvidia's inference software, Dynamo. He said engineers from both companies and Nvidia's solution design organisation are working together, and that Nvidia is also helping connect potential customers and partners.

Attention is also on the fact that Nvidia invested in a company with technology that reduces GPU usage. If GPU efficiency improves, more AI services can be run on the same hardware, potentially lowering computing costs per task. At the same time, if lower costs lead to more companies and applications adopting AI, overall demand for GPU infrastructure could expand.

The competitive environment is not easy. SimplySmart must compete not only with Nvidia's own products but also with major cloud providers such as Amazon Web Services, Google Cloud and Microsoft Azure. As AI services spread, competition is also intensifying among software companies seeking to improve inference costs and operational efficiency as well as GPU performance.

Meanwhile, India's business daily Economic Times reported SimplySmart is recently discussing an additional funding round of $40 million to $50 million and a valuation of $300 million to $400 million. Jain said the company has not been actively pursuing additional fundraising. He said the Series B round has already closed and there are no plans to increase the size of the round.