NVIDIA backs Simplismart's $14 million Series B to make AI inference cheaper
Simplismart, founded by BITS Pilani alumni Amritanshu Jain and Devansh Ghatak, has raised $14 million led by Dallas Venture Capital with NVIDIA, Accel and Shastra VC to expand software that gets more out of every GPU.
The AI boom has a cost problem, and most of it sits in inference: the computing needed every time a model answers a question, writes code or transcribes a call. Simplismart, a start-up founded by two BITS Pilani alumni with teams in India and the US, is betting that companies will pay for software that makes that step cheaper. On 27 October 2026 it said it had raised $14 million in a Series B round, with NVIDIA among the investors.
The round
The company says Dallas Venture Capital led the round, with NVIDIA, Accel, Shastra VC and Micromax participating.
What Simplismart does
Simplismart was founded in 2022 by Amritanshu Jain, its CEO, and Devansh Ghatak, its CTO. Jain previously worked at Oracle and Ghatak at Google; the two met as students at BITS Pilani.
The company builds software that sits between AI models and the GPUs that run them. Its job is to make inference faster and cheaper: squeezing more requests out of each GPU, cutting response times, and in some cases letting customers run the same workload on fewer chips. Simplismart also helps cloud providers turn raw GPU capacity into ready-to-use AI services they can sell to their own customers.
The company says its annualised revenue run rate grew 14 times over the past six months. An annualised run rate is a current operating measure rather than a guarantee of future recurring revenue.
Where the money goes
Simplismart said it will use the funds to:
- strengthen its inference engine and orchestration platform;
- expand partnerships with cloud providers and sovereign AI programmes;
- grow in North America and other markets.
The founder letter says the company is also building a dynamic GPU scheduling layer, software that decides in real time which workloads run on which chips, so that expensive hardware sits idle less often.
Why NVIDIA's cheque is interesting
At first glance, NVIDIA backing a company that helps customers use fewer GPUs looks odd. It makes more sense on a closer look. When inference gets cheaper, more companies can afford to put AI into production, and total demand for computing tends to grow. NVIDIA has invested widely in software companies that make its hardware easier and cheaper to use, and inference is where most AI spending is heading as models move from training into everyday use.
The investment also gives NVIDIA another link into India's AI build-out, where it already works with cloud providers, data centre operators and the government's IndiaAI Mission.
Why it matters for India
India's AI strategy rests heavily on access to computing. The IndiaAI Mission has subsidised tens of thousands of GPUs for start-ups and researchers, and companies such as Tata Communications, Neysa, Yotta and CtrlS are building out GPU capacity and AI data centres. But GPUs are expensive and scarce, and a sovereign AI stack is only affordable if that capacity is used efficiently.
That is the gap Simplismart is targeting. Software that improves utilisation and helps Indian cloud providers package GPUs as managed services makes each rupee of AI infrastructure spending go further. The company's focus on sovereign AI partnerships suggests it sees government-backed computing programmes, in India and elsewhere, as a significant market.
The round also says something about where Indian AI founders are building. Several outlets list Simplismart as headquartered in San Francisco, others in Bengaluru. Like many Indian AI infrastructure companies, it sells mostly to global customers while keeping much of its engineering in India. Its growth adds to India's base of engineers who understand low-level AI systems: compilers, kernels, schedulers and GPU memory management. These skills are in short supply worldwide, and India will need them if it is to run its own models on its own infrastructure.
The competitive picture
Inference optimisation is a crowded field. Simplismart competes with well-funded US start-ups, with open-source serving tools, and with the in-house tools of the large cloud providers. Its argument is that a specialised, hardware-aware layer can beat general-purpose tools on cost and speed.
The next year will test that argument. A 14-fold rise in run rate over six months is a strong signal, but usage-based revenue can be volatile. Simplismart now has $14 million and a set of strategic backers to turn early traction into a durable business.