Simplismart Raises 14 Million Dollars With Nvidia Aboard to Make AI Inference Cheaper

Simplismart, a Bengaluru startup that builds the software layer between AI models and the GPUs that run them, has raised 14 million dollars in a Series B round announced October 27. Dallas Venture Capital led the all-primary round, with Nvidia, Accel, Shastra VC, and Micromax participating. Reports put the post-money valuation at about 100 million dollars.

The company is four years old, employs 58 people, and was founded in 2022 by Amritanshu Jain, a former Oracle engineer, and Devansh Ghatak, a former Google engineer who met Jain while studying at BITS Pilani. Its previous round was a 7 million dollar Series A led by Accel in October 2024. The new money will fund US expansion, hiring, and deeper development of its inference engine and orchestration platform.

What Simplismart Series B money buys

Inference is the unglamorous half of the AI business. Training a model gets the headlines, but running it for real users, millions of times a day, is where the bills pile up. Simplismart sells software that deploys, scales, and operates models in production, aiming to hit customers' latency and cost targets with fewer GPUs than they would otherwise need.

The company says the pitch is working. Its annualized revenue run rate has grown 14 times over the past six months into eight figures, according to Jain, who calculated that number from current customer usage rather than signed contracts. For one customer, enterprise AI company Ema, Simplismart says it targeted roughly 200 millisecond latency at about 9 cents per million tokens. Another report says it cut one customer's GPU serving costs by 56 percent. Those are company claims, not audited results, but they explain why Nvidia wanted in: cheaper inference sells more GPUs in the long run.

Why Nvidia's name on the cap table matters

Nvidia does not need to invest in a 58-person startup for the money. It invests to shape the ecosystem around its hardware. Simplismart says its engineers work directly with Nvidia to make AI infrastructure easier to deploy at scale, and its software powers the serverless inference layer of Yotta's Shakti Studio. For a startup selling to cloud providers and sovereign AI programs across India, North America, the Middle East, Australia, and East Asia, Nvidia's backing is a credibility shortcut with exactly the customers it needs.

There is also a telling detail in what did not happen. The Economic Times had reported Simplismart was negotiating a larger raise of 40 to 50 million dollars at a 300 to 400 million dollar valuation. Jain declined to address that report directly but said the Series B had closed and would not be extended. A smaller, closed round at a lower valuation is not a failure, but it is a reminder that the funding market in late 2026 rewards proven usage over big stories. Simplismart chose to bank a real round rather than chase a bigger headline.

The inference gold rush is getting crowded

Simplismart is not alone in this layer of the stack. The past month alone has seen large rounds for AI infrastructure, including Upwind's 400 million dollar raise and Collective's 100 million dollar round for its AI financial operating system. Inference is attractive because it is a cost problem every AI company shares, and cost problems produce paying customers faster than research breakthroughs do.

The risk for Simplismart is the same as for every infrastructure startup: the cloud providers it partners with are also its competitors, and the model providers keep making their own serving stacks better. Its defense is focus. It does one job, running models efficiently in production, and it does it across clouds rather than inside one vendor's walled garden. If AI spending keeps shifting from training to serving, as most industry watchers expect, that focus is well placed. The Series B gives it the runway to prove it at a bigger scale, with Nvidia watching from the cap table.