Nvidia Backs Simplismart's $14 Million Round: Inside the Indian Startup Making AI Run on Fewer GPUs

Posted on 27th Oct 2026 11:03:28 in Artificial Intelligence, Machine Learning

Tagged as: Nvidia, Simplismart, AI inference, GPU costs, Indian startups, Series B

Nvidia has taken a stake in Bengaluru-based Simplismart, joining a $14 million all-primary Series B round that values the four-year-old AI infrastructure startup at roughly $100 million including the fresh capital. The investment, reported by TechCrunch on October 27, 2026, is a bet on the least glamorous layer of the artificial intelligence stack: the software that decides how a trained model actually runs once real users start hitting it.

The round was led by Dallas Venture Capital, with Nvidia participating alongside Accel, Shastra VC and Micromax. For Indian businesses watching the AI bill climb, the deal carries a more practical signal than the headline number. It suggests the next round of value in enterprise AI will not come from bigger models or bigger GPU clusters, but from squeezing more useful work out of the hardware companies have already bought.

What Simplismart Actually Sells

Simplismart does not build foundation models. It builds the orchestration layer that sits between a model and the computing infrastructure underneath it, adjusting how that model is deployed and served so it meets a customer's specific speed, throughput and cost targets.

In practice, the platform handles four jobs. It deploys models, either through Simplismart's own managed platform or inside a customer's existing cloud, on-premises estate or air-gapped environment. It fine-tunes models on customer data. It scales computing resources automatically as traffic rises and falls. And it monitors performance so engineering teams can see where latency and spend are leaking.

There is a fifth line of business that matters for India specifically: Simplismart lets cloud providers build and operate their own AI inference services on GPU infrastructure using its software. That turns the company from a vendor into a supplier to the very data centre operators competing for the same enterprise workloads.

Co-founder and CEO Amritanshu Jain frames the product around a blunt observation. "Nobody actually needs the fastest or the cheapest inference," he told TechCrunch. What companies need, he argued, is to run models at speeds and costs that make sense for their own application, rather than at whichever extreme a vendor happens to optimise for.

Inside the $14 Million Round

The funding is all primary, meaning the money goes into the company rather than buying out early shareholders. Dallas Venture Capital led, with Nvidia, Accel, Shastra VC and Micromax participating. The round values Simplismart at about $100 million post-money.

The founding story is the familiar Bengaluru engineering pipeline. Jain, a former Oracle engineer, and CTO Devansh Ghatak, a former Google engineer, met while studying at BITS Pilani and started the company in 2022 out of their experience building and deploying machine learning systems. The startup now employs 58 people.

There is a valuation gap worth noting. The Economic Times had reported that Simplismart was in talks to raise $40 million to $50 million at a valuation of $300 million to $400 million. Asked about that report, Jain declined to comment, saying the company was not actively fundraising and that the Series B was closed and would not be extended. The gap between the reported talks and the closed round is a useful reminder that leaked term sheets are not signed term sheets, and that 2026's AI funding headlines routinely run ahead of the actual wire.

Why Nvidia Is Funding Software That Saves GPUs

The investment looks like a paradox. Nvidia is the dominant supplier of AI accelerators, and Simplismart sells software that, in some cases, lets customers run the same workload on fewer of them. Why would the chip seller fund the efficiency layer?

Jain's answer is that cheaper inference expands the market rather than shrinking it. If serving a model costs less per request, more applications become economically viable, more requests get served, and total demand for compute rises even as each individual workload becomes leaner. That is the classic Jevons argument applied to tokens, and Nvidia has been making versions of it for years.

The relationship is already deeper than a cheque. Nvidia has been working with Simplismart on technical integrations, including around its Dynamo inference software, with Simplismart's engineers collaborating directly with Nvidia solutions architects. Nvidia has also introduced the startup to potential customers and partners. For a company of Simplismart's size, that distribution help is arguably worth more than the capital.

The Numbers Behind the Pitch

Simplismart's growth claim is aggressive. Jain said the company's annualised revenue run rate grew 14-fold over the past six months to an eight-figure sum. That figure is based on recent customer usage rather than contracted recurring revenue, and it excludes revenue passed through to infrastructure providers. He did not disclose the exact number.

More than half of Simplismart's revenue comes from the United States, even though it has more customers in India. That split tells you where the willingness to pay for inference optimisation currently sits, and it explains why the new funding is earmarked for expanding US presence, hiring, and further product development.

Customer evidence is more concrete than the revenue claim. Simplismart says it helped AI video-generation platform InVideo cut GPU serving costs by 56 percent. Its customer list includes AI video companies HeyGen and Higgsfield, and enterprise AI startup Ema. An AWS case study published on the company describes infrastructure cost reductions of up to 40 percent, an eight-fold increase in GPU hours deployed, and six-fold business growth within six months, with warm-pool autoscaling bringing scale-up time down to 60 to 70 seconds.

The company's own published figures claim peak GPU usage cut by more than half and roughly 2.5 times more data processed at half the cost. Those are vendor numbers and should be read as such, but they point consistently in one direction: the savings come from scheduling and serving discipline, not from exotic hardware.

From Bengaluru to Yotta's Shakti Studio

The most instructive Indian deployment is Yotta Data Services. The Mumbai-based cloud provider uses Simplismart's technology to power the serverless inference service inside its Shakti Studio platform, a browser-based environment that gives developers access to serverless GPUs, production AI endpoints and fine-tuning workflows covering SFT, GRPO, DPO, RFT, LoRA and QLoRA.

Shakti Studio bills on a pay-as-you-go basis, either by token or by GPU-minute, and supports PyTorch, TensorFlow, ONNX, Hugging Face and Nvidia NeMo. For Indian enterprises with data residency or compliance constraints, the combination of domestically hosted GPU capacity and an orchestration layer built by an Indian team is a meaningful alternative to routing inference through a US hyperscaler.

That is the strategic logic behind a sovereign AI cloud: the GPU capacity is necessary but not sufficient, because raw silicon without a serving layer is an expensive space heater. Simplismart's role in Shakti Studio is precisely that serving layer, and it is the kind of unglamorous plumbing that determines whether Indian AI infrastructure is actually usable by mid-market companies.

What This Means for Indian Businesses Running AI

For an Indian SME or mid-market firm experimenting with AI, three practical takeaways stand out.

First, inference is now the dominant cost, not training. Most companies will never train a frontier model. Nearly all of them will pay to serve one, request after request, and that recurring bill is where margins quietly disappear. Optimisation work on the serving layer therefore has a faster payback than almost any other AI investment.

Second, buying GPUs is not the only route. Serverless endpoints, warm pools and autoscaling mean a company can start with pay-as-you-go capacity and only commit to dedicated hardware once usage patterns are stable. The 60-second scale-up figures matter most to businesses whose traffic is spiky, such as e-commerce during festive sales or a services firm handling seasonal document volumes.

Third, the layer is becoming crowded and consolidating. Simplismart competes with Nvidia's own inference offerings, with Amazon Web Services, Google Cloud and Microsoft Azure, and with a growing field of inference startups. That competition is good news for buyers: pricing pressure on the serving layer is one of the few places where AI costs are genuinely falling rather than rising.

The Crowded Road Ahead

Simplismart's stated edge is not raw speed but fit. Jain argues the company wins where customers face stringent service-level agreements and tight total cost of ownership targets, because it tunes the balance per workload rather than offering a preset faster or cheaper tier. That is a defensible position, but it is also the position every serious inference vendor claims.

The harder question is durability. Inference optimisation is software, and software advantages erode as open-source serving engines improve and hyperscalers bundle similar capabilities into their platforms. Nvidia's involvement cuts both ways here: it validates the company and opens enterprise doors, while also placing a strategic investor with its own inference products on the cap table.

For now, the $14 million buys Simplismart runway, a US push and a stronger engineering bench. For the Indian AI ecosystem, it marks something less visible but more consequential than another funding announcement: the emergence of a domestic inference infrastructure layer, built by Indian engineers, backed by the world's most important chip company, and increasingly likely to sit underneath the AI products Indian businesses actually use.

Sources

• TechCrunch — Nvidia backs Indian startup helping companies run AI models with fewer GPUs
• Amazon Web Services — Simplismart.ai Scales Generative AI Workloads with Faster Inference and 40% Lower Infrastructure Costs
• Simplismart — What Is an AI Inference Platform? A Complete Guide for 2026
• Yotta Shakti Cloud — Shakti Studio: India's AI Inference and Fine Tuning Platform
• Dallas Venture Capital — Simplismart portfolio profile