Simplismart Raises $14M Series B for AI Inference Software

Cloud providers can put their own names on an AI service built with Simplismart's software. That arrangement gives the Bengaluru-rooted inference company a customer beyond the application developer: the organization that owns computing capacity and wants to turn it into a service somebody will keep buying.

Simplismart's $14M Series B, announced October 27, 2026, finances that ambition. Dallas Venture Capital led the round, joined by NVIDIA, Accel, Shastra VC and Micromax. Co-founders Amritanshu Jain, CEO, and Devansh Ghatak, CTO, are building a business around the software between an AI model and the hardware doing its work.

The distinction matters to both sides of the transaction. An application company wants responses that arrive within its users' tolerance and its operating budget. A cloud provider needs to turn installed GPU capacity into a dependable product, carrying the customer relationship under its own brand. Simplismart's opportunity sits inside the engineering required to satisfy those promises together.

What Simplismart's Series B finances

The company identifies deeper development of its inference engine and orchestration platform, expanded cloud partnerships and international growth as priorities for the capital. North America is a particular focus. Those plans extend a software strategy the founders have pursued since starting Simplismart in 2022, rather than introducing a newly announced GPU-fleet ownership business.

The financing follows a $7M Series A led by Accel, announced October 17, 2024. In that earlier letter, Amritanshu Jain described working on infrastructure at Oracle Cloud while Devansh Ghatak worked on search algorithms at Google. The founders' frustration with deploying machine-learning systems became the starting point for the company. The present round takes that operating problem into a broader set of customer relationships.

Dallas Venture Capital now leads an investor group that includes an important hardware supplier in NVIDIA. Simplismart says its engineers work directly with NVIDIA solutions architects and that NVIDIA helps facilitate industry connections. That gives the financing a technical and commercial relationship alongside the investment, while leaving individual check sizes and ownership stakes undisclosed in the company's announcement.

Simplismart has also reported a 14x increase in annualized revenue run rate over 6 months. The figure is management's account of recent growth. Without an absolute revenue disclosure or a complete period-by-period series, it provides a sense of momentum while leaving the size and durability of that usage outside the public announcement.

How inference becomes an operating decision

Inference is the work an AI model performs when responding to a request. For the application team, its cost and timing affect the product the customer experiences. Spending less on a response can become expensive in another way if the user abandons the interaction while waiting for it. A workload running overnight can tolerate a different balance.

Simplismart's product addresses those differences through a configurable serving engine and orchestration software. Hardware selection, numerical precision, batching, caching and scheduling all influence the result. The engineering task is to find a configuration that fits the application and keep it useful as demand changes, while respecting the environment where the customer needs to run it.

An AWS case study makes part of that operational work visible. AWS describes Simplismart's use of Amazon EKS, EC2 Auto Scaling warm pools and customized pod autoscalers. The partner account reports scaling in roughly 60-70 seconds and infrastructure-cost reductions of up to 40% in the deployments described. Those are attributed case-study outcomes, rather than a promised result for every model or customer.

For a platform leader, this is a purchasing question about who will own a continuing engineering job. Models, usage patterns and deployment choices can change after the initial launch. Buying an inference service brings another organization into those decisions, so the useful comparison includes operational responsibility alongside the price of processing a request.

A cloud provider becomes a software customer

Simplismart says its software powers the serverless inference layer of Yotta's Shakti Studio. The relationship illustrates how a provider can sell access to AI models while Simplismart supplies software beneath the service. The provider keeps its customer-facing brand, with the inference platform contributing to the experience that brand has promised.

That division of work creates a different sales conversation from supplying an application endpoint alone. A cloud operator has infrastructure to put to work and customers expecting a usable service. The software supplier must fit into that provider's commercial offering, deployment environment and operating processes. Its customer is itself taking responsibility for other customers.

Deployment control belongs in that conversation early. Simplismart describes support for chosen clouds and private environments, including on-premises operation. An enterprise with a requirement about where its models run or data resides has already constrained the list of usable capacity. Available computing equipment elsewhere may never enter that buying decision, however attractive the price looks.

For the founders, serving both application teams and cloud providers can create a useful view of demand and supply. It also gives the company two sets of requirements to understand. Application teams judge the experience and expense of using AI; providers judge whether the software helps them operate and sell a service within their own commitments. Product development has to keep those perspectives connected.

Scheduling capacity within the customer's boundaries

Simplismart describes a further ambition to match workloads with suitable capacity across regions and time zones. The company presents that scheduling layer as work it is building toward. Treating it as an already completed global network would outrun the announcement and obscure the practical constraints the founders themselves identify.

Matching a workload to a machine includes the question of when it must respond and where it is allowed to execute. A provider planning capacity also has to consider whether the required model is ready when a request arrives. These details determine whether nominally available equipment can do something useful for the customer at that moment.

The Series B puts additional resources behind the engine, orchestration and partnerships through which Simplismart is pursuing that work. Its founders have chosen a position where success depends on fitting software into somebody else's hardware and somebody else's product promise. As cloud partnerships expand, more of Simplismart's engineering can reach an end user through a provider whose name is on the service, carrying the expectation that the next request will receive an answer.