It is no secret to anyone paying attention that Generative AI has experienced an unprecedented growth in adoption over the past 15 months. Almost overnight, it seems nearly every company has accelerated the integration of AI into their product or operations roadmaps. 

But as AI applications have proliferated, so have a new set of challenges in managing the associated demands of model complexity and computing power, which are quickly restricting the ability of AI to scale for broad use across various markets. 

In order for AI applications to expand at a level which is financially and environmentally sustainable, AI inference—the process of using live data in trained models for predictions or tasks—must find a way to be more power-efficient than current solutions. Even a single ChatGPT query can generate more than 100x more CO2 than a normal Google search. The need for solutions that directly address the key challenges in AI inference processing—compute capability, scalability, accuracy, and energy savings—is more urgent than ever. 

‍

With a proven AI inference system that can deliver 10x higher compute density, along with 10x power efficiency, and 13x lower cost per query, Recogni is expanding its solutions to address GenAI in addition to autonomous driving applications. 

‍

Enter Recogni, an AI computing innovator positioned to raise the bar on power performance and address the large compute needs of AI workloads. Recogni has just announced their Series C funding round of $102 million, which Celesta is delighted to co-lead along with GreatPoint Ventures. 

Celesta first invested in Recogni in 2019, and the company has made significant progress in that time as a leading developer of AI chips and systems. Focused the past few years on intelligent autonomy for mobility markets, Recogni introduced Recogni Scorpio in late 2022, its first low power AI compute product, and has since captured a number of customer wins. Scorpio is the world’s first 1000 TFLOPS-class inference solution at the lowest power, validating the company’s ability to deliver scalable performance at low power for a range of AI inference applications. 

With a proven AI inference system that can deliver 10x higher compute density, along with 10x power efficiency, and 13x lower cost per query, Recogni is expanding its solutions to address GenAI in addition to autonomous driving applications. 

From its inception, Recogni has focused on high compute and low power solutions from convolutional neural networks to transformers, with an architecture that is well suited to address the high compute demand of large language models (LLMs), ChatGPT, and other generative models. As market forces have converged over the past year to accelerate growth in the GenAI space, it has become increasingly clear that Recogni is incredibly well positioned to deliver solutions purpose-built for this new range of applications. Rather than an aspirational pivot into a new market (as many companies are doing now) this expansion is about serving an identified market need that the company’s technology is already well designed to address, which happens to align with a projected $400B+ TAM by 2030. 

Today’s cloud-based AI training and inference solutions predominantly utilize power-intensive GPUs (graphics processing units), exerting substantial stress on the compute capacity, cooling, and power systems of data centers. The proprietary design of Recogni’s universal inference engine delivers a fundamental change in inference processing that solves the AI compute issue from the ground up. By dissecting the compute efficiency bottlenecks, their team has managed to uniquely solve the fundamental problem of high compute at low power (which are immensely difficult to deliver in tandem), along with scalability that can stand the test of time as model sizes and applications continue to grow exponentially. 

Recogni’s technology perfectly embodies a core tenet of our investment thesis: foundational innovation of systems-level infrastructure that will enable the mass adoption of emerging tech in the years to come. 

On behalf of the entire Celesta team, we’re pleased to support Recogni’s ability to meet this stunning market demand with its innovative technology. We look forward to continuing to work alongside their fantastic team as they expand the company’s exciting product offering.