imec spin-off, Vertical Compute raises €20 million

The round was led by imec.xpand and supported by an investor base which includes Eurazeo, XAnge, Vector Gestion and imec. The funding will support Vertical Compute’s ambition to develop a novel vertical integrated memory and compute technology, unlocking a new generation of AI applications.

Vertical Compute says that its technology will have a "transformative impact, enabling next-generation applications with unparalleled efficiency and privacy". By minimising data movement and bringing large data closer to computation, the company claims that the innovation can ensure energy savings of up to 80%, unlock hyper-personalised AI solutions, and eliminate the need for remote data transfers, protecting user privacy.

“Memory technologies face limitations in both density and performance scaling, while processor performance continues to surge. The extreme data access requirements of AI workloads exacerbate this challenge, making it imperative to overcome the memory wall to enable the next wave of AI innovations. We believe going Vertical is the path to 100X gains,” says Sébastien Couet, CTO of Vertical Compute, in a press release.

Advancements in large language models and generative AI are reshaping industries at an unparalleled rate. Despite their potential, these AI models depend heavily on complex cloud infrastructure and high-bandwidth memory, leading to data transfer latency, high energy consumption, and the need to send sensitive data to remote servers. While edge computing offers a solution to these challenges, running large AI models directly on smartphones, PCs, or smart home devices presents significant hurdles related to cost, power efficiency, and scalability.

According to Vertical Compute, the core issue is the "memory wall." Static Random Access Memory (SRAM), integrated as caches of the CPU or GPU, is fast but small and costly. Dynamic Random-Access Memory (DRAM), the main memory in compute systems, is larger but expensive and energy-consuming. Both memory technologies are facing slowing advancements in density and performance, while processor speeds and market demands continue to rise, creating a significant bottleneck. This challenge is intensifying due to the growing demand for AI workloads, which require rapid access to massive amounts of data. Overcoming this memory wall is crucial for advancing AI inference.

The intersection of large-scale AI models and edge computing necessitates a fundamental transformation in data processing. Vertical Compute aims to seize this opportunity by developing chiplet-based solutions that leverage a new approach to storing bits in high-aspect-ratio vertical structures. The core of Vertical Compute’s patented technology, invented by Sébastien Couet, former Magnetic Program Director at imec, integrates vertical data lanes directly atop computation units.

It has the potential to outperform DRAM in terms of density, cost and energy, by reducing data movements from centimeters to nanometers. This promising technology, coupled with an ambitious commercialisation plan, has led to the creation of this new semiconductor venture.