With the development of artificial intelligence technology, the efficiency of data processing and energy consumption management have become the key factors restricting its further development. Faced with this challenge, Vertical Compute is opening up new possibilities for the next generation of AI applications through its innovative vertically integrated memory and compute technologies. Vertical Compute recently announced the successful completion of a €20 million seed investment round from imec.xpand, Eurazeo, XAnge, Vector Gestion and imec. The funding will support the company's revolutionary breakthroughs in in-memory technology to advance AI and edge computing applications.
Memory technology bottlenecks: Challenges and opportunities coexist
Over the past few decades, advances in computer memory technology have not kept pace with improvements in processor performance. This phenomenon is known as the "memory wall". While traditional static random-access memory (SRAM) is fast, small in capacity, and costly, dynamic random-access memory (DRAM) can provide more storage space, but its energy consumption and cost make it challenging for large-scale applications. With the popularity of applications such as artificial intelligence and generative AI, these issues have become more prominent, especially in large-scale data processing scenarios that require high bandwidth and fast access.
Vertical Compute's technology is designed to break this bottleneck. Sébastien Couet, the company's chief technology officer, said: "Memory technology has reached a bottleneck in terms of density and performance, while processor performance is constantly improving. Faced with the extreme data access demands of AI workloads, we must break through the memory wall to unlock the next wave of AI innovation. ”
Vertical integration technology: Fundamentally solve the memory bottleneck
In order to cope with the bottleneck of memory technology, Vertical Compute proposes a new idea based on vertical integration. This innovative vertical memory and computing structure greatly optimizes data transmission by combining memory with computing units to reduce the distance that data travels inside the chip. The high energy consumption and high latency problems faced by traditional memory technologies have been effectively alleviated in the vertical structure.
Vertical Compute's core patented technology tightly integrates data storage and computing units through the design of vertical data channels. By reducing the distance of data transmission, it is reduced from centimeters to nanometers, which greatly improves the connection efficiency between memory and computing units. This not only significantly reduces energy consumption, but also improves system performance, helping AI applications run efficiently on edge devices without relying on cloud infrastructure.
"Our approach addresses the performance limitations of current memory technologies by consolidating data storage and compute into a more efficient vertical structure," said Sylvain Dubois, CEO of the company.”
Figure: Vertical Compute: Breaking through memory bottlenecks and driving the future of AI applications (Source: IMEC)
Ushering in a new era of AI and edge computing
With the rise of generative AI, natural language processing, and large-scale machine learning models, traditional cloud computing architectures have struggled to meet their requirements for bandwidth, latency, and privacy protection. In particular, AI inference on edge devices such as smartphones, PCs, and smart homes presents multiple challenges such as cost, energy consumption, and scalability.
Vertical Compute's vertically integrated memory technology is designed to address this challenge. Not only does it reduce latency and power consumption for data transmission, but it also protects user privacy and avoids the need to transfer sensitive data to a remote server. This allows AI applications to be processed on local devices, improving the user experience and effectively solving the privacy concerns of large-scale AI applications.
Kickstart R&D and team building, the future of vertical computing
Vertical Compute is headquartered in Lowanranho, Belgium, and has R&D centers in Leuven, Belgium, Grenoble, and Nice. The company is actively recruiting top engineers to support its R&D efforts in vertically integrated memory and computing technologies and accelerate the commercialization of its products.
"We want to attract top talent from all over Europe and bring Europe's technological advantages to the global stage," said the company's founder.”
The disruptive potential of vertical computing
The successful seed round not only reflects investors' high recognition of the potential of Vertical Compute's technology, but also demonstrates the great confidence of the global capital market in the company's future development. "We have great confidence in the leadership of the Vertical Compute team and this disruptive technology, and we look forward to working with them to drive change in the semiconductor industry," said Tom Vanhoutte of imec.xpand. ”
