Vertical Compute, a Belgian semiconductor startup focused on optimizing AI chip architecture, has raised €20 million in an all-equity seed round just three months after its launch. Co-founder Sylvain Dubois is a former Google semiconductor strategist while Sebastien Couet served imec, for over ten years as its leader in R&D in the semiconductor field and hopes to bridge memory bottlenecks that restrain the scalability of AI.
The funding round was led by imec.xpand, the venture capital arm of R&D hub imec, with participation from French VCs Eurazeo and XAnge, and Swiss investor Vector Gestion. The acquired funds will be used to develop its innovative chip architecture, which integrates memory directly into processors, eliminating the need for costly and space-intensive high-bandwidth memories (HBMs).
Tackling the Memory Bottleneck:
The current AI chips like Nvidia GPUs depend on memory HBMs coming primarily from Samsung, Micron, and Hynix. It is sold for around $10,000 a chip. All this brings along high cost and energy use as well as reduces scalability. Memory in the Vertical Compute’s solution will be inside the chip, eliminating centimeters-level data transfer that turns into nanometers, boosting performance, reducing cost, and also decreasing energy use.
“The data is immediately available without back-and-forth movements,” explained Dubois. “This innovation drastically reduces costs, and infrastructure needs while enhancing energy efficiency.”
Boosting Edge Computing:
The technology developed by Vertical Compute is likely to revolutionize edge computing as it enables the execution of AI computations locally on devices such as smartphones and laptops, rather than on cloud data centers. This promises to reduce latency, lower energy consumption, and increase data privacy—important factors as AI models grow in size and complexity.
“It’s not sustainable to depend solely on cloud infrastructure operated by giants like Google, Amazon, and Microsoft,” said Dubois. “We need localized, energy-efficient alternatives.”
Though not designed for training AI models locally, the chips will be crucial for inference—the application of trained AI models to new data.
Strategic Collaborations and Growth Plans:
Vertical Compute is not going to manufacture complete chips but will design memory layers create test products in-house and then outsource the production to global foundries such as TSMC. The company will sell its technology to large semiconductor companies such as Nvidia, AMD, and Intel for a competitive advantage in AI memory solutions.
Despite its global aspirations, Vertical Compute remains a European company at heart.
“Europe needs companies like this,” Dubois said. “We are building a European company with European talent to foster innovation locally.”
The company will hire 30 new people in Belgium and France within a year. It will open a French subsidiary shortly, tapping into the semiconductor hubs of Grenoble and Nice.
Talent Repatriation:
Dubois, who has been back in the U.S. for six months, is keen to tap into European talent while repatriating experts from Silicon Valley.
“I have seen the exodus to California and the return since Covid. Repatriating talent is a priority—there’s still untapped potential in the U.S. to bring back.”
Addressing a major industry challenge and stimulating European innovation, Vertical Compute puts itself at the heart of AI semiconductor technology for the future.
If you’re as excited about DeFi, blockchain, and the evolving Web3 universe as we are, join our community! Subscribe to our newsletter for the latest updates, trends, and insights—let’s navigate the world of Web 3 together!