€259 million for KANDOU: the invisible market set to be worth hundreds of billions

In data centers training the most advanced artificial intelligence models, the challenge is no longer just about computation, but about moving ever-growing volumes of data between GPUs, memory, and internal networks. As models increase in size and complexity, this data movement is becoming the primary determinant of performance.

This is precisely where Kandou AI positions itself. The Lausanne-based company has just raised approximately €259 million (£225 million) in a strategic round led by Maverick Silicon, alongside SoftBank Group, Synopsys, Cadence Design Systems, and Alchip Technologies.

The shift: from compute to data movement

For more than a decade, computing performance has been largely framed as a function of raw processing power. The rise of artificial intelligence extended this paradigm, with investments heavily concentrated on GPUs and specialized accelerators. But this equation is now reaching its limits.

In modern AI clusters, components must continuously exchange volumes of data that are growing faster than compute capacity itself. Memory is expanding, architectures are becoming more distributed, and models now rely on thousands of chips. In this context, interconnect bandwidth and latency become critical constraints.

Overall performance is no longer solely determined by available compute power, but by the ability to efficiently orchestrate data flows between components. As Kandou puts it, this marks a paradigm shift: “GPUs and CPUs must be connected to exponentially growing memory volumes, making memory bandwidth and interconnects major bottlenecks.”

A discreet yet foundational market

This trend is bringing renewed attention to the high-speed interconnect market, long operating in the background. It encompasses the technologies that enable data transfer between chips, racks, and systems—SerDes (serializer/deserializer), retimers, and copper or optical interfaces.

These components are rarely visible and are not end-user products, but they form essential intermediate layers. Their importance is increasing as system architectures become more complex.

The challenge is twofold:

• increasing bandwidth to handle growing data flows
• reducing energy consumption as data centers approach physical and economic limits

In this context, Kandou advocates an approach based on reinventing copper interconnects, leveraging signaling technologies derived from information theory and semiconductor design. The goal is to reach performance levels comparable to certain optical solutions while maintaining more favorable cost and energy efficiency.

The copper bet versus the rise of optics

One of the key trade-offs in AI infrastructure today lies between copper and optical interconnects.

Optical solutions deliver high performance over long distances but remain costly and energy-intensive at scale. Copper, historically more limited, offers lower costs and easier integration. The innovation challenge, therefore, lies in pushing its physical limits further.

Kandou is aligned with this approach. “Our goal is to transform AI hardware through products built on our copper interconnects and system IP,” explains Srujan Linga, co-founder and CEO of the company. “Our first products will target AI systems and high-speed rack-level connectivity beyond 448G, with the ambition to progressively cover the entire infrastructure stack.”

A cap table reflecting systemic stakes

The investor base combines profiles across the semiconductor value chain. Alongside Maverick Silicon, Synopsys and Cadence Design Systems are central to chip design tools. Alchip Technologies operates in custom chip design and manufacturing. SoftBank Group, for its part, continues to build positions across critical AI infrastructure.

AI’s physical infrastructure becomes a new investment frontier

Beyond Kandou, this round reflects a broader trend: increasing investment in infrastructure and industrial layers of AI (energy, cooling, interconnects, and memory).

While such deals remain relatively rare in Europe, several funds are beginning to position themselves in these areas. Firms such as Atlantic Bridge, IQ Capital, and Vsquared Ventures have developed theses around semiconductors, compute architectures, and the physical constraints of AI. In France, activity remains more limited, with players such as Supernova Invest. At this stage, however, few European funds have the financial and organizational capacity to independently lead multi-hundred-million-euro rounds in such capital-intensive technologies. As a result, AI infrastructure financing remains largely structured around international coalitions, where European investors still play more of a supporting than leading role.

Over $500 million raised to structure a breakthrough technology

Since its founding in 2011 by Srujan Linga and Dr. Amin Shokrollahi, Kandou AI has followed a progressive funding trajectory typical of capital-intensive deeptech companies. After an initial pre-seed backed by Venture Kick, the company raised approximately $10 million as early as 2012, followed by multiple growth rounds led notably by Bessemer Venture Partners. The recently announced strategic round of $225 million (approximately €209 million) marks a clear step change, both in scale and in the profile of its investors. It brings Kandou’s total funding to over $500 million.