Kandou AI Raises $225M in Series A Funding

Why It Matters

By pushing copper interconnects toward Shannon‑limit performance, Kandou could delay costly migration to optical links, lowering AI infrastructure expenses and enabling faster scaling of generative‑AI workloads.

Key Takeaways

• $225M Series A led by Maverick Silicon.
• Copper MIMO boosts bandwidth, cuts power consumption.
• Claims up to 12× cost reduction, 10× reach improvement.
• Analog design avoids complex DSP and silicon interposers.
• Targets AI data‑center interconnects beyond 448 Gb/s.

Pulse Analysis

The AI boom has exposed a hidden bottleneck: moving terabytes of data between processors and memory fast enough to keep up with ever‑larger models. Traditional copper links, once sufficient at 112 Gbps, now struggle as generative AI pushes required rates past 224 Gbps. While many vendors look to optical solutions, those introduce conversion latency and higher bill‑of‑materials. Kandou AI’s approach re‑imagines copper by applying information‑theoretic concepts, treating crosstalk as usable signal energy rather than noise, thereby extracting more bits per hertz without adding physical lanes.

At the heart of Kandou’s technology is Copper MIMO, a chord‑signaling scheme that leverages analog vector‑matrix transformations across multiple wires. By inverting the channel distortion at the transmitter and applying the inverse at the receiver, the system converts interference into constructive data, moving the link closer to the Shannon limit. The analog implementation sidesteps power‑hungry digital signal processing, delivering up to three‑fold power savings and eliminating the need for expensive silicon interposers. With roughly 500 patents protecting the method, the startup positions itself as the sole commercial player exploiting this paradigm in the wired interconnect market.

The $225 million Series A, backed by heavyweight investors, gives Kandou the runway to scale manufacturing, deepen ties with hyperscale data‑center operators, and expand its product portfolio beyond 448 Gb/s to multi‑terabit solutions. If the promised cost reductions—up to twelve times cheaper than competing designs—materialize, customers can defer costly optical upgrades and accelerate AI deployment timelines. As enterprises seek sovereign, low‑latency AI infrastructure, Kandou’s copper‑centric chips could become a strategic component, reshaping the economics of next‑generation compute fabrics.