According to the report, Vertical Semiconductor — which emerged from research at Massachusetts Institute of Technology (MIT) — has developed a novel chip architecture using gallium nitride (GaN) instead of traditional silicon, stacking transistors vertically rather than horizontally.

This innovation promises to reduce heat and inefficiency in voltage conversion — a step-change advantage for AI data centres which currently waste large amounts of energy converting high-voltage input into chip-ready voltages. Vertical plans to ship prototypes this year and scale commercial deployment next year.

Key Highlights

• Vertical Semiconductor raises $11 million for its GaN power-chip tech.

• Switches to gallium nitride and vertical transistor stacking for efficiency gains.

• Targets large AI data-centre operators facing challenges with energy and heat waste.

• Potential to disrupt established players in the power-electronics market (e.g., silicon-based suppliers).

Why This Matters

1. Energy bottleneck in AI: As model sizes and compute demands grow, so does the energy and heat problem — solutions like these tackle a foundational layer, not just the model.

2. Hardware layer diversification: Moving beyond CPUs, GPUs, and networking, we’re now looking at power-electronics innovation as a competitive frontier.

3. Commercial scale in sight: With prototypes this year and full deployment next year, this is more than lab-theory — it’s near-term infrastructure change.

4. Investment signal: Even modest VC rounds (here $11 m) in deep-tech show investor confidence in the “hidden layers” of AI infrastructure.

Source
Reuters – Full Article