We solve the 'power problem' AI data centers are facing: Onsemi CEO

Key Concepts

• Silicon Carbide (SiC): A wide-bandgap semiconductor material that is more efficient at handling high power and heat than traditional silicon.
• Power Tree: The entire infrastructure of power conversion and distribution, from the electrical grid into the data center and down to the individual chips.
• Compute-Power Paradox: The observation that as AI models (CPUs/GPUs) become more powerful, they require exponentially more electricity, creating a bottleneck in data center expansion.
• Supply Chain Resilience: The strategy of localizing manufacturing (specifically in North America) to mitigate geopolitical risks and logistics disruptions.

---

1. Main Topics and Key Points

The interview focuses on ON Semiconductor (onsemi) and its role in the AI infrastructure boom. While companies like NVIDIA and AMD focus on the "compute" side (GPUs/CPUs), onsemi focuses on the "power" side.

• Performance: onsemi reported a strong first quarter with top and bottom-line beats, following an 84.8% stock gain in 2026.
• The Power Bottleneck: CEO Hassane El-Khoury argues that the current AI boom is not a bubble but a supply-constrained market. There is a shortage of compute, a shortage of chips, and a critical shortage of power.
• Efficiency: onsemi’s chips are capable of reducing power loss in data centers by 50% compared to traditional silicon chips.

2. Real-World Applications

• Data Center Optimization: onsemi’s technology is used to manage the "power tree" within data centers, ensuring that the massive energy requirements of AI training are met with minimal waste.
• Grid Infrastructure: Beyond the data center walls, the company provides solutions for energy storage systems to supplement the grid while long-term infrastructure upgrades (like transformers) are pending.

3. Methodologies and Frameworks

• The "Power Tree" Strategy: onsemi manages power conversion at every stage—from the grid, into the data center, and down to the server level.
• Hybrid Manufacturing Strategy: To ensure supply chain security, onsemi utilizes a hybrid model, producing 8-inch silicon carbide substrates in North America. This reduces reliance on foreign foundries and mitigates risks associated with U.S. Commerce Department export restrictions.

4. Key Arguments and Perspectives

• AI is Supply-Constrained: El-Khoury rejects the "AI bubble" narrative, asserting that demand is growing faster than the industry can build the necessary infrastructure.
• Discipline in Power Usage: Because building new power grid capacity takes years, the industry must focus on "disciplined" power usage. El-Khoury notes that a 1% improvement in efficiency can power a million homes or allow a data center to increase compute capacity without requiring additional grid input.
• Strategic Positioning: By focusing on power efficiency, onsemi positions itself as a necessary partner to the compute giants (NVIDIA/AMD) rather than a competitor.

5. Notable Quotes

• "United States is short power. We're short compute, we're short chips, there will be shortages in all three." — Hassane El-Khoury
• "I would argue 1% improvement can power a million homes for a year, or can increase the compute power of an existing data center without adding more power as a pipeline into it." — Hassane El-Khoury

6. Supply Chain and Geopolitical Resilience

Addressing concerns regarding U.S. export restrictions on Chinese chipmakers (specifically Hua Hong), El-Khoury stated there is no impact on onsemi. He emphasized that the company learned from the COVID-19 pandemic and invested heavily in a North American-based footprint, making them the only U.S.-based power semiconductor company capable of managing the entire "power tree."

7. Synthesis and Conclusion

The core takeaway is that the AI revolution is currently hitting a physical limit: the availability of electricity. While the market is fixated on the chips that perform calculations (GPUs), the underlying infrastructure—specifically power management—is the true bottleneck. onsemi has successfully pivoted to address this by utilizing Silicon Carbide technology to maximize energy efficiency. By localizing their supply chain in North America, they have insulated themselves from geopolitical volatility, positioning the company as a critical, stable player in the broader AI hardware ecosystem.