800VDC architecture promises lower costs, higher efficiency, and better integration with solar energy
San Jose, California, 12 February 2026 – The explosive growth of artificial intelligence is transforming how data centers are designed and powered. As AI training clusters become more advanced, they now require between 50 kW and 100 kW per rack, and future systems may demand up to 200 kW per rack. Traditional AC-based electrical systems are struggling to keep up with these rising energy needs.
According to a recent white paper from Enteligent, legacy alternating current (AC) architectures are reaching both physical and economic limits. The company, which develops DC-to-DC power electronics, proposes an 800-volt direct current (800VDC) backbone as a more efficient solution for AI data centers.
For a typical 10-megawatt AI data center, the 800VDC model could save about $5.8 million in initial capital expenses. These savings mainly come from eliminating costly uninterruptible power supply systems and rack power distribution units. On top of that, annual operating expenses could drop by around $711,000 due to better energy efficiency and reduced cooling needs.
In simple terms, today’s solar-powered data centers waste energy by converting electricity multiple times. Solar panels generate DC power, which is converted to AC for distribution and then changed back to DC before it reaches servers. Each conversion leads to energy loss.
The 800VDC design allows solar panels and battery storage to connect directly to the data center’s power system. By removing two major conversion steps, energy losses can be reduced by 8% to 10% compared to traditional AC systems.
Efficiency is one of the biggest advantages. Standard AC setups often operate at around 78% end-to-end efficiency. The 800VDC system can reach 94% to 95% efficiency. This improvement not only lowers electricity use but also reduces heat generation, which means cooling systems do not have to work as hard. For renewable energy operators, it ensures more clean energy powers AI computing instead of being wasted as heat.
Another key benefit is the reduction in copper usage. High-density AI racks running at 240VAC require very high currents, often 400 amps or more, and thick copper cables. At 800VDC, the current is reduced by about 70%. A 100 kW load requires only 125 amps, allowing for smaller cables. This can cut copper mass by 50% to 80% and lower installation costs significantly.
Reliability also improves. Traditional server power supplies often fail due to high heat and complex components. The 800VDC architecture replaces them with shared, high-efficiency DC power shelves at the rack level. These units are hot-swappable, meaning they can be replaced without shutting down the system.
As hyperscale data centers expand to meet AI computing demand, direct-DC power systems are gaining attention for their scalability and sustainability advantages. By integrating solar energy and battery storage directly into a high-voltage DC backbone, data centers can reduce power costs, improve energy efficiency, and even support grid stability.
With AI workloads growing rapidly, high-voltage DC architecture could become a key foundation for next-generation data centers focused on clean energy, cost savings, and long-term scalability.
