Free tools Windows power users keep installed
One-click scans. No signup required.
PUE remains useful for measuring data-center facility overhead, but it does not account for electricity lost inside a server as power is converted and regulated before reaching an AI processor. In an October 2, 2026, Electronic Design article, Hans Hasselby-Andersen, CEO of Lotus Microsystems, argues that operators should pair PUE with stage-by-stage reporting on server power delivery—not treat PUE as a complete measure of AI infrastructure efficiency.
What PUE measures—and where its boundary ends
Power usage effectiveness, or PUE, is the ratio of total facility energy to energy delivered to IT equipment. It captures facility overhead such as cooling, lighting and power distribution. A lower PUE indicates less facility energy spent beyond the energy delivered to IT equipment; it does not describe every loss after power enters a server. The basic ratio is also described in Electronic Design’s PUE explainer.
As an Amazon Associate I earn from qualifying purchases.
In Hasselby-Andersen’s account, PUE’s boundary stops at the server threshold. Electricity is still converted and regulated inside the server before it reaches a processor, and those internal conversion losses are not separately measured by facility PUE. That is a boundary distinction, not a claim that PUE is defective: it remains a facility-level metric, but it cannot answer every question about server-level efficiency.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
How power reaches an AI processor
A simplified server power path begins with AC input. A power supply converts that input to a high-voltage DC bus; one or more DC-DC stages reduce the voltage, and a point-of-load (POL) converter regulates power near the processor. Actual architectures vary, so this sequence should not be read as a universal schematic.
Each conversion stage can lose some energy as heat. The losses increase the power that must ultimately be removed as heat, but facility PUE does not identify them as a separate server-internal category. PUE describes the facility-to-IT relationship; stage efficiency describes how input power changes as it is converted and regulated within the server. The two measures answer different questions.
Why the distinction matters more at AI rack power levels
Hasselby-Andersen argues that higher rack power makes percentage losses more consequential in absolute terms, while accelerator load transients place additional demands on power electronics. The Electronic Design article reports the following rack-power figures, but does not independently validate them or provide test data quantifying conversion losses:
| Rack-power context | Figure reported by the article |
|---|---|
| Typical data-center rack about five years before publication | Roughly 5–8 kW per rack |
| Current AI-facility designs, as described in the October 2, 2026 article | 15–50 kW per rack |
| GPU-dense configurations, as described in the article | More than 100 kW per rack |
These are the author’s reported ranges, not independently established industry averages. Rack power varies with hardware generation, workload and deployment. The article’s figures illustrate why operators may want to examine conversion stages; they do not establish how much energy a particular server loses or which design is more efficient.
Keep PUE, and add stage-level questions
Hasselby-Andersen’s proposal is to pair PUE with reporting on power-delivery efficiency at individual conversion stages, especially close to the processor. This is a proposed way to broaden efficiency discussions, not an industry-wide metric or an adopted test protocol. His point is not that PUE should be discarded: facility overhead and server-internal conversion efficiency are complementary concerns, not interchangeable scores.
For operators comparing vendors or architectures, ask how each reported efficiency figure was obtained. Useful questions include:
- What are the input and output measurement boundaries for each conversion stage?
- At what load, voltage and current was efficiency measured, and what transient profile was applied?
- Was the figure measured instantaneously, averaged over a workload, or rated at a specified operating point?
- How does the point-of-load stage behave under the processor’s relevant operating conditions?
These questions extend the article’s stage-by-stage recommendation into practical comparison criteria; the article itself does not prescribe a complete measurement protocol. Comparisons are meaningful only when the boundaries and operating conditions are sufficiently alike. Without comparable measurements, a single efficiency figure cannot support a reliable product ranking.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to interpret the product example
The article names vStrata, a vertical power module intended for ultra-high-current AI accelerators, as an example from Lotus Microsystems. Hasselby-Andersen is the company’s CEO, so the example is a vendor’s product reference—not independent evidence of performance, savings or compatibility. The article does not establish those results for the product.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesQuick Recap
Best Value
- Family farms not data design for people against AI server farms, data center expansion, rural land buyouts, corporate agriculture, and industrial tech development replacing farmland and open space. Rural conservation and anti data center message.
- AI protest design for farmers, land conservation supporters, anti AI activists, sustainability groups, environmental advocates, rural communities, and people opposing server farm construction, power grid strain, and farmland destruction.
- Lightweight, Classic fit, Double-needle sleeve and bottom hem
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
