Are data center and AI power ETFs expensive right now?

Updated September 11, 2026

How XLU's valuation compares with its own past

The line is XLU's trailing price-to-earnings ratio, one reading per month across 114 observations since May 2017 — every one of them our own calculation from the fund's holdings rather than a figure quoted from a provider.

All six funds, and where each sits in its own range

Value, quality and risk scores are ETF Copilot's own 0–10 measures computed across every fund we cover. Higher is better on all three, so a low value score means expensive. Where a column is highlighted, amber marks the least favourable reading of the group and green the most favourable.

The six funds are not interchangeable

These funds are bought for one story and built out of three different businesses: real estate that happens to be data centers, industrial equipment that happens to serve the grid, and regulated utilities that sell the electricity. They are priced the way their own industries are priced, which is why they do not move together.

GRID is at once the most expensive of the six, at 29.6 times earnings, and the weakest on the quality of the companies it holds — scoring 2.3 out of 10 where the strongest of the group scores 6.9. It holds 117 companies against XLU's 31, and reaching that much further down the industry is what the quality score is measuring. Paying more for less profitable companies is a real outcome, and no fund name signals it.

At the other end, SRVR is the only fund of the six that is cheaper than the group's midpoint on our value score, at 9.1 out of 10, without also sitting below it on company quality — 6.1 out of 10.

Trailing earnings are the harshest lens available

What would have to be true for today's reading to hold

Stated as conditions, because a condition is checkable and because nothing here says what to do about any fund:

• The earnings have to arrive. XLU's multiple falls from 19.0 to 16.5 only if the forecast earnings are delivered in full. Delivered halfway, it is around 17.8 times.
• The way each fund is built has to keep mattering. XLU holds 31 companies and scores 5.2 out of 10 on company quality; GRID spreads across 117 and scores 2.3. Those are two different bets wearing one label.
• The ride has to be tolerable. Our risk score runs 0 to 10, where 10 is the steadiest fund in the universe we cover. Across this group it runs from 3.4 for DTCR to 5.1 for SRVR, against 7.5 for VOO. The same funds produced the multiples above and those scores.

How this compares with the wider market

Data center and AI power ETFs are not the only funds trading away from their own past. QQQ is more expensive than 93% of its own record, at 30.8 times against a median of 25.1, and VOO is more expensive than 84% of its own record, at 25.2 times against a median of 21.5. The same measurement, applied to the other themes people ask about: semiconductor ETFs sit above 89% to 97% of their own records; AI ETFs sit above 4% to 95% of their own records; defense ETFs sit above 25% to 89% of their own records.

Is XLU expensive right now?

Against its own record, no on trailing earnings: XLU trades at 19.0 times against a median of 21.0, cheaper than 77% of its own 114 monthly readings, and it has been above that median in 11 of the last 24 months. Against the earnings expected next year the same fund is at 16.5 times, so both lenses point the same way and the page above shows both.

Which data center ETF is cheapest against its own history?

RSPU, cheaper than 100% of its own 40 monthly readings — the lowest of the six funds here relative to its own past. That is a different question from which fund is cheapest outright: on price alone the lowest multiple of the group is RSPU at 17.5 times. A fund can carry the lowest multiple in its group and still be expensive for itself.

Why do two data center ETFs show different P/E ratios?

Because two data center ETFs hold different companies in different proportions. A fund's price-to-earnings ratio is the weighted average of the ratios of what it holds, so a fund capped at the largest and most profitable names produces a different figure from one that equal-weights a much wider list. Providers also differ on trailing versus forward earnings, on how loss-making companies are handled, and on the date of the holdings file. Every figure on this page is computed the same way for every fund, from each fund's own published holdings.

Does a high P/E mean these funds will fall?

No. The ratio measures price against earnings already reported. It carries no information about what happens next, and reading a forecast into a measurement is the most common mistake made with it. What it does tell you is where a fund sits relative to what buyers have been willing to pay for the same fund in the past, which is a fact about the past and the present only.

Cite this
ETF Copilot, “Data center and AI power ETFs measured against their own price-to-earnings history”, figures as of September 11, 2026. https://etf-copilot.com/learn/are-data-center-etfs-expensive

Method. The price-to-earnings ratio is trailing twelve-month, computed by ETF Copilot from each fund's own published holdings rather than quoted from a data provider: every holding's ratio, weighted by its weight in the fund, across the whole portfolio and not the top ten. The monthly series is rebuilt from the holdings as they stood at each date, so the line is comparable with itself. “More expensive than” is the share of that fund's own monthly readings sitting below today's. History lengths differ by fund, from 31 monthly readings to 114, because the funds launched at different times; each percentile is measured against that fund's own record and no other. Value, quality and risk scores are ETF Copilot's own 0–10 measures computed across every fund we cover; higher is better on all three, so a low value score means expensive. Figures are as of September 11, 2026 and are recalculated daily.

See XLU in full, and any of the funds we cover →

Nothing on this page is investment advice, a recommendation, or a forecast of returns.