Report 168 · Energy Storage
In September, Donut Lab published a report from Finland's national research lab, VTT, putting its much-disputed battery cell at 409.3 watt-hours per kilogram. The number is a real measurement. The report is also only seven pages long, and reading all seven tells you exactly how far the number goes.
Some background. Donut Lab, a Finnish company, launched what it calls a solid-state battery at CES in January 2026. Electrek, which has followed the story closely, lists the headline claims as 400 Wh/kg of energy density and a design life of 100,000 cycles, alongside very fast charging. Since then the company has published a series of commissioned test reports on a site it calls "I Donut Believe." By Electrek's count, VTT customer report VTT-CR-00391-26, dated 1 September 2026, is the sixth VTT report and the first to measure energy density.
I am not going to argue about whether this battery is what the company says it is. That is not something a reader, or I, can settle from the documents available. What I can do is read the test report the way I would read any lab report, and separate what it measured from what people are using it to claim.
The method section is short and clear. VTT received one cell, designated DL6, described as "an energy-optimised variant of the V1 cell previously tested at VTT." They weighed it on a lab balance, measured its dimensions with a caliper, and ran a capacity test in a climate chamber at 25 °C.
The capacity test was two cycles. The cell was charged at 0.1C to 4.25 V, held at 4.25 V until the current fell to 0.05C, rested, then discharged at 0.1C down to 2.3 V. Repeat once. The energy figures come from the second discharge. Then the cell was parked at about half charge and the test stopped.
Result: 409.3 Wh/kg and 804.7 Wh/L. The report adds that, on visual inspection, the cell did not swell. That is the whole finding.
It is a legitimate measurement from a competent lab, with instrument specifications listed and a named author and reviewer. Energy density really is that simple: energy out divided by mass. Nothing in this report looks sloppy to me. The limits are in the scope, not the execution.
The discharge rate. By Donut Lab's own definition, 1C takes a battery from empty to full in one hour, so a 0.1C discharge takes roughly ten hours. Slow discharge is a standard and fair way to measure capacity, and it is also the gentlest case: the slower you pull current, the less energy is lost inside the cell, so the figure lands near the top of what the cell can give. That is fine, as long as nobody reads it as the energy the cell delivers at a working rate.
The voltage window. VTT's Table 1 says the 2.3 to 4.25 V range was a "preliminary specification of the device under test, as provided by the customer." The first VTT report, in February, tested the earlier V1 cell with a customer-supplied recommended range of 2.7 to 4.15 V and discharged it at 1C. The new window is wider at both ends and the rate is ten times slower. Both changes add watt-hours. Neither is illegitimate if the cell is rated for it, but the two reports are not measuring the same thing on the same terms.
The cell. The 409 figure belongs to the V1.5. The fast-charge result from February was on a V1 cell. A reader who stacks the headline numbers from different reports into one sentence ("fast charging and 400 Wh/kg") is combining results from different cells.
The volume. VTT measured volume as the box that holds the pouch with its side seams folded and its tabs excluded, following Annex B of IEC 62902:2025. That is a defined, reasonable convention. It is still a cell-level box, not the space the cell takes up inside a finished product.
Here is the part that matters most. The word "solid" appears exactly once in the text of the report, in the project name ("Solid-State Battery Performance Test 3"). The report makes no determination of chemistry at all: nothing on the electrolyte, the anode or the cathode. The February report is even more explicit, describing the test articles as devices "which the customer identified as solid-state battery cells." That is careful, correct lab language. A contract lab measures what it is asked to measure and says whose claim the rest is.
VTT's own public statement from 20 February reads the same way. It says VTT "has conducted measurements on the properties of Donut Lab's battery," that Donut Lab would publish the results, and that "Donut Lab also responds to questions related to measurements and the battery technology it has developed." Measurement service, yes. Endorsement of the technology, no.
Also absent: cycle life. Two cycles tell you nothing about 100,000. Electrek, reviewing the series, puts it bluntly: "This test ran two capacity cycles and stopped."
For an electric motorcycle, Wh/kg matters because every kilogram is carried. For stationary storage (a home battery, a backup system, a grid container), weight is rarely the constraint. What decides whether a storage battery is a good buy is the stuff this report did not test: how much capacity survives years of daily cycling, how the cell behaves at the charge and discharge rates it will actually see, and how it fails. I have written about each of those separately: what wears grid batteries out, what a warranty's end-of-life figure means, and what a fire test does and does not certify.
That is also the lens I use in my own work. I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator. I do not own that company and earn nothing from this link. Full policy here. Nothing in this report involves them or any product they sell.
This case is a useful template because the documents are public. When a battery company cites a third-party test, ask five questions, and the report itself will usually answer them:
1. Who wrote the test plan? Here, the customer ("carried out in accordance with the customer's test plan"). That is normal for commissioned work, and it means the lab tested what the company chose to have tested.
2. How many samples? One cell. A single good cell shows the design can reach a number; it says nothing about how consistent production cells are.
3. At what rate, in what window? 0.1C, 2.3 to 4.25 V, 25 °C. Compare that with how the product will actually be used.
4. How many cycles? Two.
5. What does the lab say it did not determine? Chemistry, durability, pack-level performance. Read the summary and conclusions for what is missing, not just what is there.
I read the report as published by the company. The PDF I read was downloaded from the copy Electrek hosts; the company's own results site links a file with the same report number. VTT marks it "VTT Confidential" and says its name may only be used in publishing with written authorisation; I am assuming the company has that authorisation, since it chose to publish. I have not seen any VTT data file behind the summary numbers.
I did not read the other test reports in the series in full (high temperature, self-discharge, pack charging, cycling, or the Intertek reports). My comparison is with the first VTT report only, which I did read.
I make no finding about the chemistry. Electrek reports allegations and denials about what is inside these cells. I have not evaluated those and nothing here depends on them.
VTT measured one Donut Lab cell at 409.3 Wh/kg, on a slow ten-hour discharge, over two cycles, in a voltage window the customer supplied. That is a real number and a narrow one. It says nothing about whether the cell is solid-state, how long it lasts, or how it performs at working rates, and the report does not pretend otherwise. When a battery headline says "independently tested," the test report is usually more honest than the headline. Read it.
Scope note: this report reads published test documents. No battery was tested and no finding is made about any cell's chemistry. The five questions are the author's framework for reading commissioned test reports.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.