Report 174 · Energy Storage
A five-digit cycle count is a powerful sales number. Hithium's launch release for its first one shows how that kind of number gets built: 4,000 cycles measured, 20,000 projected, at a looser end-of-life line than the industry usually quotes. Nine months later, the qualifiers were gone.
Hithium is a Chinese maker of storage cells and systems. In December 2024 it launched the ∞Cell N162Ah, which it called the first sodium-ion battery designed specifically for utility-scale energy storage. The chemistry is a sodium iron ortho-pyrophosphate cathode with a hard carbon anode. I am not picking on this company; its release is just unusually clear about how its headline number was made, which makes it a good teaching case for every cycle-life claim you will read.
Here is the launch release, verbatim, on cycle life:
At 25°C and 1P power conditions, its capacity retention reaches 94.2% after 4,000 cycles and is expected to achieve a cycle life of over 20,000 cycles at a state of health (SOH) of 70%.
Read it slowly and there are three separate facts in it. First, what was measured: 4,000 cycles, with 94.2% of capacity left, at room temperature and one stated power rate. Second, what was projected: the other 16,000 cycles. The word doing the work is "expected." Third, where the finish line was drawn: 70% state of health, meaning the cell counts as alive until it has lost 30% of its capacity.
Here is my own arithmetic, not anything the company published. Going from 100% to 94.2% is a loss of 5.8 points over 4,000 cycles. If the cell keeps losing capacity at exactly that rate forever, it needs about 20,700 cycles to lose 30 points and reach 70%. So "over 20,000 cycles" is what you get by drawing a straight line through the measured data and extending it about five times past the end of the test.
I do not know that Hithium's engineers used a straight line. They may have a fitted model with more behind it. The release does not say, and that is the point: the reader is given the projection without the method.
The same line, run to 80% instead of 70%, gives about 13,800 cycles. Eighty percent is a common end-of-life line in battery research; the review of battery aging I cite below uses "the number of cycles until 80% retention" as a lifetime metric in its own analysis. Moving the finish line from 80% to 70% is worth roughly 6,900 cycles on paper, before anything about the cell changes. I covered the same move in home-battery warranties in what a battery warranty's end figure means.
Battery capacity does not have to fall in a straight line. In 2022 a team of seventeen battery researchers from Stanford, Oxford, Warwick, NREL, Sandia and elsewhere published a review in the Journal of The Electrochemical Society on exactly this problem. They describe three shapes an aging curve can take: sublinear (slowing down), linear, and superlinear, where fade suddenly speeds up. The field calls that bend a "knee," and the review lists other names for it, including "sudden death" and "capacity plunge."
Their warning about prediction is the part a buyer should remember. Some of the processes behind a knee have signals the authors call electrochemically undetectable, and others require "extrapolation of an exponential function, which is often an error-prone exercise." Even more pointed for anyone reading a state-of-health figure: "batteries with identical states of health (i.e., estimated capacity or energy retention) may have entirely different remaining useful lives."
That review covers lithium-ion cells. I have not found a published equivalent for sodium iron pyrophosphate cells, and I am not claiming this cell will hit a knee. Its curve may stay gentle all the way. The narrower point is that 4,000 good cycles cannot show you whether a bend is waiting at 9,000 or 15,000, because the test stopped before either.
The next sentence in the release reads: "Additionally, at 45°C, the cycle life improves by more than 5 times, with a capacity retention of 92.5% after 4,000 cycles." I cannot make that sentence add up. At 45 °C the cell kept less capacity after 4,000 cycles (92.5%) than it did at 25 °C (94.2%), so "improves by more than 5 times" must be a comparison against something else, perhaps a competing chemistry at high temperature. The release does not say what. When a datasheet line can't tell you its own baseline, treat it as marketing until someone shows the comparison.
In September 2025 Hithium announced that the same cell had passed a national sodium-ion battery assessment, run by a standards institute under China's Ministry of Industry and Information Technology against an industry-association specification. That is a real milestone. But the cycle claim in that release reads simply that the cell is "offering over 20,000 charge cycles." No "expected." No 70%. No 4,000.
Nothing in the release says the assessment measured cycle life, and a quick sanity check suggests it could not have measured 20,000 cycles. If "1P" means what it usually means on storage datasheets (charging or discharging at full rated power, about an hour each way), one cycle takes at least two hours. Twenty thousand cycles is then at least 40,000 hours of nonstop testing, about four and a half years. The cell was launched in December 2024. Again, my arithmetic and my reading of "1P," not the company's.
This is how most cycle-life numbers travel. The first document carries the conditions. Each repetition sheds one. By the time the number reaches a sales deck or a news headline, a projection reads like a measurement.
1. How many cycles were actually run? Here, 4,000. Everything past that is a model.
2. To what end-of-life line? 70% and 80% are not the same promise. Ask which one, and convert if you're comparing two products.
3. At what rate, depth and temperature? Duty matters as much as chemistry; I wrote about that in what actually kills a grid battery, and temperature can even flip direction between chemistries, as in does heat kill your battery.
4. How was the rest projected? A straight line, a fitted model, accelerated testing? If the document won't say, you can't judge it.
5. What does the warranty actually cover? The warranty is the number the seller is willing to pay for. If it is smaller than the brochure number, believe the warranty.
These are also the questions I work on day to day. I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator, run by fellow veterans. 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.
I read Hithium's two English-language releases, not a test report. I have not seen the cycling data, the test protocol behind "1P," or how the 20,000 figure was derived. If Hithium publishes the curve or the method, I will add it here.
I did not read the assessment specification (T/CNESA 1006-2025) or the assessment results, so I make no claim about what that evaluation did or did not test. My point is only that the September release does not say it measured cycle life.
I make no prediction about this cell. It may well outlast the projection. The arithmetic above describes the claim, not the battery.
Hithium measured 4,000 cycles at 94.2% retention and projected more than 20,000 cycles to 70% health. A straight line through the measured point lands right on that number, and battery researchers warn that aging curves often stop being straight. The conditions were in the launch release and gone from the next one. When you see a five-digit cycle count, find the first document that used it, and read the sentence it came in.
Scope note: the straight-line projections (about 20,700 cycles to 70% and 13,800 to 80%) and the minimum test-time estimate are the author's arithmetic from the figures in the launch release, labeled as such in the text. No battery was tested.
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.