At the end of September 2026 it emerged that Garmin had bought Moxy — the American maker of wearable muscle oxygen saturation sensors one of the oldest players in this market. No press release, no regulatory filing: the news came out via the industry blog the5krunner, and six days later Garmin confirmed the deal to DC Rainmaker in a single sentence with no details. The price was not disclosed.

Why Garmin wants it is not hard to guess: back on 19 February 2026 the company filed a trademark application for Muscle Battery — by its description, for an algorithm built on SmO₂ data. By analogy with Body Battery, it will probably be a “charge level” for a specific muscle. It can't be calculated from the wrist: you need a separate sensor on the muscle itself. That said, Garmin watches and bike computers have long been able to read Moxy via its ANT+ profile.

Before you start waiting for a “quad battery percentage”, it's worth understanding what SmO₂ is and how accurately it is measured.

How the sensor works

The sensor, stuck onto a muscle (most often the vastus lateralis on the outside of the thigh), shines near-infrared light into the skin and picks up what is reflected back. Haemoglobin with and without oxygen absorbs this light differently, and from the ratio the device calculates what fraction of haemoglobin in a small volume of tissue beneath it is saturated with oxygen. The method is called NIRS — near-infrared spectroscopy.

The key words are “in a small volume of tissue beneath it”. This is not the saturation of the whole muscle, let alone of the whole body.

What's wrong with the number: a 2026 review

A review by Feldmann and colleagues from Bern, L'Aquila and Montpellier (Sports Medicine — Open, June 2026) untangles the confusion between SmO₂ (“saturation of the muscle specifically”) and StO₂ (“tissue saturation”). The signal passes through skin and subcutaneous fat, and how much of the final number comes from muscle depends on fat thickness, skin pigmentation, sensor geometry and the device's mathematical model. The authors propose writing rSmO₂ — “regional” saturation — so as not to fool ourselves.

A practical check from the review: occlude the artery with a cuff. If the signal is mostly muscle, it covers a wide range — dropping to about 20%, and jumping to almost 80% once the cuff is released. If skin and fat strongly influence it, the range shrinks noticeably. No commercial system, the authors stress, fully isolates the muscle.

An important caveat: the first author is a minority shareholder and co-developer of Moxy, which is honestly disclosed in the paper.

How stable is the number from test to test

New Zealand, 2017. 10 high-level cyclists did a graded test twice. At low and moderate power SmO₂ is well reproducible, but as intensity rises, the spread grows — because the contracting muscle squeezes its blood vessels, and because of motion artefacts. Moxy's second metric, total haemoglobin (THb), was judged by the authors to be unsuitable for assessing oxygenation.
Germany, 2024. 12 trained cyclists and triathletes did a graded test with 5-minute stages three times. The reliability coefficient (ICC) was 0.79–0.92, formally “good”. But the standard error of measurement was 5–9 points of SmO₂, and the minimal real change was 14–18 points. In other words, if today you see 45% at a given power and a week later 35%, that may still be noise.
Left and right legs in the same study differed by an average of 2 points, but with limits of agreement of ±19.9. And this was not explained by leg dominance. A sensor on your left thigh won't tell you what's going on in the right one.

Where SmO₂ really works: the slope instead of the number

The good news came from the University of Oregon (Medicine & Science in Sports & Exercise, March 2026). 30 endurance athletes (15 women) did a protocol of four 4-minute stages on a cycle ergometer four weeks in a row, with a Moxy on the rectus femoris. The researchers looked not at the SmO₂ value but at its slope within each stage: below threshold, saturation stabilises or rises; above it, it keeps falling. The power at which the slope is zero is an estimate of maximal metabolic steady state (the boundary between the heavy and severe domains).

The “zero slope” power did not differ between visits.
ICC was 0.953, the standard error 2.6 W (1.2%), and the minimal real change 7.3 W (3.4%).

So the same “noisy” SmO₂, used as a trend within a single session, gives a threshold that reproduces from visit to visit as well as a good lab test. A caveat: this is a check of repeatability, not accuracy — in this study the result was not compared with a direct determination of maximal lactate steady state.

How to apply this

Watch the dynamics within a session, not the absolute number. How fast SmO₂ falls during an interval and how fast it recovers during the rest is informative. “I'm at 52% and my friend is at 60%” is not.
One leg, one spot. Mark the spot with a marker or a landmark and stick the sensor in the same place; don't compare your left leg with your right.
Don't compare between people. Fat thickness and skin colour change the signal, so for a lean cyclist and for someone with more subcutaneous fat, the “same” 40% means different things.
Threshold by slope is a practical method. Several 4-minute stages around your expected threshold, and the power where the SmO₂ slope becomes zero, make a workable alternative to lactate sampling, though direct comparisons with lactate are still scarce. You can cross-check the estimate with the lactate threshold calculator or the FTP calculator.
Sprints and running will be noisier. Almost all reliability data come from cycling at submaximal power; running involves more movement and sensor shifting.
Treat Muscle Battery the way you treat Body Battery. It is not yet known how Garmin will turn SmO₂ into a “charge”. We have already looked at how poorly validated different brands' readiness scores are — in our article on Body Battery, Recovery and Readiness.

The bottom line

Garmin bought Moxy quietly and back in February filed a trademark application for Muscle Battery — judging by the description, a SmO₂-based feature that requires a sensor on the muscle.
SmO₂ is the oxygen saturation of a small patch of tissue under the sensor, not of the whole muscle; fat and skin affect the number.
The absolute value is noisy: a real change starts at 14–18 points, and legs differ by up to ±20.
The threshold from the SmO₂ slope across stages reproduces from test to test within about 7 W — this is where the sensor is genuinely useful.

Sources: Feldmann A., Quaresima V., Perrey S., Ferrari M. “Distinguishing Muscle Oxygen Saturation from Tissue Oxygen Saturation: Accuracy, Limitations, and Practical Utility in Muscle Oximetry”. Sports Med Open, 2026;12(1):61. DOI: 10.1186/s40798-026-01033-w · Hudgins J.H., Schoeberlein M.I., Wilkins B.W. “Reliability and Repeatability of Determining Power Associated with Maximal Metabolic Steady State Using Changes in NIRS-Derived Muscle Oxygenation”. Med Sci Sports Exerc, 2026;58(3):517–526. DOI: 10.1249/MSS.0000000000003875 · Skotzke P., Schwindling S., Meyer T. “Side differences and reproducibility of the Moxy muscle oximeter during cycling in trained men”. Eur J Appl Physiol, 2024;124(10):3075–3083. DOI: 10.1007/s00421-024-05514-2 · Crum E.M. et al. “Validity and reliability of the Moxy oxygen monitor during incremental cycling exercise”. Eur J Sport Sci, 2017;17(8):1037–1043. DOI: 10.1080/17461391.2017.1330899