The world's best-selling Bluetooth health device is a disposable

By Shailendra Kumar Ram · 2026-09-09 · 14 min read

I have marked derived figures as derived, inferences as inferences, and gaps as gaps. Where the honest answer is that nobody has measured something, that is what it says.

Why the question has no lookup answer

IDC, Canalys, Omdia and Counterpoint all publish wearable and medical device numbers. None of them counts BLE radios, and none treats "health device with a BLE radio and a companion app" as a category. The answer below is on one specific definition, radios shipped per year. Count cumulative units of a single product line instead, or installed base of a durable device, and a different product wins.

One methodological point does most of the work. For a disposable sensor, users and units are different questions with answers two orders of magnitude apart. Every source I found reports Libre users. None reports sensors. For a product whose radio is consumed roughly 25 times per user per year, that distinction is the whole story, and I did not find an article on this topic that makes it.

The finding: the Libre's BLE radio is a consumable

A durable health device ships one radio and keeps it for years. A continuous glucose monitor ships a new radio inside every sensor, and the sensor lasts 14 or 15 days. That single architectural difference dominates every commercial factor:

Turning that ratio into an annual figure requires a derivation, so here it is in full, with both cross-checks, marked clearly as an estimate rather than a citation:

• From users. Over 8 million users, times roughly 24 to 26 sensors a year at 14 or 15 day wear, gives roughly 190 to 210 million sensors a year. This assumes near-continuous wear and therefore overstates the real figure, since adherence is never total.
• From revenue. Abbott reported Diabetes Care sales of $7,998 million for Abbott's full year 2025 results. At an assumed blended selling price of $35 to $40 per sensor, that implies roughly 200 to 230 million sensors. The price assumption is mine and varies enormously by reimbursement regime. This method also treats the whole Diabetes Care segment as sensor revenue, which it is not, so read it as an upper bound rather than a point estimate.

Two methods landing on the same order of magnitude is reassuring, though the second is a plausibility check rather than a confirmation, since I chose the price. Note also that the user count is from 2026 and the revenue from 2025, so they describe different years. Treat around 200 million BLE radios a year as a Fermi estimate with stated assumptions, not as a statistic. What it does support is the comparative claim: no durable health wearable ships anything like that many radios annually, and Libre does it by discarding the radio every fortnight.

What it sells for, and what Abbott makes from it

Abbott does not report Libre as a line item. The closest published figure is the Diabetes Care segment, and the segment is very nearly all Libre: in the fourth quarter of 2025, continuous glucose monitors were $2.0 billion of a $2,133 million segment, about 94 per cent. Abbott gives that CGM number for the quarter only, never for the year, so the annual figures below are segment totals.

The revenue is useful for a second reason. The sensor estimate derived from users above does not use price anywhere, so it can be run in the other direction to get a price. Roughly $7.5 billion of CGM revenue across 190 to 210 million sensors implies a realised price of about $36 to $40 per sensor. That is not an independent confirmation of the earlier estimate, since both rest on the same user count, but it is a number Abbott has never published.

It is worth putting next to what the sensor costs a person at the counter:

Abbott realises somewhere around a third of the US cash price. That gap is the whole commercial shape of the product: this is a business built on reimbursement and volume, not on retail margin, which is also why a disposable radio is affordable in the first place. Nobody would throw away a $123 radio every fortnight. Throwing away a $37 one, most of which somebody else pays for, is a different proposition.

Does the FreeStyle Libre 2 use Bluetooth?

Yes. The FreeStyle Libre 2 uses both Bluetooth Low Energy and NFC. A large number of comparison articles state that Libre 2 is NFC-only and that Libre 3 introduced Bluetooth. That is wrong. The FDA's substantial equivalence review 510(k) K193371, received in December 2019 and decided in June 2020, documents Libre 2 with both radios: NFC for scanning the sensor and BLE originally carrying the glucose alarms. The distinction matters here, because the original 2014 Libre had NFC only, so cumulative BLE units for the product line start from the Libre 2 era, not from the 2014 launch.

Inside a Libre 2 sensor: two radios, one coin cell

Two independent public teardowns of the Libre 2 sensor agree on what is in it: iFixit and Gough Lui. A third, paid TechInsights analysis covers the same part, but it is behind a paywall and I have not read it.

The architecture is a deliberate choice rather than an obvious one. This is a two-radio medical device running fourteen days on a single silver oxide button cell, and it gets there by splitting the work: a sensing domain that can be powered by NFC field harvesting, and a separate BLE companion part that handles the radio. A single integrated SoC would have been simpler and would not have made the power budget.

The protocol is proprietary and, from Libre 2 onward, encrypted. The Bluetooth SIG has published an adopted Continuous Glucose Monitoring Service for years, covering exactly this device class, including handing over the readings taken while the phone was away. Abbott appears to route around it, though I am inferring that from the reverse-engineering literature rather than from an audit of what the sensor actually exposes.

That literature is public. The original Libre was unencrypted and opened up in 2016, and Libre 2's encryption setup was partially mapped in 2020. The community consensus that Libre 2 uses AES-128-CCM with an ECDH key agreement comes from the AndroidAPS and DiaBLE ecosystems rather than from a vendor specification or a paper, so treat it as practitioner knowledge rather than established fact. Documentation for Libre 3 is thinner, and at least one researcher reports receiving DMCA pressure from Abbott over publishing it.

The LibreLink app half, and what regulation does to it

A finding I did not expect. There is no single install figure for this app, by construction. Abbott does not ship one LibreLink. It ships a separate Google Play listing per country, each with its own bundle identifier, because each one is a separately regulated medical device in its own jurisdiction. `com.freestylelibre.app.us`, `com.freestylelibre.app.gb`, `com.freestylelibre.app.fr` and so on are different apps. Any article citing an install count for LibreLink is citing one country.

That fragmentation is invisible to anyone who has only built consumer BLE apps, and it has consequences for release engineering, crash triage and support that a consumer app never faces. It is also why the ratings only mean anything one country at a time:

A paper that reported the device volume and stopped there would be telling half the story. The best-selling BLE health device in the world has a companion app rated below three stars in its two largest English-speaking markets. For a product where the app is the alarm surface rather than a convenience, that is not a cosmetic problem.

The failure mode that matters: background alerts

Every engineer who has built a BLE companion app knows the background execution problem. On iOS, Core Bluetooth's state preservation and restoration will relaunch a terminated app for a Bluetooth event, but not if the user force-quit it, and the system can terminate it under memory pressure regardless. On Android, Doze and the background service limits do comparable damage. In a consumer product this produces a support ticket. Here it produces something else, and unusually for anything in BLE engineering, the regulator has written it down.

The FDA issued a safety communication on 5 February 2025 about diabetes device alerts not reaching users. The contributing causes it names are a list of platform behaviours: notification permissions, Do Not Disturb and Focus modes, battery saver, apps being put to sleep after disuse, operating system updates the device does not support, and newly connected audio accessories redirecting sound. The agency reports that missed alerts contributed to cases of severe hypoglycaemia, severe hyperglycaemia, diabetic ketoacidosis and death.

There is a peer-reviewed case report of the specific mechanism. Sato and colleagues describe a woman in her late forties with type 1 diabetes whose phone ran an automatic operating system update overnight; at four in the morning she treated hypoglycaemia herself with no alert having fired, and hours later was found in a hypoglycaemic coma requiring emergency care (Therapeutic Advances in Endocrinology and Metabolism, 2025). She recovered without lasting harm. The authors note that the FDA's advice to disable automatic updates is not widely known. Nothing here is clinical advice; it is a note about platform behaviour for the people building these systems.

Limitations

• The headline unit figure is derived, not sourced. Both cross-checks rest on assumptions I have stated, and the revenue method depends on an average selling price I chose.
• All the teardown evidence is for Libre 2. I found no component-level teardown of Libre 3 from a source I would cite.
• The encryption specifics come from reverse-engineering communities rather than from vendor documentation or peer review, so treat them as practitioner knowledge.
• That Abbott bypasses the standard CGM GATT profile is inference. It follows from the reverse-engineering literature rather than from a direct audit.
• App ratings are a Play Store snapshot. I did not pull App Store figures and I did not analyse the review text, so I am not claiming to know why users rate these apps as they do.
• Treating the Diabetes Care segment as Libre revenue rests on a 94 per cent CGM share measured in one quarter and applied to the year. Abbott does not publish the annual split.
• The price table mixes a 2026 US cash price with a 2017 UK retail figure, and the implied realised price shares its user count with the sensor estimate, so it is a restatement rather than a check.
• Where this post says no measurement exists, it means I did not find one, not that none was published.

Where I can help with a BLE health device

Most of the problems above are not research problems for the teams building these products, they are Tuesday. These are the ones I am usually brought in for:

• Alert reliability. I audit what actually happens to your notifications across iOS and Android background states, force-quit, Doze, battery saver and OS upgrades, and make the recovery path deliberate rather than accidental.
• Connection and reconnection strategy. Advertising and connection intervals set against a real power budget rather than a default, and reconnection after a dropped link made boring.
• Firmware and app together. Most of these problems live in the seam between the two, which is a bad place for a boundary between two suppliers.

If any of that is live for you, hiring a firmware and Bluetooth app developer and the cost and timeline questions have their own posts, so you can price a project before talking to anybody. For the build itself, what it takes to launch a health band covers the ground.

The best-selling Bluetooth radio in healthcare is one nobody thinks of as a radio, in a product people throw in the bin every fortnight.