← Back to blog

Essay No. 025 · AI Infrastructure · Melbourne, Australia

AI Infrastructure Wearable Health Rockley Photonics Apple Watch Non-Invasive Glucose Silicon Photonics Biomarkers FDA Dexcom Continuous Monitoring Medical Devices

The Clinic-on-Wrist Reality Check.Original analysisNot medical adviceNot investment advice

Why non-invasive biomarkers are still the holy grail of wearable health.

Pugalenthi Magendran

April 2026 · Melbourne, Australia

12 min read

In 2021, Rockley Photonics looked like the company that could turn the Apple Watch into a “clinic on the wrist.” The dream was right. Continuous biomarker monitoring is one of the biggest opportunities in consumer health. But the product timeline was wrong. Non-invasive glucose and wearable biomarkers are not just sensor features. They are physics, clinical validation, FDA, privacy, manufacturing, and workflow problems.

In 2021, Rockley Photonics looked like the future of the Apple Watch. The promise was beautiful. Put a spectrometer on the wrist, shine light into the skin, measure multiple biomarkers continuously, and turn a smartwatch into a clinic on the wrist.

The uploaded SemiAnalysis article framed Rockley as a silicon-photonics company that could measure glucose, lactate, alcohol, hydration, body temperature, blood pressure, blood oxygen, and heart rate using lasers on a chip. It also said Apple had already paid Rockley $70M in engineering fees and suggested Apple Watch integration could begin around 2022 or 2023.¹

That did not happen. There is still no Apple Watch today that directly measures blood glucose non-invasively. The FDA has explicitly warned that no smartwatch or smart ring has been authorised, cleared, or approved to measure or estimate blood glucose values on its own.²

That does not make the original idea worthless. It makes it more interesting.

The destination was right. The timeline was wrong. The difficulty was underestimated.

Key idea

The correct claim is not that Rockley failed and wearable biomarkers are dead. The correct claim is that Rockley was right about the direction and wrong about the timeline. Continuous biomarker monitoring is still one of the biggest opportunities in consumer health, but the winning product will need clinical accuracy, regulatory approval, privacy trust, manufacturing scale, reimbursement logic, and integration into real healthcare workflows.

I. The 2021 thesis

In June 2021, Dylan Patel published a SemiAnalysis piece on Rockley Photonics. The framing was bullish. Rockley was a silicon-photonics company that wanted to miniaturise spectroscopy into a wearable sensor module. The claimed biomarker list included glucose, lactate, alcohol, hydration, body temperature, blood pressure, blood oxygen, and heart rate. The article reported that Apple had paid Rockley $70M in engineering fees and pointed to Apple Watch integration in roughly 2022 or 2023. The deeper thesis was that wearables could move from basic optical sensing into continuous biochemical monitoring, with Rockley as the full-stack platform behind it.¹

I revisited that piece because the 2021 destination still matters, but the 2026 reality is very different.

2021 thesis

Rockley’s promise was not just a better smartwatch sensor. It was a wearable spectrometer: a silicon-photonics platform that could turn the wrist into a continuous health-monitoring surface.

II. The dream was correct

Most current smartwatches rely on LED-based photoplethysmography (PPG). PPG looks at changes in reflected light related to blood volume. It is good enough for heart rate, blood oxygen estimation, heart-rate variability, respiratory rate proxies, and activity and sleep signals. It is not designed to read biochemical state.

Rockley wanted to push past PPG into spectroscopy. Short-wave infrared sensing. Laser-based illumination. Photonic integrated circuits paired with an interpretation ASIC. Multi-modal biochemical and biophysical monitoring instead of one optical signal.¹

Today’s wearable

PPG and motion

Sensors · LEDs and accelerometers.
Signals · heart rate, SpO2 estimate, HRV.
Use · activity, sleep, wellness.
Limits · not biochemical, not diagnostic.

Clinic-on-wrist dream

Spectroscopy + biomarkers

Sensors · SWIR / laser / PIC + ASIC.
Signals · hydration, temperature, glucose trends, lactate, alcohol, blood pressure signals.
Use · continuous health monitoring.
Required · clinical validation, FDA, privacy, workflows.

The future wearable will not just count steps. It will become a continuous health interface.

III. The reality check

The most important 2026 fact about wearable biomarkers is regulatory, not technical.

FDA safety communication

The FDA has not authorised, cleared, or approved any smartwatch or smart ring intended to measure or estimate blood glucose values on its own, and warns that inaccurate glucose readings can lead to dangerous medication errors.

Source: U.S. FDA safety communication on smartwatches and smart rings.²

A watch face can be wrong. A calorie estimate can be wrong. A sleep score can be fuzzy. Glucose cannot be casually wrong. Wrong readings can drive wrong insulin doses, wrong food corrections, missed hypoglycemia, and overcorrection. That is the line between a wellness feature and a medical decision.

A watch face can be wrong. A glucose reading cannot.

IV. The SPAC-era mistake

The 2021 story had a classic SPAC-era shape. Huge TAM. Apple connection. Patents. Healthcare disruption. Wearable adoption. FDA upside. Explosive projections. A “clinic on the wrist” narrative that sounded clean in a deck.

Medical sensing does not scale like a normal gadget feature. A prototype is not a product. A partner demo is not regulatory clearance. A biomarker trend is not a medical-grade measurement. Consumer wellness is not clinical diagnosis. An Apple relationship is not a guaranteed Apple Watch integration. A patent portfolio is not adoption. A TAM is not revenue.

The bottleneck was not only miniaturisation. It was trust.

V. Rockley became a commercialization-risk case study

Rockley filed for Chapter 11 in January 2023. The restructuring plan was designed to reduce debt and provide working capital, with approximately $35M of cash for ongoing operations cited in the filing.³ That does not mean the technology was fake. It means the path from photonic sensor to regulated consumer-health product was far harder than the original market story implied.

Diagram · Nine different mountains, not one

Sensor physics

Optics, lasers, detectors, signal-to-noise.

Working prototype

Hardware that shows the idea is feasible.

Sampled platform

Reference design shipped to partners.

Clinical study

Real patients, real conditions, real comparators.

Regulatory clearance

FDA / equivalents authorise medical claims.

Mass production

Yield, packaging, supply chain, support.

Apple-scale integration

Industrial design, battery, thermals, software.

Reimbursement & workflow

Insurers, doctors, EMRs, behaviour change.

Real patient adoption

Daily use that actually improves outcomes.

Those are different mountains. The mistake was thinking they were one mountain.

VI. The technology still matters

Post-restructuring, Rockley began sampling its Bioptx Biosensing Band and Platform to strategic customers and partners, streaming SWIR-based body temperature and hydration signals alongside LED/PPG-based heart rate, HRV, respiratory rate, and blood oxygen saturation.⁴ The company has also reported development progress on non-invasive blood glucose monitoring, including 5 mg/dL prediction accuracy in simulated skin samples and a human study with 40 diabetic subjects over 10 weeks.⁵

That is meaningful technical progress, but not proof of a mass-market medical product. A promising study is not a consumer product. A development platform is not a medical-grade watch. A glucose trend is not automatically a CGM replacement.

A promising study is not a consumer product. A development platform is not a medical-grade watch.

VII. The watch became the interface, not the sensor

The cleanest 2026 reading of the market is in how Dexcom positions Direct to Apple Watch. Dexcom says G7 readings can stream directly to the Apple Watch, without requiring a nearby iPhone, but the glucose measurement still comes from the Dexcom CGM sensor on the body, not from the watch itself.⁶

Apple Watch

The health interface

Wins: display, notifications, behaviour nudges, ecosystem integration, accessibility. Does not (yet): directly measure glucose on its own.

Dexcom CGM

The medical sensor

Wins: clinically validated, FDA-cleared continuous glucose monitoring, integrated with insurance and care pathways. Costs: a minimally invasive sensor worn on the body.⁶

The wrist won the interface. The invasive CGM still won the measurement.

VIII. Apple’s health strategy is incremental and regulated

Apple’s health roadmap is conservative in a specific direction. Apple Watch Series 11 and the surrounding watchOS release added features such as a sleep score and hypertension notifications. Apple describes hypertension notifications as a way to look for signs or patterns of chronic high blood pressure over time. It is explicitly not a replacement for a cuff-style blood pressure monitor or a clinical diagnosis.⁷

That pattern is the rule, not the exception. Apple tends to ship a health feature only when it is validated, explainable, and safe enough for mass-market use. The headline feature you see on a launch slide is usually the conservative end of what the underlying sensor can detect.

Apple’s health strategy is not “ship the most ambitious sensor.” It is “ship the safest validated health signal that can survive mass-market use.”

IX. Why non-invasive glucose is so hard

Non-invasive glucose is hard for a small set of stubborn reasons. The glucose signal is weak. Water dominates absorption in the relevant optical bands. Skin varies across people. Skin tone varies. Tissue thickness varies. Hydration changes. Temperature changes. Blood perfusion changes. Motion creates artifacts. Sweat and contact pressure change readings. Calibration drifts over time. Algorithms can overfit to individual or study conditions. Clinical populations are diverse. Medical decisions have high consequences.

That is why the FDA warning is so direct. Non-invasive glucose is not a feature. It is a medical-grade measurement problem under consumer-device constraints.²

Non-invasive glucose is not a feature. It is a medical-grade measurement problem.

X. Why the long-term opportunity is still huge

The reality check does not delete the opportunity. People dislike finger pricks. Invasive sensors are uncomfortable and expensive. Chronic disease management is one of the largest healthcare spending categories in the world. Aging populations need more monitoring. Healthcare systems are strained. Remote patient monitoring is growing. Wearables are normal. Insurers and physicians want earlier intervention.

The winning product still needs clinical-grade accuracy, low false positives, low false negatives, a defensible calibration strategy, regulatory clearance, privacy protection, a reimbursement model, physician trust, consumer UX, manufacturing yield, and long-term reliability.

The market is not waiting for another wellness chart. It is waiting for a trusted measurement.

XI. The full stack of wearable biomarkers

The 2021 deck already understood part of this stack. The missing part was proving all of it at medical-grade scale.

Diagram · The full stack of wearable biomarkers

Reimbursement & behaviour change

Outcomes

Doctor / insurer workflow

Workflow

Consumer UX

Trust

Privacy & security

Risk

Regulatory validation (FDA / EU)

Approval

Clinical datasets

Evidence

Biomarker models

Signal processing

DSP

Firmware

Embedded

Packaging / module integration

Integration

ASIC / electronics

Silicon

Photonics / sensor hardware

Optics

A wearable biomarker company is not only a chip company. It is a medical evidence company.

Quick terms

XII. What could break the optimistic thesis

The bear case for wearable biomarkers is not that the science is fake. It is that the science is hard.

Bear case · what slows the dream

Persistent noise. Non-invasive glucose may stay too noisy for broad medical use.
Calibration burden. Per-user calibration may be too painful for consumers.
FDA bar. Regulatory standards for medical claims may remain very hard to meet.
Apple chooses safety. Apple may keep shipping incremental, validated health features instead of the spectroscopy dream.
CGM incumbents. Dexcom and Abbott may keep dominating glucose through minimally invasive sensors.
Trust gap. Consumers may not trust wrist-based biomarker readings for medical decisions.
Reimbursement. Insurers may not reimburse wellness devices that lack clinical evidence.
Population diversity. Algorithms may fail across skin tones, body shapes, and clinical conditions.
Real-world conditions. Sensor contact, motion, sweat, and skin differences may limit reliability.
Privacy. Health-data rules may slow adoption.
Manufacturing cost. Photonic modules may stay too expensive for mass market.
Trial generalisation. Small studies may not generalise to large populations.
Trend vs measurement. Companies may overpromise “trends” as if they were medical readings.

The hardest risk is not that the sensor never works in a lab. It is that it works well enough, for enough people, under enough real-world conditions, to deserve medical trust.

XIII. What could break the bear case

The bull case is also worth taking seriously. The same forces that make wearable biomarkers hard are the forces that make them valuable when they finally arrive.

Bull case

Why the dream still wins

Optical sensors keep getting smaller, cheaper, and more sensitive.
Photonic integration lowers size and cost over time.
AI signal processing improves at extracting weak biomarker signals.
Larger clinical datasets improve population coverage.
Wearable adoption normalises continuous monitoring.
Distribution via Apple, Dexcom, Abbott, Samsung, Fitbit ecosystems.
Remote patient monitoring is a structural healthcare trend.
Trend-based features may arrive before full diagnostic accuracy.

Most likely path

Early-warning layer first

Step 1 · better wellness signals on existing wearables.
Step 2 · regulated trend-based features (e.g. hypertension notifications).⁷
Step 3 · integration of medical sensors with wearable interfaces.⁶
Step 4 · biomarker monitoring with strict labelling and disclaimers.
Step 5 · medical-grade non-invasive biomarkers after clinical evidence and regulatory clearance.

The first winning wearable biomarker may not replace a lab test. It may become an early-warning layer.

XIV. What to watch

Working checklist, not a prediction. These are the signals to watch.

What to watch

FDA guidance on non-invasive glucose devices.
Apple Watch health sensor roadmap.
Dexcom and Abbott direct-to-watch features.
Rockley / Bioptx customer announcements.
Peer-reviewed clinical validation studies.
Sample size and diversity in studies.
Real-world accuracy, not just benchtop results.
Calibration requirements for users.
False-positive and false-negative rates.
Reimbursement pathways for wearables.
Insurer pilots and outcome data.
Doctor workflow integration.
FDA clearances for trend-based biomarkers.
Spectroscopy miniaturisation.
Privacy and health-data rules.
Apple / Samsung / Google partnerships with CGM or biosensing companies.

XV. The clinic-on-wrist reality check

Rockley was not wrong about the future. It was early about the product.

The wrist will become one of the most important health interfaces in the world. But the clinic on the wrist will not arrive through hype, SPAC decks, or Apple rumours. It will arrive when optics, silicon, algorithms, clinical validation, regulatory approval, privacy, manufacturing, reimbursement, and healthcare workflows finally line up.

The dream is still alive. The shortcut was not.

¹ Patel, D. (Jun 2021). Rockley Photonics Will Revolutionize Healthcare By Measuring Biomarkers Such As Glucose With Lasers In The Apple Watch. SemiAnalysis. Historical anchor for the 2021 Rockley framing, including the clinic-on-wrist visual, claimed biomarkers, $70M Apple engineering fees, suggested 2022/2023 integration, and the full-stack platform argument. Used as inspiration only. No content, structure, or charts reproduced.

² U.S. Food and Drug Administration. Do Not Use Smartwatches or Smart Rings to Measure Blood Glucose Levels: FDA Safety Communication. The FDA states it has not authorised, cleared, or approved any smartwatch or smart ring intended to measure or estimate blood glucose values on its own, and warns about safety risks of inaccurate glucose readings.

³ Rockley Photonics Holdings Limited. 8-K filing, January 2023. Chapter 11 restructuring plan with associated debt-reduction terms and approximately $35M of cash for ongoing operations as reported in the filing.

⁴ Rockley Photonics / BusinessWire (Dec 2023). Rockley Photonics begins sampling Bioptx Biosensing Band and Platform. Description of Bioptx Biosensing Band sampling to strategic customers and partners, including SWIR-based body temperature and hydration signals and LED/PPG-based heart rate, HRV, respiratory rate, and blood oxygen saturation.

⁵ Rockley Photonics / BusinessWire (Sep 2023). Rockley Photonics advances non-invasive blood glucose monitoring. Reported 5 mg/dL prediction accuracy in simulated skin samples and a human study with 40 diabetic subjects over 10 weeks, framed as development progress rather than FDA-cleared product.

⁶ Dexcom. Dexcom G7 Direct to Apple Watch. Apple Watch streams Dexcom G7 CGM readings directly without requiring a nearby iPhone, with the glucose measurement coming from the Dexcom CGM sensor on the body.

⁷ Apple. Apple debuts Apple Watch Series 11. Apple Watch Series 11 health features including sleep score and hypertension notifications, framed as trend-based health insights rather than replacements for clinical measurements.