Most IoT devices have a battery problem. You deploy sensors across a factory floor, a farm, or a building — and then spend the next several years replacing batteries, or watching nodes go offline because someone forgot. For a 20-node pilot, that’s manageable. For thousands of devices across remote or hazardous sites, it starts to undermine the whole value proposition.

Energy harvesting has been trying to solve this for over a decade, and in 2026 the technology has genuinely matured. “Batteryless IoT” is no longer a research project — it’s a realistic engineering option for the right applications. Here’s the current state of the three main approaches.

Solar Harvesting

Solar is the most mature and typically most efficient approach for IoT. The cells used on edge devices aren’t the same as rooftop panels — they’re optimised for low-light environments, using amorphous silicon or organic PV materials that outperform crystalline silicon under diffuse indoor lighting. A sensor transmitting temperature and humidity every 30 seconds can often run indefinitely from a 20-30cm² indoor solar cell with a supercapacitor or thin-film battery as a buffer.

For outdoor applications, even small panels generate more than enough energy through most of the UK year. The challenge is overcast days and winter — buffer sizing and sleep cycle design has to account for days-long periods of low generation. This is where many early deployments came unstuck. Sizing the buffer for worst-case generation rather than average, and designing firmware that degrades gracefully (transmitting less frequently when energy is low), is what separates deployments that work from ones that go dark in February.

Standard ICs like the Texas Instruments BQ25570 and e-peas AEM40940 handle maximum power point tracking (MPPT), storage management, and load switching in a single package, removing a lot of the analog design complexity.

Thermoelectric Harvesting

Thermoelectric harvesting uses the temperature differential between a hot surface and ambient air to generate electricity via the Seebeck effect. Industrial IoT is the natural home for this: a sensor mounted on a pipe, motor, or heat exchanger can harvest energy from the very source it’s monitoring. No separate power supply, no cable run, no battery.

Power output is modest — typically microwatts to a few milliwatts — but for a sensor that wakes up occasionally to sample and transmit, that’s often sufficient. Companies like Perpetuum (part of SKF) and Mide Technology have been shipping thermoelectric vibration harvesters for predictive maintenance for years. What’s changed in 2026 is that wireless protocols have become efficient enough to run on these power budgets. BLE 5.4 in connection-based mode, LoRaWAN with duty cycling, and sub-GHz Thread all have standby currents in the microamp range, which was not true of their predecessors.

The main constraint is geometry: the harvester needs to be in thermal contact with the hot surface, and the ambient side needs enough exposure to maintain the temperature differential. For pipe-mounted sensors, this usually isn’t a problem. For general-purpose deployments, it limits where you can place nodes.

Vibration Harvesting

Piezoelectric and electromagnetic transducers convert kinetic energy from machinery, vehicles, or structural vibrations into electrical power. The energy profile is spiky and unpredictable, which makes it harder to design around than solar or thermal — significant capacitive storage is needed to smooth output and ensure the device can always complete a transmission. For applications where the vibration source runs predictably (a pump, a motor, a rotating shaft), it works reliably and eliminates battery maintenance entirely.

Piezoelectric harvesters are mechanically simple but frequency-sensitive — the resonant frequency of the element needs to match the dominant vibration frequency of the source. Electromagnetic harvesters are more broadband but larger and heavier. For most industrial predictive maintenance deployments, electromagnetic is the practical choice.

Power Management in Firmware

The hardware side of energy harvesting has been solved better than the software. Getting maximum energy out of a harvester, storing it efficiently, and then allocating it intelligently across sensing, processing, and radio transmission requires careful firmware design.

Current best practice uses ultra-low-power microcontrollers — nRF52840, STM32L0, SAMD21 — with aggressive duty cycling. Most of the time the device is in deep sleep, waking only to sample, process locally, and transmit. The radio is almost always the biggest power draw, so minimising transmit frequency and payload size is critical.

Zephyr RTOS has built-in power management infrastructure that makes duty cycling tractable — you declare power states and transitions rather than implementing them from scratch. Combined with harvesting ICs that handle MPPT and storage management, most of the complex power management can be handled at the component and RTOS level rather than requiring custom analog design.

Where It’s Actually Deployed

The most production-ready applications are industrial predictive maintenance (vibration sensors on motors and pumps), smart agriculture (soil moisture and microclimate nodes), building management (occupancy and air quality sensors), and cold chain monitoring (refrigerator temperature logging).

The common thread is high node density at sites where battery replacement is genuinely costly — both in labour and in operational disruption. Replacing 500 batteries annually across a large facility, or sending a technician to remote field sensors every few months, has real costs that batteryless designs eliminate.

Getting Started

For new designs, solar harvesting with a supercapacitor buffer is the lowest-risk starting point — the energy source is predictable, the hardware is well-documented, and the reference designs are mature. The Nordic nRF9161 (with integrated LTE-M/NB-IoT) and STM32U5 are current reference MCUs for energy-harvesting nodes. Evaluation kits from e-peas and TI include reference solar harvesting designs with power management ICs.

Thermoelectric and vibration harvesting are worth reaching for when the application has a natural thermal or kinetic source and solar exposure isn’t practical. For most new IoT deployments in 2026, energy harvesting deserves to be in the design discussion from the start rather than treated as a future enhancement.