Ambient IoT in 3GPP Release 19: When Battery-Free Tags Beat RFID

What did 3GPP Release 19 actually standardize for ambient IoT?

3GPP finished its ambient IoT study in TR 38.848 and carried the work into Release 19, which froze in 2025. The work covers two device classes. Device 1 has no energy storage and communicates by backscatter, reflecting and modulating a carrier signal on roughly 1 microwatt of harvested power. Device 2a adds a small energy store and amplification, drawing a few hundred microwatts. Both target indoor deployments, data rates on the order of kilobits per second, and ranges in the tens of meters. Two read topologies made the cut: a base station interrogates tags directly, or a phone or other UE acts as an intermediate reader between tag and network. What this means in practice: a tag with no battery, priced to be disposable, that a cellular network can inventory without portals or handheld sweeps. Expect silicon and network support to trail the spec by a year or more, as they did for RedCap. Backscatter just joined the cellular roadmap.

When do battery-free tags beat passive RFID?

Passive UHF RFID only speaks when spoken to. A tag answers during the seconds it sits in a reader field, at a dock door portal or under a handheld sweep, and everything between those moments is invisible. Battery-free ambient IoT tags flip the model: they harvest RF or light, then beacon on their own schedule to any gateway in range. That changes what you can see. Wiliot already ships battery-free Bluetooth tags that report identity and temperature continuously, which is why early adopters cluster in cold chain and reusable packaging. Pick ambient IoT when:

• events between scan points carry cost, like a pallet parked in the wrong staging lane for four hours
• you need sensor data, temperature or humidity, riding on the same tag as the identity
• you cannot install portals at every transition, and ceiling gateways on a 20 to 30 meter grid are easier to justify

Choose ambient when the gaps cost you money.

When does passive RAIN RFID still win?

RAIN RFID is a 20-year-old ecosystem operating at a scale ambient IoT will not touch for years. The RAIN Alliance reports tens of billions of UHF tag chips shipped per year, and that volume is why an inlay costs 3 to 5 cents at quantity. EPC Gen2 readers pull hundreds of tag reads per second, so a loaded pallet clears a dock door in one pass. Auburn University's RFID Lab has documented apparel retailers moving inventory accuracy from roughly 65 percent to above 95 percent with regular RAIN cycle counts, and retail mandates from Walmart and others assume that exact stack. If your items are high volume, low value, and mainly need counting at known transitions, receiving, dock doors, point of sale, passive RFID remains the right answer. Ambient tags cost several times more per unit and only earn that premium if you actually use the continuous signal. At five cents, RFID is hard to beat.

What does item-level tracking really cost per tag?

Compare systems, not stickers. RAIN RFID pairs a 3 to 5 cent tag with expensive read points: a fixed reader with four antennas typically lands between $2,000 and $5,000 installed, and handhelds run $1,500 to $3,000 each, plus the labor of walking them. Battery-free ambient IoT inverts it: the tag costs tens of cents rather than single cents, but the read layer is a mesh of BLE gateways under $200, existing access points, or, in the Release 19 version, the carrier network itself. So run the math on trips, not units:

• Single-trip items (apparel units, cartons tagged once and sold): tag cost dominates, RFID wins
• Multi-trip assets (totes, roll cages, returnable pharma shippers): a 30 cent tag amortized over 50 trips costs under 1 cent per trip, and the gateway mesh is a fixed cost
• Sensor-bearing loads (cold chain): the temperature log alone can justify the tag

Amortize the tag over trips, not units.

How do you run an ambient IoT warehouse pilot?

An 8 to 12 week pilot answers this better than any vendor deck. Weeks 1 to 2: pick two zones with different RF conditions, one open staging area and one dense racking aisle, and record your baseline, current cycle count accuracy and time to find a lost item. Weeks 3 to 4: tag 1,000 to 5,000 items or totes with battery-free tags and mount gateways on a 20 to 30 meter grid. Weeks 5 to 10: measure four numbers: read rate per zone per hour, time to detect a deliberately misplaced item, zone-level location accuracy against ground truth, and read reliability at the bottom of dense stacks, where harvested energy is weakest. That last one is where pilots fail quietly. Metal and water absorb the carrier that powers the tag, the same physics that hurts UHF RFID. Weeks 11 to 12: price the full rollout from measured read rates, not datasheet range. Pilot both, and let the read logs decide.

Ambient IoT tells you an item exists. Where is it?

A Release 19 ambient IoT read tells you a tag responded to a specific reader, which places the item within tens of meters. That is presence, not position. The spec includes no fine ranging, so an ambient tag will not tell you which shelf, which bay, or which cart, and a passive RFID read says nothing beyond the antenna zone that heard it. Most item-level programs end up layered: cheap tags for identity at scale, plus a positioning layer, UWB ranging, BLE, or camera-based tracking, to resolve location down to the bay or bin where labor actually gets spent. That layer is where the hard, patented engineering lives. Position Imaging licenses hundreds of granted patents across RF ranging, computer vision, and machine learning tracking, IP cited by Apple and Bosch, so a team building on ambient IoT or RFID can add the positioning layer in months, with freedom to operate, instead of re-deriving it. Presence is not position. Plan for both.