The Three Things That Stopped Building Monitoring at Scale Just Got Fixed at Sensors Converge

I've been deploying sensors in small commercial buildings for two years. The technical problems are interesting and mostly solved. The reasons proposals don't get signed are not technical. They are three boring, structural problems that the industry has been quietly working around with duct tape and willpower.

The Sensors Converge 2026 show floor opened today in Santa Clara. Three vendors announced products that, together, remove all three of those structural problems at the same time. Not in a press-release "transforms the industry" way. In a way you can put on a proposal next week and the math actually works.

Here are the three problems, and what showed up on the show floor today.

Problem 1: "Who's going to change all those batteries?"

This is the question that kills more sensor proposals than any other. A 30,000 square foot commercial building benefits from 40 to 60 distributed monitoring points. HVAC return temps, supply temps, refrigeration cabinet temps, walk-in cooler humidity, mechanical-room CO and CO2, panel current sensors, sump pump float verification, hot water recirc loop temps, freezer thresholds, exhaust fan vibration. None of these individually justifies a per-sensor cloud subscription. All of them together justify a service contract.

And every single one of them, until today, has had a battery in it.

Three CR2032s in a temp/humidity tag. A LiPo cell in a vibration logger. Two AA cells in a leak detector. They last 12 to 36 months and then they don't. The labor cost of a once-a-year battery walkthrough on a 60-sensor deployment is four to eight hours of skilled tech time. At even modest billing rates that's $400 to $800 per year per building, every year, forever, on a service contract that quotes for $2,400 to $4,800. The battery line is meaningful.

It's also the thing that breaks "set it and forget it." A facility manager who agrees to a sensor deployment is implicitly agreeing to a recurring annoyance. They don't say no because of that, but they don't sign quickly either.

What showed up today: Powercast's wireless-power backbone

Powercast came to Sensors Converge with a wireless-power layer designed specifically for distributed edge devices. Their Powerharvester chipsets convert RF energy into usable DC power at distances up to 85 feet (26 meters) from the transmitter. Live demonstrations at the booth: a 1.2-millimeter-thick battery-free Bluetooth Low Energy sensor tag co-developed with InPlay, plus battery-free RFID environmental sensors already deployed in production for server-rack thermal monitoring in data centers.

Powercast's own description of the target applications calls it out directly: "across commercial buildings, industrial facilities, healthcare environments, logistics networks, and data centers… temperature, humidity, occupancy, vibration, asset location, and system health."

30 million Powercast-enabled products are already in the field globally. This is not a vaporware show demo. It's an existing technology stack that has now been productized for distributed sensors specifically.

The implication for a building-monitoring proposal: strike the battery-replacement line. The capex difference between a battery-powered and an RF-harvesting sensor at volume is approaching zero. The opex difference, over a 10-year operating horizon on a 60-sensor deployment, is $4,000 to $8,000 of avoided service-call labor. That's not a marketing number. It's a line item.

Problem 2: "We can't get a signal in the boiler room."

Every building survey includes a moment where I walk into the mechanical penthouse, the basement boiler room, or the back of the kitchen, hold up my phone, and watch Wi-Fi disappear. 2.4 GHz Wi-Fi loses 6 dB through every poured-concrete wall. A typical commercial mechanical room sits behind two of them. By the time the signal reaches the equipment that needs monitoring, it's not a signal, it's noise.

The standard workaround is a Wi-Fi extender, or a powered repeater, or a rats-nest of mesh nodes, all of which require power, network configuration, and a place on the wall that nobody objects to. None of those are zero-effort. All of them are part of why a building survey takes two hours instead of forty minutes.

Bluetooth Low Energy doesn't fix it; BLE has even shorter range. LoRa fixes the range problem but introduces gateway cost and bandwidth limits that make it a poor fit for sensor populations that need 5-second-resolution time series.

What showed up today: Murata's Wi-Fi HaLow modules

Murata is showcasing two Wi-Fi HaLow modules at Sensors Converge: the LBWA0ZZ2HK at +23 dBm and the LBWA0ZZ2HL at +13 dBm. Both run on the IEEE 802.11ah standard — sub-1 GHz Wi-Fi — at up to 1 kilometer of range with PHY data rates up to 15 Mbps. Both are ultra-low power. Both are explicitly targeted at "smart metering, smart building, industrial, and camera IoT applications."

The translation for a building-monitoring practitioner: a single HaLow access point in a normal IT closet covers the entire mechanical envelope of a typical 30,000 square foot commercial building. Not "covers most of it." Covers all of it. At sub-1 GHz, the propagation through poured concrete is dramatically better than 2.4 GHz, and the link budget at 1 km nominal range gives you tens of dB of margin to spend on attenuation through walls.

The building-survey step shrinks. No repeaters. No mesh nodes. No "let's see if we can get the signal to the boiler room." The answer, before you walk in, is yes.

The relevance of module availability from Murata specifically — as opposed to reference designs from chipset vendors — is that the sensor-vendor ecosystem can now move. A company building a tank-level sensor or a vibration logger doesn't need to do their own RF certification; they drop in a Murata module. That's how the 12-month catalog of sensors I'll be deploying in 2027 ends up with HaLow connectivity at the same price point as Wi-Fi today.

Problem 3: "Vibration monitoring sounds great, but the install is brutal."

Vibration is the single highest-value signal in a commercial building. Bearing failure on an HVAC blower, cavitation on a circulator pump, contactor chatter, refrigeration compressor wear — all of them announce themselves in vibration spectra weeks before they cause a service call. Vibration is also the signal where on-device inference matters most: the data rate on a piezoelectric accelerometer at any useful sampling rate is incompatible with cloud-tethered monitoring. You either run the FFT and the anomaly detector locally, or you don't run it at all.

Until recently, that meant a relatively expensive industrial vibration logger ($300 to $1,500 each), an awkward mounting bracket, drilled holes or magnetic mounts that don't hold reliably on painted surfaces, and a separate edge gateway to consolidate the streams. The deployment friction was real. I have skipped vibration monitoring on more than one building proposal not because the value wasn't there but because the per-point install time multiplied through 12 motors made the proposal too expensive.

What showed up today: Upbeat Technology's Tiny AI Engine

Upbeat Technology came to the Sensors Converge startup zone with the UPM01 and UPM02 — vibration processing units that measure 3.2 × 2.5 millimeters. Pair them with the UP201, a dual-core RISC-V AI MCU at 3.0 × 3.0 millimeters, and you have a complete on-device vibration anomaly detection node in a footprint smaller than 2% of a CR2032 coin cell. The company's claim, in its own words: machine vibration analysis with on-device inference detecting "subtle anomalies in real time" to enable "proactive maintenance, reducing unplanned downtime by over 90%."

The 90% number is a vendor claim, not an independent measurement. Treat it as marketing. The sensor format, however, is what matters.

A 3.2 × 2.5 mm sensor adheres to the housing of an HVAC blower, a circulator pump, or a refrigeration compressor with a single drop of cyanoacrylate adhesive. No drilled holes. No bracket. No magnetic mount that fails on a painted surface. The install time per point drops from twenty minutes to two. The cost of the sensor is negligible at quantity. The on-device inference means the data stream off the sensor is compact (an anomaly score and a small vector, not raw accelerometer streams), which means the connectivity layer (HaLow, see Problem 2) doesn't bottleneck.

Vibration monitoring on every motor in a building stops being an upgrade and starts being a default.

Why it matters that all three landed at the same show on the same day

Each of these announcements, in isolation, is incremental. Powercast has been doing wireless power for years. Murata's HaLow modules were introduced last year. Upbeat is a startup with vendor-claim numbers. None of them, alone, transforms anything.

What makes today's combination interesting is that the three structural barriers to wide deployment of building monitoring — battery maintenance, wireless coverage, and per-point install cost on vibration — were all addressed by separate companies on the same show floor at the same time. The supplier ecosystem assembled this answer in parallel without coordinating. That's the pattern of a market tipping. Independent vendors converging on the same set of unlock points, because each of them watched the same friction in the field and built for it.

The EDGE AI FOUNDATION keynote tomorrow morning, "How Edge AI Will Save the World," is the editorial framing of the same shift. The Wednesday panel on "communication-constrained environments" is even more on-the-nose for our use case — that phrase was clearly coined for autonomous robots and drones operating in adversarial RF environments, but it describes a 1950s church basement just as well.

What this looks like in a proposal

The two case studies on this site — the agentic sump pump running 97 recovery cycles in a single rainy night, and the smart-building deployment that cut 42% off energy costs in six weeks — both ran into all three of these problems and worked around them with effort.

The sump pump uses a ZigBee mesh because Wi-Fi didn't reach the basement. It needed two repeaters. The smart building uses CR123A batteries on the door sensors because of mounting constraints; we change them annually. The vibration monitoring on the boiler circulator didn't make the first deployment because the install time on the available off-the-shelf sensors made it cost more than what it would catch in the first year.

Translate today's announcements into the next deployment that lands on my desk:

• Wireless coverage problem: solved by a single HaLow access point in the IT closet. Building survey time on a 30,000 sq ft commercial property drops from two hours to forty minutes. The proposal's RF-coverage caveat goes away.
• Battery maintenance problem: a Powercast transmitter in each major mechanical envelope eliminates 80% of battery line items. The proposal's annual maintenance line drops by $400 to $800 per year. Lifetime savings on a 10-year contract: $4,000 to $8,000. The proposal closes faster.
• Vibration install problem: 3 mm × 2 mm Upbeat-style sensors on every motor in a building add a dozen monitoring points to the deployment at minimal install time. Vibration monitoring becomes table stakes instead of an upcharge. The proposal value goes up because the catch rate goes up.

None of these are speculative. Each is a vendor announcement at a major industry show that opened today. The integration work, in a typical small-building deployment, is a quarter's worth of engineering. By the second half of 2026 the catalog of available sensors will reflect this. By 2027 the proposal template I write will reflect it.

What I'm watching the rest of this week

Three threads to watch as the show runs Tuesday through Thursday:

1. Pete Bernard's keynote tomorrow morning. The CEO of EDGE AI FOUNDATION speaks at 9:30 am Wednesday under the title "How Edge AI Will Save the World." The interesting thing isn't the framing; it's any quantitative claims about deployment volume, cost per sensor, or on-device inference economics that get printed. Those numbers travel into client conversations.
2. Microchip's 3D ToF lidar launch at 2:55 pm Wednesday. Single-package time-of-flight in a small footprint is the cheap-presence-detection layer for occupancy, maintenance access, and safety zones. The pricing point on this part is what determines whether ToF replaces PIR sensors in a building proposal in 2027 or in 2028.
3. The Wednesday afternoon panel on edge AI in "communication-constrained environments." Building-monitoring practitioners are not the marquee customer for that framing, but the framing is exactly right for our use case. If the panel produces quotable definitions or specific use-case math, it goes into the next proposal I write.

I'll write a recap on Thursday or Friday. The show is producing real signal, not just announcements.

The bottom line

For two years I've quoted building-monitoring proposals around three known structural costs: annual battery service, RF coverage workarounds, and vibration-sensor install labor. Today, three independent vendors at one industry show announced the answers to all three.

The interesting question for the second half of 2026 stops being "can edge AI monitor this building?" The answer to that has been yes for two years. The new question is "how thoroughly can we monitor this building, given that the structural costs that capped the sensor count just collapsed?"

Buildings get a lot more sensors next year. The sensors get cheaper, smaller, longer-range, and battery-free. Local inference handles what the cloud can't. The work to deploy this stack is real but the friction is dropping. That's the practitioner's read on Sensors Converge opening day.

Read the case studies and related posts: How edge AI prevented a basement flood | 42% off energy costs on a community center | Cloud latency is a safety hazard for robots, a cost hazard for buildings | The edge AI hardware wave has arrived