Smart Tangibles News Digest #2608
- Yoel Frischoff
- 19 hours ago
- 10 min read
February 23, 2026
Global smart tangibles news from around the world - connected hardware, IoT infrastructure, edge intelligence, standards, and the business models behind long-lived products.
This week, edge AI crosses a threshold: it is no longer something bolted onto industrial boards or sold as a cloud spillover. Two major chip vendors shipped embedded AI acceleration directly into automotive grade MCUs and production ready developer stacks, while IoT Analytics declared the "dashboard era" over and the autonomous operations era underway.
Meanwhile, the security side of the ledger is consolidating fast - Mitsubishi Electric closed its billion dollar acquisition of Nozomi Networks, and Dragos published its annual OT threat report showing adversaries have moved beyond reconnaissance into mapping industrial control loops.
The connecting thread is maturity. Edge AI, fleet security, and enterprise IoT are all graduating from pilot stage infrastructure into baseline expectations. For product teams, that shifts the conversation from "should we add AI or security" to "how do we operate and update these capabilities across a product's full lifecycle."
Cross-Cutting Signals
Edge AI is migrating from board level modules into MCU silicon - STMicroelectronics and Microchip both shipped embedded AI acceleration and production toolchains in the same week, signaling that on-chip inference is becoming a standard MCU feature rather than an add-on board.
Enterprise IoT is leaving the dashboard era - IoT Analytics frames 2026 as the year enterprise IoT moves from monitoring and visualization toward autonomous, agentic operations - a shift that changes what "connected product" means for OEMs.
Industrial security is being absorbed by industrial companies - Mitsubishi Electric's $1B acquisition of Nozomi Networks follows ServiceNow's purchase of Armis, showing that OT security is merging into the operational stack rather than remaining a standalone vendor category.
Threat actors are mapping control loops, not just networks - Dragos reports that adversaries have advanced from scanning OT environments to understanding how physical processes work and how to disrupt them, raising the bar for what product security must protect.
This Week at a Glance
Edge AI, autonomous IoT, and OT security each hit new milestones this week - together they redefine what "production ready" means for connected hardware.
Microchip Technology launches full stack edge AI for MCUs and MPUs - Pre-trained models, unified toolchains, and applications spanning arc fault detection to facial recognition make edge AI a turnkey capability for embedded teams.
IoT Analytics reports enterprise IoT reached $324B and is shifting toward autonomous operations - The market is outgrowing horizontal platforms and dashboard centric deployments, moving instead toward vertical, AI driven, self-optimizing stacks.
Mitsubishi Electric completes its $1B acquisition of Nozomi Networks - The deal gives one of the world's largest industrial automation companies direct ownership of OT/IoT security and visibility tools used across oil and gas, utilities, pharma, and mining.
Dragos 2026 OT report identifies three new threat groups and a 49% rise in ransomware actors - Adversaries are now mapping control loops in industrial infrastructure, moving past reconnaissance toward the ability to disrupt physical processes.
News In Detail
1. STMicroelectronics Stellar P3E: First Automotive MCU with Embedded AI
STMicroelectronics ships the first automotive MCU with an integrated neural network accelerator, pushing edge AI into safety rated, MCU class silicon for vehicles.
Announced February 10, the Stellar P3E is built around four 500 MHz Arm Cortex-R52+ cores and ST's Neural-ART Accelerator, delivering inference at microsecond speeds with up to 30x greater AI efficiency compared to traditional MCU core processing.
The chip uses 19.5 MB of phase change memory (PCM) based xMemory, roughly double the density of embedded flash, and targets multi-function electronic control units (ECUs) that consolidate functions to reduce system cost, weight, and wiring.
For automotive OEMs and tier one suppliers, the P3E means predictive maintenance, smart sensing, and diagnostics can run locally inside the ECU without offloading to more expensive domain controllers or cloud links.
That shifts the architecture: functions that previously required separate boxes or high bandwidth connections can now be co-located on a single MCU, simplifying vehicle wiring and BOM. Engineering samples are available now, with full automotive qualification planned for H2 2026.
For smart tangibles teams beyond automotive, this is a leading indicator. When edge AI becomes a checkbox feature inside MCUs rather than a board level add-on, product managers can plan for on-device intelligence as a standard capability across portfolios. The harder question becomes software: how do you version, validate, and update AI models running on MCU class hardware over a vehicle's 10-15 year service life?
Signals to Watch
Other automotive MCU vendors (NXP, Infineon, Renesas) announcing embedded NPU roadmaps to match ST's first mover positioning.
Tier one suppliers citing Stellar P3E in X-in-1 ECU consolidation strategies that reduce box counts per vehicle.
Toolchain and model management solutions designed specifically for MCU class AI in automotive safety environments.
OTA update frameworks that can handle AI model refresh on safety certified automotive MCUs.
Key Links
2. Microchip Technology Ships Full Stack Edge AI for MCUs and MPUs
Microchip's new full stack edge AI offering turns its MCU and MPU lines into production ready AI platforms with pre-trained models, unified toolchains, and turnkey application templates.
Microchip brings AI into the everyday embedded workflow - high-level summary
Microchip expanded its MPLAB development environment with tools that let engineers add ready-made AI features directly into their firmware - things like arc-fault detection, predictive maintenance monitoring, face recognition with liveness checks, and voice keyword commands.
The key idea is simplicity: teams can prototype on small 8-bit microcontrollers and later move to more powerful chips without changing tools or learning a new AI framework. That continuity means embedded developers can build AI features without needing dedicated machine-learning specialists. AI becomes just another step in the normal development cycle: build, test, deploy, refine.
For product teams, this lowers both cost and risk. When the chip vendor supplies the models, tooling, and documentation as part of its standard stack, adding on-device intelligence stops being a research project and becomes a practical product feature.
The harder problem moves to operations: once devices ship, teams must monitor model performance, decide when retraining is needed, and maintain reliable update pipelines so accuracy holds up over years of real-world use.
Signals to Watch
Adoption metrics from Microchip's MPLAB ML Development Suite, particularly around production deployments versus proof of concept downloads.
Competing full stack edge AI offerings from other MCU vendors (ST, NXP, Renesas, Silicon Labs) aimed at the same embedded developer audience.
Reference designs and application notes that pair Microchip's AI stack with OTA update and model management frameworks.
Real world reliability data from deployed edge AI applications on MCU class hardware in industrial or consumer products.
Key Links
3. Enterprise IoT Hits $324B as Industry Shifts from Dashboards to Autonomous Operations
IoT Analytics frames 2026 as the year enterprise IoT leaves the monitoring era and enters autonomous, AI driven operations, raising the bar for what OEMs must build and support.
IoT Analytics reports the enterprise IoT market grew 13% year over year in 2025 to $324B, with 14% growth projected for 2026. But the headline number matters less than the structural shift underneath it: enterprise IoT is no longer primarily about connecting devices and displaying dashboards. The market is moving toward autonomous, connected operations where AI agents, self-optimizing systems, and prescriptive analytics replace manual interpretation of sensor data.
Platform strategy feels the pressure directly. Traditional horizontal IoT platforms - the "one platform for all use cases" model - are losing ground. Enterprises are favoring vertical specific solutions, tightly integrated stacks, or composable architectures that fit alongside existing IT/OT systems rather than replacing them. Meanwhile, chip vendors are embedding AI accelerators and NPUs directly into MCUs, which moves inference from cloud to edge and reduces dependency on platform middleware.
For smart tangibles product teams, this reframing matters. If your customers are moving from "monitor and alert" to "predict and act autonomously," the value your product delivers must evolve accordingly.
Devices that only send data upstream become commodity sensors. Those that can run local models, make decisions at the edge, and integrate into agentic workflows will command higher margins and stickier customer relationships. The operational requirement is clear: you need update pipelines, model versioning, and observability built into the product, not bolted on after launch.
Signals to Watch
Enterprise RFPs that specify "autonomous operations" or "agentic AI" capabilities rather than generic "IoT platform" requirements.
Vertical IoT platform vendors gaining share against horizontal platforms, particularly in manufacturing, energy, and logistics.
New pricing models that tie recurring revenue to autonomous outcomes (uptime, yield, efficiency) rather than device connectivity or data volume.
Edge AI workloads migrating from gateways to MCU class devices as NPU integration becomes standard.
Key Links
4. Mitsubishi Electric Completes $1B Acquisition of Nozomi Networks
Mitsubishi Electric's billion dollar acquisition of OT/IoT security leader Nozomi Networks marks a structural shift: industrial automation companies now own the security stack, not just the control stack.
Completed January 28, the $1B deal gives Mitsubishi Electric - one of the world's largest industrial automation companies - direct ownership of Nozomi Networks, which protects OT, IoT, and cyber physical systems across critical infrastructure. Nozomi recently surpassed $100M in annual revenue and serves 5 of the top 10 oil and gas companies, 7 of the top 10 pharma manufacturers, 7 of the top 10 utilities, and 4 of the top 10 mining operations. Nozomi will operate as an independent subsidiary, maintaining its brand, leadership, and partner relationships.
This deal follows the same pattern as ServiceNow's $7.75B acquisition of Armis in 2025: OT/IoT security is being absorbed into larger operational stacks rather than remaining a standalone vendor category. But the Mitsubishi deal has a different shape. ServiceNow is an IT service management company that bought into OT. Mitsubishi is an industrial automation company that bought into security. That means Nozomi's visibility and threat detection tools could eventually integrate directly with Mitsubishi's PLCs, factory automation systems, and building management platforms - an operational depth that IT-centered acquirers cannot match.
For smart tangibles teams, the signal is that security is now a first party concern for the companies that build and operate industrial hardware. If your supply chain includes Mitsubishi automation gear, expect security tooling to show up as an integrated feature, not a third party overlay. Product teams that already design for security will find this transition natural. Those that treat it as somebody else's problem will face growing friction in procurement, compliance, and partner integration.
Signals to Watch
Integration roadmaps that connect Nozomi's visibility platform with Mitsubishi's factory automation and building management systems.
Competitive responses from Siemens, Schneider Electric, ABB, and Rockwell - do they build, buy, or partner for OT security?
Procurement requirements that bundle OT security with industrial automation contracts as a single vendor obligation.
Impact on Nozomi's partnerships with competing automation vendors now that it is owned by Mitsubishi.
Key Links
5. Dragos 2026 OT Report: Adversaries Map Control Loops as Ransomware Surges
Dragos reports that OT threat actors have moved beyond network scanning into mapping industrial control loops - understanding how to disrupt physical processes, not just IT systems.
The Dragos 2026 OT Cybersecurity Year in Review, released February 17, identifies three new threat groups (AZURITE, PYROXENE, and SYLVANITE) and documents a major escalation: adversaries are now mapping control loops across industrial infrastructure to understand how physical processes can be manipulated. KAMACITE systematically mapped control loops across U.S. infrastructure in 2025, while ELECTRUM targeted distributed energy systems in Poland. PYROXENE deployed destructive wiper malware against critical infrastructure during regional conflict.
The ransomware picture is equally stark. Dragos tracked 119 ransomware groups targeting industrial organizations in 2025, up from 80 in 2024 - a 49% increase. Manufacturing accounted for more than two thirds of all victims, with 3,300 organizations impacted overall. The report emphasizes that threat groups are operating as coordinated ecosystems: SYLVANITE hands off established footholds to VOLTZITE for deeper OT intrusions, demonstrating specialization and supply chain logic within the threat community itself.
For smart tangibles teams shipping hardware into industrial environments, the risk profile now extends well beyond network perimeters. It is no longer enough to secure network boundaries or apply IT security patterns. If adversaries understand how your control loops work and can manipulate process variables, your product's security posture must include process level anomaly detection, validated control logic, and firmware integrity checks that go beyond traditional patching. Products without these capabilities face growing procurement resistance from CISOs who now own OT risk.
Signals to Watch
RFPs and vendor questionnaires that specifically ask about control loop integrity, process anomaly detection, and firmware attestation.
Insurance carriers adjusting premiums or coverage terms based on OT security posture assessments that reference Dragos threat intelligence.
Product security teams adding OT threat modeling to their development lifecycle, not just IT-centric STRIDE or DREAD frameworks.
Industrial procurement standards that require verifiable security capabilities before devices can be deployed on OT networks.
Key Links
From TheRoad / Smart Tangibles
Previous issue: Smart Tangibles News Digest #2607
Deep dive and case submissions: Smart Tangibles case study submission page - share real world examples of connected products, smart infrastructure, and service backed hardware.
Smart Tangibles book progress: The manuscript is incorporating new chapters on edge orchestration, domain specific IoT platforms (like smart retail), and security baselines for long lived devices.
How to Use This Digest
Treat these stories as prompts for roadmap reviews - where should your next generation hardware assume edge AI, higher security baselines, or Matter and Thread as default plumbing.
Use the “Signals to Watch” bullets as inputs to risk registers and opportunity maps, especially around lifecycle management, platform dependencies, and standards adoption.
Bring one story per week into cross functional discussions between product, hardware, security, and operations to stress test assumptions about stacks and partners.
For strategy and finance teams, map the market and standards trends here against your own unit economics and portfolio bets to see where assumptions are shifting under your feet.
If you find this useful, share it with a colleague or subscribe for future issues at theroadtlv.com.
This news digest is compiled weekly. Register here to receive it in your inbox.
Comments