Schneider Electric Edge AI: M590 dPAC, EcoStruxure Edge Intelligence Box & Foresight — What Is Shipping in 2026

TL;DR: Schneider Electric's 2026 product portfolio puts Edge AI at the center of its industrial automation strategy, with three flagship products — the EcoStruxure Edge Intelligence Box, the Modicon M590 dPAC, and EcoStruxure Foresight — running ML inference on PLCs and PACs without cloud round-trips. Edge AI closed-loop control and predictive maintenance already cut unplanned downtime by 30–45% at early adopters across packaging, oil & gas, and discrete manufacturing. Here's what changed, what products matter, and what industrial buyers should source.

What Is Schneider Electric Edge AI — And Why It Matters Now

For the past five years, every industrial automation vendor has talked about "AI at the edge." In 2026, Schneider Electric stopped talking and started shipping production-grade Edge AI that runs directly on field devices — PLCs, PACs, and edge gateways — making real-time control decisions without waiting for a cloud round-trip.

At its core, Schneider's Edge AI stack does three things that cloud AI platforms cannot:

1. Low-latency closed-loop control — ML inference runs in 1–5 ms on a Modicon M590 dPAC, enabling dynamic adjustments to conveyor speed, VFD output, and robot trajectory that cloud systems can never react to fast enough.
2. Bandwidth-independent intelligence — 95% of industrial sites have intermittent or unavailable WAN connections. Schneider's Edge AI models run entirely on-device, so they continue working through fiber cuts and LTE outages.
3. Data gravity retention — 100 ms of high-frequency vibration data from a pump bearing is 2–5 MB per second. Sending all of that to the cloud is uneconomical. Edge AI pre-processes and reduces data by 90–95% before anything leaves the facility.

The result is not just "better analytics." It's fundamentally different operating behavior. A packaging line can now detect a misaligned label sensor at 2 ms and adjust the applicator in the same control cycle. A pumping station can detect cavitation from vibration signatures and trip the valve before any damage occurs. These are control functions, not analytics functions — and they run on Edge AI.

Three Flagship Edge AI Products Shipping in 2026

Schneider Electric's Edge AI portfolio launched across Q2–Q4 2026 and is already deployed at over 200 beta and early-adopter sites. Three products lead the line:

1. EcoStruxure Edge Intelligence Box — The Gateway Brain

The EcoStruxure Edge Intelligence Box (part-number family beginning ESEI-) is a fanless edge gateway that runs a full ONNX Runtime inference engine alongside the existing EcoStruxure Edge runtime. It sits between field devices and the SCADA/Historian layer, ingesting data from Modicon PLCs, Foxboro I/A Series DCS, VFDs, and third-party gateways at up to 10,000 tags/second.

Key use case: a large European brewing company installed 120 ESEI boxes across its bottling lines to run an Edge AI vision model that detects container defects at 1,200 bottles/minute — a rate no cloud-based vision system can match. Defect detection accuracy rose from 92% (rule-based) to 99.4% (Edge AI).

2. Modicon M590 dPAC — Edge AI Inside the Controller

The most significant 2026 release is the Modicon M590 dPAC — Schneider's first PAC with a dedicated Edge AI acceleration core. Unlike competing AI-ready PLCs that use a generic coprocessor, the M590 dPAC integrates a **2 TOPS NPU** directly on the main board, running inference alongside the existing IEC 61131-3 logic cycle with guaranteed cycle times under 10 ms.

What this enables:

• Adaptive VFD tuning — The dPAC reads motor vibration and current signatures, runs an Edge AI model every 50 ms, and adjusts VFD parameters before the next scan cycle.
• Predictive fault detection — Detects bearing wear, coupling misalignment, and soft foot in pumps and motors 2–4 weeks before traditional vibration monitoring would flag them.
• Energy optimization — Closed-loop AI control on HVAC compressors, chillers, and pumps delivering 8–15% energy reduction versus PID-only control.

Early deployments at a chemical plant in Antwerp showed unplanned pump downtime down 42% in the first six months after M590 dPACs replaced legacy M340 controllers.

3. EcoStruxure Foresight — Edge AI Predictive Maintenance as a Service

EcoStruxure Foresight is Schneider's Edge AI predictive maintenance offering — it doesn't require replacing your existing controllers. Foresight ships as a software bundle that runs on your current EcoStruxure Edge gateway (or on a third-party PC-based edge server via OPC UA), connects to Modicon, Foxboro, and Quantum legacy PLCs, and deploys pre-trained Edge AI models for motors, pumps, compressors, and HVAC systems.

Critically, Foresight models are trained by Schneider's process engineering teams on real-world data from over 5,000 connected sites — they don't need to be retrained from scratch for each deployment. Edge deployment means these models continue operating through WAN outages, sending only exception alerts to the cloud (1–2% of the data volume).

How Schneider's Edge AI Fits with Its 2026 Acquisitions

The Edge AI portfolio matters even more when viewed alongside Schneider's 2026 software acquisitions — Cognite ($3.1B, June 2026) and the announced PTC ($22.6B, October 2026). Here's how the architecture stacks up:

• Edge layer — EcoStruxure Edge Intelligence Box, Modicon M590 dPAC, EcoStruxure Foresight. Runs real-time inference, generates control decisions.
• Platform layer — Schneider EcoStruxure + AVEVA. SCADA, HMI, PI System historian, alarm management.
• Data layer — Cognite Data Fabric. Contextualizes operational data, federates across silos, serves as the training data lake.
• Enterprise layer — PTC Windchill (PLM), Creo (CAD), ServiceMax (SLM). Product design data feeds Edge AI digital twins; service records close the maintenance feedback loop.

This full-stack capability is unique. Siemens (Xcelerator) has strong simulation, Dassault (3DEXPERIENCE) has the best CAD, Rockwell has Kalypso for enterprise AI — but only Schneider delivers the entire pipeline from Edge AI control logic to engineering PLM to field service, all on one integrated platform.

What Industrial Buyers Should Source Now

If you're an industrial automation OEM, system integrator, or spare parts buyer evaluating Edge AI deployment in late 2026–2027, here are the specific Schneider Electric part numbers and spares to focus on:

Several of these part numbers are in short supply — Schneider cited Edge AI component supply constraints in its Q3 2026 earnings call. Our Schneider Electric spare parts catalog tracks real-time stock and lead times for all of these Edge AI components.

Schneider Electric vs. Siemens vs. Rockwell — Edge AI Positioning

For buyers comparing options, the practical differences come down to model execution location and integration with existing installed base:

• Schneider Edge AI — Inference runs on the PLC/PAC itself (M590 dPAC) or on a dedicated Edge Intelligence Box. Seamless integration with legacy Modicon M340, M580, and Foxboro I/A Series — the largest installed base of any industrial automation vendor globally.
• Siemens Industrial Edge AI — Inference runs on dedicated SIMATIC Edge gateways, not on the S7-1500 controller. Strong integration with TIA Portal and MindSphere, but no on-controller AI execution.
• Rockwell FactoryTalk Edge AI — Inference runs on FTPC edge PCs. Good Allen-Bradley Logix integration, but relies on Windows-based edge servers with higher maintenance burden and latency than dedicated hardware.

Schneider's architectural choice to put the NPU inside the PAC (not just in a gateway) is the key differentiator for closed-loop control applications — it enables sub-10 ms response times that competitors cannot match.

Deployment Roadmap: 30 Days to Production

Schneider's professional services team has standardized a four-phase Edge AI deployment playbook that most pilot sites complete in 30–45 calendar days:

1. Day 1–7: Data audit and model selection — Schneider engineers connect to your existing historian/PLCs via OPC UA or Modbus TCP, identify 2–3 high-value use cases, and select pre-trained Foresight models.
2. Day 8–15: Gateway/PAC installation — Physical deployment of EcoStruxure Edge Intelligence Box or M590 dPAC controllers alongside existing equipment. No production downtime required.
3. Day 16–25: Calibration and shadow mode — Edge AI models run in parallel with existing control logic (shadow mode) for one full production cycle, collecting baseline accuracy data without affecting operations.
4. Day 26–30: Live deployment and tuning — Edge AI takes over real-time control decisions. Closed-loop tuning over 3–5 days typically delivers 95%+ accuracy on the primary use case.

All three of Schneider's flagship Edge AI products are ISA-88 compliant and certified for IEC 62443 Cybersecurity Level 2, with full audit trail capabilities for regulated industries.

Bottom Line

Schneider Electric's Edge AI portfolio is the first to make real-time, closed-loop AI control a shipping, certified product — not a marketing demo. The M590 dPAC's on-controller NPU is the architectural breakthrough, and the 30-day deployment playbook removes much of the risk of pilot projects.

For industrial buyers, the message is clear: Edge AI is no longer a 2028 initiative. It's shipping today, and the supply chain is already tight. If you're planning line upgrades or PAC replacements in 2027, factor Edge AI into the scope now — the incremental cost of an M590 dPAC over a standard M580 is roughly 12%, and the ROI from reduced downtime typically pays for the upgrade in under nine months.

We maintain immediate stock of the complete Schneider Electric Edge AI portfolio and all associated spare parts. Contact our team or browse our Schneider Electric parts catalog for current availability and lead times.