ConnX Puts Intel Edge Chips Inside MaestroIQ Operations

ConnX said on Oct. 2, 2026 it will put Intel Core Ultra Series 3 processors under its MaestroIQ platform.

The Plainsboro, N.J., firm is integrating Intel edge computing technology so transit and factory operators can correlate live signals at the site. Joint proof-of-value tests are due to start within 90 days.

That calendar is the whole deal. Intel already sells the chip into industrial PCs. ConnX already sits on voice, camera and operations networks. The bet is that those two layers, run together, turn a local processor into something a dispatcher can use when a station, depot or line goes wrong.

Core Ultra Series 3 Still Needs Software Tenants

Intel’s edge brief for Core Ultra Series 3 describes an embedded industrial computer, the AP-6121, with integrated Arc graphics and up to 180 TOPS of AI acceleration at 25W to 65W. The intended jobs are edge AI, robotics, machine vision and factory automation, which is a long way from a laptop NPU demo.

Hardware partners have already boxed the same family into fanless PCs, COM modules and wide-temp controllers. Those boxes still need software that already talks to cameras, fare gates, radios and plant systems, or they stay as compute on a shelf.

SERIES 3 AT THE EDGE

ConnX is not building that box. It is offering MaestroIQ as the tenant that decides which alerts fire, which tickets open and which cameras a supervisor actually sees.

The Software Layer ConnX Already Sells

MaestroIQ is ConnX’s AI command layer. The company says it can unify IT, OT, network, and applications so operators stop hopping across 15 to 25 or more tools. The pitch is correlation and automated action, not another dashboard that only pages a human.

Buyers can take MaestroIQ 360, a fully managed AIOps service with 24/7 coverage, ticketing and SLA reporting, or MaestroIQ Core, a platform-as-a-service model where ConnX runs the stack and the customer keeps the controls. Both assume ConnX already has a path into the customer’s live systems.

On paper that path is wide. ConnX says it has more than 30 years in communications and security, was named Cisco’s 2024 Managed Services Partner of the Year, and counts 100+ Fortune 500 customers, more than 10,000 sites under management, 500+ communication professionals, 135+ countries supported, and more than 2 million alarms resolved with AIOps.

Intelligence increasingly needs to operate where the physical business happens, across vehicles, factories, stores and critical infrastructure, not just in the data center. Working with Intel’s edge technology helps us move customers from isolated AI pilots toward operational AI they can act on in real time.

Indrajit Ghosh, CEO of ConnX, Oct. 2, 2026 announcement

Ghosh is describing a move the firm has been selling in stages: first unify voice and tickets, then watch cameras and sensors, then put inference next to the asset so the command layer does not wait on a central hall.

90 Days to Start Joint Proofs of Value

ConnX and Intel plan to begin joint proof-of-value demonstrations within the first 90 days of the collaboration. Counted from Oct. 2, 2026, that window closes on Dec. 31, 2026. The companies did not name the first sites, the first workloads or how they will score a pass.

ConnX posted the same notice from its own account the day of the release.

We’re excited to announce a new collaboration with 𝗜𝗻𝘁𝗲𝗹, integrating the 𝗜𝗻𝘁𝗲𝗹 𝗖𝗼𝗿𝗲 𝗨𝗹𝘁𝗿𝗮 𝗦𝗲𝗿𝗶𝗲𝘀 𝟯 𝗽𝗹𝗮𝘁𝗳𝗼𝗿𝗺 with 𝗖𝗼𝗻𝗻𝗫 𝗠𝗮𝗲𝘀𝘁𝗿𝗼𝗜𝗤™.

📣 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗽𝗿𝗲𝘀𝘀 𝗿𝗲𝗹𝗲𝗮𝘀𝗲: https://t.co/gWeT8LWjwW pic.twitter.com/4HdRwnpGOG

— ConnX Inc. (@ConnXInc) October 2, 2026

The useful reading of that post is not the press language. It is that a comms-and-operations vendor is now willing to put a named Intel client processor under the product it already runs for agencies, rather than keeping MaestroIQ as a cloud overlay on other people’s servers.

HOW CONNX GOT TO THIS CHIP

1. Sept. 27, 2024: Wins a SEPTA contract to replace aging Centrex voice with Maestro orchestration, Cisco unified communications and contact center gear.
2. Oct. 30, 2024: Cisco names ConnX Managed Services Partner of the Year.
3. Aug. 18, 2026: Lands a seat on Wisconsin’s statewide enterprise voice contract, 505ENT-M25-ENTVOICE, covering legacy voice transition, cloud telephony, contact center and PBX replacement.
4. Oct. 2, 2026: Says it will run Core Ultra Series 3 inside MaestroIQ and start joint proofs with Intel within 90 days.

The Wisconsin award, weeks before the Intel notice, is the tell. ConnX is still winning phone-system work while it asks the same command layer to host on-site inference.

Live Signals From Vehicles, Cameras and Assets

The combined design is supposed to ingest and correlate feeds that already exist in a transit agency or plant, then process them closer to the gate, vehicle or line instead of shipping every frame to a data center. ConnX’s own transit package, Connected Transit as a Service, already lists the classes of input MaestroIQ claims to see.

WHAT THE STACK SAYS IT WILL WATCH

• Networks and apps: IT, OT and application events that today live in separate monitoring tools.
• Cyber systems: Security alerts tied to the same sites as the operational tickets.
• Vehicles and CAD/AVL: Location, status and radio context from the fleet.
• Cameras and fare events: Station video correlated with validation, gates and agency rules, as in ConnX’s fare-compliance walkthrough.
• Industrial assets: Plant and depot equipment, including HVAC, elevators and lighting on the transit side.

A ConnX fare-compliance demo shows the loop in miniature. MaestroIQ compares movement, payment data, station location, gate activity, timestamps and agency rules, then opens a structured alert for review instead of leaving a clerk to hunt across systems. That is a software problem first. The Intel chip is there to keep that loop local when the video and the rules cannot wait on a backhaul.

Inference next to a bus yard or a packing line is a different market from training large models in a rented cluster. The pairing only pays if those local models change what a supervisor does in the next minute, not if they produce a prettier after-action chart.

SEPTA’s Voice Overhaul Opened the Door

The cleanest picture of ConnX’s real footprint is not a TOPS slide. It is the Southeastern Pennsylvania Transportation Authority job announced on Sept. 27, 2024.

SEPTA, the 6th-largest U.S. rapid transit authority, was running 6,000 Centrex lines across 378 locations. ConnX won the work to collapse that plant into a cloud voice and network service on Maestro, with Cisco unified communications, contact center and zero-trust access in the mix. Ghosh said at the time that Maestro would add AI-driven business insights and passenger sentiment analysis on top of the voice cutover.

That contract is a communications migration with an analytics promise. Connected vehicles were on the wish list. The Intel work is the next ask of the same customer type: once the radios, cameras and tickets share a command layer, run the models at the station so a disruption does not have to round-trip through a central hall.

Cities and operators are the ones who will feel a miss. If the proofs never leave a lab, they keep the same split between the camera vendor, the CAD vendor and the radio vendor. If the proofs hold, they get one queue for safety, service and cyber events, with the compute sitting in the facility they already staff.

Intel’s Edge Campaign Reaches Past One Partner

On May 31, 2026, Intel said the Series 3 family had more than 130 edge design engagements in AI and edge computing. The same briefing introduced OpenVINO Physical AI, an open-source robotics library with a silicon-tuned runtime, with general availability listed for 2H 2026, and Physical AI Studio for data collection, fine-tuning and export of vision-language-action models.

Physical AI models are transforming robotics, but deployment has been slowed by fragmented software stacks and one-off integrations for every robot. With Intel Core Ultra Series 3 and OpenVINO Physical AI, we provide a unified, open, and scalable path from AI experimentation to production-grade robots delivering hardware-accelerated, high-performance inference.

Dan Rodriguez, Corporate Vice President, Edge Computing Group, Intel, May 31, 2026

The showcase design was Sensory AI’s Ella store robot, which Intel said moved from a split CPU-plus-accelerator layout onto a single Core Ultra Series 3 SoC running three agents at once. That is retail robotics. ConnX is the other half of the same campaign: find vertical software that already has operators, and park the SoC where those operators work.

Intel did not issue its own quote in the ConnX notice. The software firm is carrying the announcement, which fits a partner-led motion more than a silicon launch.

WHAT WE KNOW

• The hardware: Core Ultra Series 3, rated up to 180 TOPS in Intel’s edge brief, is the named platform under MaestroIQ.
• The clock: Joint proofs of value are due to begin within 90 days of Oct. 2, 2026, which is Dec. 31, 2026.
• The tenants: Transportation, manufacturing, retail and other live operations ConnX already names as MaestroIQ sites.

WHAT IS UNCONFIRMED

• First sites: No agency, factory or depot is named for the joint tests.
• Pass criteria: No latency, uptime or incident-time target has been published.
• Intel’s commercial role: No Intel executive is on the ConnX notice, and no SKU, price or supported OEM box is listed.

Until those three items land, the collaboration is a design win on a slide, not a load on a rail yard.

Incident Response Is the Test That Counts

Point products already sell either a rugged PC or an AIOps console. ConnX is arguing that the two should ship as one path: Intel’s on-site acceleration plus MaestroIQ’s shared layer, so a fare jump, a stalled train or a hot bearing shows up as a single, ranked event. The company says that mix should help operators detect and respond faster across transport, assets and infrastructure. It has not yet shown a timed incident against a baseline.

The next fact to watch is narrow. By Dec. 31, 2026, ConnX and Intel either start a proof on a live network they already touch, or they do not. TOPS will not decide that date. A dispatcher’s queue will.