Source record: I36
Article edition prepared: 4 November 2026. This is not the publication date of every cited source.
Research horizon: 23 and 27 October 2026; source-specific dates retained in article
External-source research synthesis, not an OfferSignal client result. Based on the retained October 2026 research and source-specific dates.
A welding cell can have a valid computer-aided design (CAD) path and still face a shifted root opening, tack, gap or fit-up error. Reprogramming each variation is slow; trusting an unverified correction is risky. NovAI addresses this engineering-to-weld problem, not the journey from request for quote to invoice.
What NovAI is designed to do
NovAI is described as a welding-intelligence layer. Capture records weld video, parameters and production context; Control lets an operator adjust conditions during a weld; Autonomy is described as using machine vision, welding intelligence and artificial-intelligence logic to adjust the weld automatically; and NovHub brings video, parameters, traceability and production records together. 1 2 3
Novarc describes a torch-adjacent optical sensor and neural-network perception for joint geometry, root shifts, tacks and weld-pool information, with real-time changes to voltage, wire-feed speed, weave amplitude and travel speed. Autonomy claims seam tracking, tack fusion, root-opening, seam-offset and fit-up adaptation. A 19 October 2026 release said Autonomy and NovHub were demonstrated on Yaskawa and ABB robots, including pre-arc root-opening evaluation and live adjustment for gaps, misalignment and tacks. This is a demonstration, not a named customer production study. 5 11
NovPlan imports CAD or CSV weld lists, simulates reach, collisions and cycle time, performs seam finding, plans multi-robot motion and exports ABB, FANUC or Yaskawa code. Novarc gives a vendor example in which 175 weld paths fell from more than 35 hours to 1.5 hours, described as 25 times faster. This is not an independently verified adaptive-autonomy result. 6
What the public outcomes actually show
The strongest named customer evidence concerns the earlier Spool Welding Robot (SWR), not isolated NovAI Autonomy. Alternative Heat's 26 May 2025 report says a rail-mounted SWR handled 125–500 mm pipe and 200 mm–4.5 m welds, adding 700 metres of weld per week across day and night shifts. It gives no controlled baseline, labour-hour, cost or inspection denominator. 8
A Kleeberg trade/customer report says SWR welding was three to four times faster on 8–14 inch pipe; 95% used SWR; output exceeded 250 factored diameter inches per shift; savings were about $25 per 6–8 inch weld and $45 per 12–14 inch weld excluding system cost; and overall time fell 12–15%. The methodology is incomplete; these figures are not NovAI Autonomy results. 9
Novarc also claims more than a decade of deployments, multi-terabyte welding datasets and nearly one million arc-on hours. Public sources do not disclose model versions, drift monitoring, rollback, update frequency or a before-and-after over-the-air experiment. No public named-customer metric for NovAI Autonomy quality or over-the-air uplift was found through 27 October 2026. 4 5
Traceability is useful, not certification
NovHub is described as storing video, parameters, job and part identity, operator identity and production timing, with process-window flags and root-cause review. NovSync adds cloud backup, reports, alerts and usage analytics. These mechanisms support review, not autonomous certification or guaranteed defect detection. No public NovAI result names a customer, sample size, inspection standard or defect denominator. 1 7
Novarc publishes installation, training, site acceptance testing, warranty and support/software-update steps for SWR onboarding, while an OMRON partner case describes the integration architecture without an independent customer outcome. 10 14
An illustrative manufacturer test
For an illustrative custom fabricator, start with one weld family and link nominal CAD, sensed seam, adaptive corrections, video, inspection and job identity. The welding or quality owner remains authoritative for process windows and release.
Test reference parts plus controlled gaps, misalignment, tacks, material or wire changes and calibration faults. Measure path success, arc-on time, rework, defects, interventions and recovery before and after a defined model version. For updates, require release notes, approval, rollback and matched denominators. A demonstration cannot establish return on investment.
Commercial boundary and next reading
NovAI requires an active subscription for upgrades, remote monitoring and support. NovSync Lite is described as free for one SWR and two users, with cloud storage, reports, tracking, videos and logs; integration is not included. Prices for Autonomy, NovHub, NovPlan, higher tiers, hardware, installation and usage are undisclosed. The record does not establish RFQ or configure-price-quote, contracts, logistics, invoicing or aftersales integration. 2 3 7
Sources and next reading
Explore OfferSignal · Focused Business & AI Workshop · All research
References
- [1] NovAI platform
- [2] NovAI Autonomy
- [3] NovAI Capture
- [4] About Novarc Technologies
- [5] Novarc to Showcase Physical AI on Robotic Welding Solutions at FABTECH 2026
- [6] NovPlan
- [7] NovSync
- [8] Innovation at Work: Advancing Welding Operations at Alternative Heat
- [9] Increasing Weld Speed with Novarc Technologies & Kleeberg
- [10] From Order to Operation: Setting Customers Up for Success
- [11] How AI and Computer Vision Are Transforming Robotic Welding
- [12] Miller Electric Advances Adaptive Welding Automation with AI and Robotics
- [13] NovAI to Transform Welding Automation
- [14] Novarc Scales Automated Welding Solutions with OMRON Sysmac Control

