What happened
Chosunbiz reports that SK Telecom signed separate strategic memorandums of understanding with KG Steel and auto parts manufacturer Connec to pilot a manufacturing-specialized AI agent based on its independent foundation model project A.X K1. Per Chosunbiz, the pilot will run in the second half of 2026 at KG Steel's Dangjin plant cold rolling line and at Connec's casting and machining processes. Chosunbiz reports that KG Steel and Connec have already provided historical process-error reports and equipment manuals to develop a demo agent, and that the partners will share additional high-quality manufacturing process data during the pilot.
Technical details
Chosunbiz reports A.X K1 is a large-scale language model with 519 billion parameters, and that only about 33 billion parameters are activated during inference, a configuration the article frames as improving run-time efficiency for industrial use. Chosunbiz also reports SK Telecom will use on-site feedback and the shared data to improve the demo agent's performance and inference speed, and that commercialization and potential replacement with follow-up models will be reviewed after pilot verification.
Industry context
Editorial analysis: Deploying proprietary foundation models in manufacturing aligns with a broader pattern where industrial firms and domestic telcos pursue in-house or vendor-controlled models to reduce data-exfiltration risk and to tailor models to operational technology (OT) data. For practitioners, this trend emphasizes the importance of robust OT-to-ML data pipelines, domain-specific prompt/agent design, and inference-efficiency engineering in constrained environments.
Context and significance
Editorial analysis: The reported inference-sparsity approach-maintaining a very large parameter count while activating a smaller subset at runtime-mirrors recent industry efforts to balance model capacity with edge or on-premise deployment constraints. Observers tracking industrial AI deployments should view this as one of several vendor-led attempts to bring foundation-model capabilities into regulated or latency-sensitive factory settings.
What to watch
Editorial analysis: Monitor pilot outcomes for concrete metrics such as inference latency, fault-detection accuracy, required data-preprocessing effort, and integration cost with factory control systems. Also watch whether commercial agreements, data governance arrangements, or model replacement plans are published after the pilots conclude, as reported review points in Chosunbiz.
Scoring Rationale #
A technically interesting MOU to pilot a proprietary sparse-activation foundation model in steel and auto-parts manufacturing, relevant to OT-AI practitioners, but MOUs represent a pre-deployment trial commitment at single-company scale rather than a confirmed deployment or sector-wide shift.
Practice with real Telecom & ISP data
90 SQL & Python problems · 15 industry datasets
[Active Residential CustomersEasy](/problems/sql/active-residential-customers)
[Unlimited Fiber Plans 500Mbps+Medium](/problems/sql/unlimited-fiber-plans-above-500mbps)
[Customer Churn Risk AssessmentHard](/problems/sql/customer-churn-risk-assessment)
250 free problems · No credit card
See all Telecom & ISP problems