DeepX Outlines 'Physical AI' Edge Chip Strategy Ahead of CES 2027 Showcase
The AMW Read
Incremental update on a known edge-AI-chip player, adding a concrete efficiency certification and a customer design-win detail ahead of a trade-show showcase, but no new funding, partner, or capability claim.
DeepX Outlines 'Physical AI' Edge Chip Strategy Ahead of CES 2027 Showcase
DeepX (딥엑스) CEO Kim Nok-won joined the Consumer Technology Association's CES 2027 Media Event in San Francisco on September 21, alongside representatives from Waymo, Zoox, and CTA leadership, to discuss AI safety, privacy, and the shift of AI workloads from data centers to real-world devices. Kim argued that continuous data-center expansion faces physical limits — not just power supply but transformer and printed-circuit-board availability — and proposed a split architecture where data centers handle training and complex inference while robots, vehicles, and factory equipment run inference locally. DeepX plans to exhibit products from global customers and industrial deployments using its chips at CES 2027, January 6-9 in Las Vegas.
The positioning builds on DeepX's existing edge AI chip business: its DX-M1 processor powers Sixfab's Alphon X5, winner of a CES 2026 Enterprise Technology innovation award, and the company is expanding into robotics, smart factories, security cameras, and servers. It also follows a South Korean Green Technology certification DeepX won September 14 for four edge inference chips, with KTL testing showing up to 26.25% lower power draw and 20°C cooler operation — the concrete efficiency data behind Kim's claim that field inference can approach roughly 100x data-center efficiency. Per the AI Market Watch index, DeepX has logged three tracked news items in the past 90 days versus one in the prior 90 (name-matched, pipeline-ingested coverage only), consistent with a company building visibility ahead of a major trade-show moment.
For investors and robotics or automotive integrators, the signal to watch is whether DeepX converts design wins like the Sixfab deployment into named enterprise customers at CES 2027 — this update names only "global customers," with no new funding or named partners disclosed beyond its earlier Series D first close. Builders evaluating edge-inference silicon should track whether DeepX's efficiency certifications hold up against rival low-power NPU vendors as physical AI deployments scale.
