DeepX Proposed Physical AI to Bypass Data Center Limits
DeepX has announced plans to showcase low-power AI chips in 2027 to address escalating energy and infrastructure constraints.
Updated on Sept. 23, 2026 in Semiconductors
Live Poll
Do you believe physical AI solutions will successfully address global data center power and infrastructure constraints?
DeepX CEO Kim Nokwon recently detailed a strategy using physical AI to reduce power demand, which he claims can reach efficiency levels 100 times higher than current data centers. The development is currently in the research and promotional stage, with product demonstrations slated for 2027.
Why it matters
The rapid expansion of AI is currently bottlenecked by power supply, transformer availability, and printed circuit board constraints. DeepX aims to shift the compute burden from centralized data centers to edge devices to resolve these industrial limitations.
DeepX is developing low-power AI semiconductors designed to operate at 100 times the efficiency of traditional data center hardware. These chips aim to mitigate heat, cost, and power supply bottlenecks within customer products.
The players
Kim Nokwon
The CEO of DeepX who is advocating for physical AI solutions to manage power efficiency.
DeepX
A semiconductor company focused on developing low-power, edge-native AI processing units.
Tekedra Mawakana
The co-CEO of Waymo who is scheduled to deliver a keynote speech at CES 2027.
The details
DeepX utilizes specialized semiconductors designed to run AI workloads directly on hardware, bypassing the need for remote cloud connectivity. By focusing on low-power architectures, the company seeks to manage the heat and cost issues currently hampering large-scale AI deployment. This approach treats power supply and physical infrastructure as the primary technical constraints, rather than just raw computational throughput.
Timeline
September 21, 2026: Kim Nokwon attended a CTA media event in San Francisco.
September 23, 2026: DeepX announced the CEO's appearance at the CTA media event.
2027: DeepX will exhibit its low-power AI semiconductors at CES 2027.
The Tech Race
DeepX is challenging the current data center AI hardware stack by proposing a shift toward high-efficiency physical AI. This effort aims to bypass the infrastructure constraints defined by the modern cloud-centric computing model.
The development of these low-power chips aims to enable more robust AI features directly on consumer hardware. Users should look for product performance updates at CES 2027 to see if these components move from prototype to commercial integration.
The takeaway
The trajectory of AI hardware is moving away from massive data centers toward power-efficient, edge-ready silicon. Watch for the 2027 CES exhibit, where DeepX intends to demonstrate whether their efficiency targets translate into practical, deployable consumer components.
Further reading
For more on the current industry focus, see Semiconductors.
Source note: This article includes information reported by 조선일보.
Live Poll
Do you believe physical AI solutions will successfully address global data center power and infrastructure constraints?
