The project and the stated timeline

DeepX Co., a South Korean startup, plans to spend more than $100 million to develop a next-generation artificial intelligence chip using Samsung Electronics' 2-nanometer manufacturing process. The Korea Herald reports this, citing a company statement. According to the company, the figure includes the costs of performance verification and production engineering.

According to DeepX, the design of the DX-M2 processor has been completed and the final designs have been sent to Samsung's foundry division for manufacturing. The startup expects to obtain first silicon by March 2027, then move to customer evaluations and mass production in 2028.

Target markets and technical profile

The DX-M2 is intended for robots, industrial equipment, and AI-enabled personal computers. According to the company, the chip is designed to run generative AI models with power consumption of about 5 watts.

The investment represents a significant bet by DeepX on running models directly on devices, rather than on the computing power of remote data centers. According to the company, the approach can reduce energy consumption, costs, and delays associated with the continuous transfer of data to the cloud.

Why it matters and what remains to be verified

The choice to focus on processing outside data centers comes amid growing criticism of the latter for their massive energy and water consumption. DeepX's semiconductors are intended for AI applications that do not require the infrastructure of large computing centers.

It remains to be verified in the field how well the claimed 5-watt consumption holds up in real-world applications and how many of the stated timelines are met: first silicon by March 2027, customer evaluations, and mass production in 2028. The material does not indicate production volumes, customers already acquired, or the amount of the investment covered by sources other than the company itself. No details are available on the terms of the agreement with Samsung foundry beyond the submission of the final designs.

Source and attribution: this original summary was prepared by AICronaca based on the report by The Korea Herald ↗. The statements are attributed to the source; any information limitations: The information comes from a DeepX statement reported by The Korea Herald/Bloomberg; there are no independent confirmations on costs, chip performance, or production timelines. The material also includes headlines about Nvidia and other news unrelated to the DeepX project, which were not used in this article..