South Korean chip developer DeepX plans to spend more than $100 million developing its DX-M2 AI chip after sending the completed design to Samsung Electronics’ foundry unit, The Korea Herald reported on 13 November. DeepX announced the design milestone on 10 November, The Elec, a South Korean electronics trade publication, reported. Samsung is to fabricate the chip on SF2P, its second-generation 2-nanometre process.
Key points
- DeepX’s planned spending of more than $100 million covers development, performance verification and production engineering.
- The DX-M2 design has reached mask tape-out for prototype production on Samsung’s SF2P process.
- DeepX expects first silicon in March 2027 and plans to provide customer evaluation samples in the second quarter.
- The company forecasts $40 million in 2026 revenue, more than double its previous $17 million target.
DX-M2 moves from design to masks
The projected spending includes performance verification and production engineering, according to The Korea Herald. DeepX completed mask tape-out for prototype production in early November, The Elec reported. That step transfers the finalised design data to the foundry and precedes photomask production. It moves the project towards physical prototypes, which DeepX expects to receive in March 2027.
DeepX develops chips for AI workloads outside data centres. Its earlier DX-M1 chip entered mass production in August 2025 using Samsung’s 5nm process. The DX-M2 is planned for Samsung’s second-generation 2nm process, placing the new design on a different manufacturing technology from its predecessor.
As of October 2026, DeepX had secured 88 purchase orders worth about $13 million across 10 countries. Those orders sit alongside the company’s plans to spend more than $100 million on developing the DX-M2, rather than describing sales of the new chip.
DeepX targets 5-watt AI workloads
DeepX says the DX-M2 is designed to run generative AI workloads at around 5 watts. The intended workloads include large language models and vision-language models, which process images alongside text. The company is targeting robotics and industrial systems, as well as AI-enabled personal computers.
The planned uses put the DX-M2 in equipment that runs AI work on the device rather than sending it to a remote data centre. DeepX’s stated power figure describes the chip’s design target. Customer evaluations are scheduled to follow the arrival of first silicon.
The DX-M2 received a CES 2027 Innovation Award in the AI category. DeepX had previously received awards in two categories at CES 2026.
DeepX forecasts $40 million in 2026 revenue
DeepX forecasts $40 million in revenue for 2026, more than double the $17 million target it set previously. The company recorded 3.3 billion won in annual revenue and an operating loss of 20 billion won in 2025, The Elec reported. Its DX-M2 spending plan covers development through performance checks and production engineering, while its timetable extends beyond the current revenue forecast.
DeepX expects its first DX-M2 silicon in March 2027 and plans to supply evaluation samples to customers in the second quarter. The Korea Herald reported that the company expects mass manufacturing in 2028. The Elec reported initial mass-production shipments scheduled for the fourth quarter of 2027, with full-scale production and an expansion of global supply planned from 2028.
.jpg){kind=link}