SK hynix writes down AI venture Gauss Labs by $25.3m

SK hynix has booked a 38.2 billion won (US$25.31 million) impairment on Gauss Labs, an AI venture it fully funded.

The write-down followed an assessment of recoverable amount in its Financial Supervisory Service filing.

Gauss Labs, founded in 2020 in Silicon Valley, has relied on SK hynix as its sole customer, and recorded cumulative net losses of 19.0 billion won (US$12.58 million) from 2021 to end-2025.

The move comes as SK hynix leans on Nvidia’s Omniverse platform for its autonomous factory plans, and has said it plans to roll out autonomous factories by 2030.

🔗 Source: Chosun Daily

🧠 Food for thought

Implications, context, and why it matters.

The write-down tracks SK Group’s shift toward Nvidia’s AI factory setup

SK Hynix is stepping back from its in-house AI effort as SK Group puts more weight behind a broader partnership with Nvidia 1.
SK Group and Nvidia plan an “AI factory” (data center) in Korea with 50,000 Nvidia GPUs, with phase one due by the end of 2027 1. The site will support digital twins (virtual replicas of real-world systems) and AI agents (software that can act autonomously), plus cloud services for SK Group affiliates and outside customers 1.
Inside chip manufacturing, SK Hynix runs Nvidia-accelerated software from Cadence, Siemens and Synopsys on Nvidia-accelerated servers to speed computational lithography and physical verification for DRAM and flash memory production 2.
With manufacturing workflows already leaning on Nvidia’s platform for AI-driven production, a smaller standalone venture like Gauss Labs matters less, so the write-down fits that direction.

Big industrial firms are leaning on AI platforms instead of building full stacks

Impairing Gauss Labs while adopting Nvidia’s platform captures the build-versus-buy choice many industrial companies face in AI.
Even for SK Hynix, building a competitive proprietary AI automation stack can be harder to justify than using a dominant full-stack ecosystem 3.
Nvidia is pushing past chips into packaged plans for designing and running AI-driven facilities, which raises the pull of its platform 3.
More software specialists are placing their tools on top of these platforms, which builds network effects and squeezes room for independent offerings 2.