Seoul National University's venture arm backs chip-agnostic AI startup Vaistrata
What's the deal? Seoul National University Technology HoldingsDealroom has a profile for this one. Try Dealroom → has made an early-stage venture investment in VaistrataDealroom has a profile for this one. Try Dealroom →, a deep-tech startup building AI software for robots and autonomous vehicles. The company announced the round on September 28; it did not disclose the amount.
What does Vaistrata do? It develops GPU control and machine-learning optimisation technology, aiming to run AI models across a range of chips — Nvidia, Qualcomm, AMD, and ARM — rather than locking users into a single vendor.
The startup applies "Zero-Copy" technology, processing sensor data directly in GPU memory without a separate copying step. That cuts the delay in data movement, letting robots and self-driving systems react faster.
The founders: Chief technology officer Park Han-gil researched GPU rendering at Pearl Abyss and NetmarbleDealroom has a profile for this one. Try Dealroom →. Chief business officer Kim Hyung-kyu, an AI and embedded specialist, previously worked at Samsung TechwinDealroom has a profile for this one. Try Dealroom → and Seoul National University's precision machinery institute.
Why now? The AI market's centre of gravity is shifting from training large models to "inference" — running AI in the field on robots and autonomous vehicles. That is driving demand to use a wider variety of semiconductors.
Why the investor bit: Seoul National University Technology Holdings said Vaistrata's low-level control over GPUs gives it an edge in the market for software supporting diverse AI chips.
"The importance of cross-platform AI software that lowers dependence on specific semiconductors and supports various chips is growing," said Cha In-hwan, chief executive officer of Seoul National University Technology Holdings. (Translated from Korean.)
What's next? Vaistrata plans to focus on building an "AI runtime layer" for executing models and to expand technical validation with robotics and autonomous-driving firms at home and abroad.
The signal: As inference moves to the edge, the fight over which chips run AI is opening room for vendor-neutral software — and Vaistrata is positioning itself in that gap.
Read more: edaily.co.kr
Image credit: Generated with Gemini
