Squeezebits, a startup focused on AI model compression and optimisation, has signed a strategic memorandum of understanding with Silicon Valley AI infrastructure startup Modular Inc., it said on Tuesday.

Modular is a startup that developed the next-generation AI programming language Mojo and the high-performance inference framework MAX. Its enterprise value is $1.6 billion, or about 2.14 trillion won.

The two companies will build integrated AI infrastructure spanning model compression through execution optimisation. Squeezebits is leading a joint project to expand Modular’s MAX framework, previously specialised in existing LLM serving, into multimodal areas such as image and video generation.

The companies also unveiled the results of the cooperation at Nvidia GTC 2026, held in San Jose, California, in March.

Squeezebits CEO Kim Hyung-jun (김형준) said the cooperation means Squeezebits’ compression technology is strategically aligned with Modular’s technology stack, which is in the spotlight as a global next-generation infrastructure standard. He said it would go beyond simple technical cooperation to lay a foothold for entry into the North American market and show innovation that solves cost and performance issues at the same time in real service environments.