Neocloud company Vessl AI held the “Vessl AI 2026 Neocloud Strategy Briefing” on Oct. 15 and unveiled a blueprint to secure eight AI data center hubs nationwide by 2028 and introduce 200,000 GPUs in phases by 2030.

Founded in 2020, Vessl AI has targeted the AI-focused cloud market not with its own infrastructure but by working with data center operators in various countries and providing software that connects their infrastructure.

Since its founding, it has provided a platform supporting AI model development, training, deployment and operations. Using GPU resource management and AI workload optimisation technologies built up in the process, it is expanding its business from supplying computing resources to running AI models in live services.

Its main “Fluid Computing” technology supports the use of GPU resources distributed across multiple data centers according to customer requirements and task characteristics. The company explains that while the structure connects the infrastructure of data center partners around the world, users can have an experience similar to using a single cloud.

Vessl AI plans to expand its data center partner network with fluid computing technology, pursue financial cooperation and gradually increase the scale of GPU supply.

In South Korea, its strategy is to establish AI data center hubs in eight regions, considering power and land in the Seoul metropolitan area and outside it, and to build data centers that can accommodate next-generation GPUs. It will also step up building its own GPU infrastructure in 2027.

The company said its sales in the first half of this year came to 130 billion won and it expects annual sales to reach about 250 billion won. The 130 billion won figure includes infrastructure costs paid to data center partners, it said.

Vessl AI plans to pursue global GPU financing worth 400 billion won to secure funds needed for medium- and long-term infrastructure expansion. It aims to secure 10,000 GPUs within this year, 50,000 in 2027 and 200,000 by 2030.

Vessl AI also unveiled an inference service to meet demand for running AI models, beyond GPU supply. Customers can choose to receive computing supplied by tokens, in addition to directly using GPU resources.

Vessl AI has worked with model providers, including providing global inference infrastructure for Upstage’s large language model (LLM) Solar Pro 4. Based on this operating experience, it plans to help model providers focus on research and product development and enable companies to stably deploy AI models in live services.

Vessl AI CEO Jae-man Ahn (안재만) said, “Starting in the field of AI development, Vessl AI has expanded its business into neocloud by solving the problem of securing necessary computing resources and operating them efficiently.” He added, “Based on eight AI data center hubs nationwide, we will grow into a global top-tier neocloud company connecting Korea’s infrastructure and global AI demand.”