Darren Su, CEO of Horizon AI and co-founder and chief strategy officer of KONST, joined episode 286 of the online program Black Swan Academy, hosted by Benny Hwang (黃齊元) of Blue Ocean Asia (藍濤亞洲). On the theme of how AI compute gets integrated, he walked viewers from the group's first compute center, converted from a crypto-mining farm, to a footprint that now spans the United States, Japan, Korea, Vietnam, and Malaysia.
Key points
-
It began as a business-model experiment. KONST's core team ran crypto-mining farms, whose hardware, built around GPUs, closely resembles an AI data center. In 2023, before the AI boom reached Taiwan, the team converted a mining site in Banqiao into a compute center for about NT$32 million (US$1 million) to test whether the compute would actually rent. It did, and long-standing supplier relationships from the mining years meant chip supply was never a bottleneck as they scaled.
-
One group, three roles. KONST handles everything hardware: building, hosting, and operating AI data centers. Glows.ai is the GPU cloud platform, serving large model-training clients on bare metal under two- to three-year contracts and smaller teams on pay-as-you-go. Horizon AI is the enterprise adoption partner, combining the group's compute strength with vetted applications to run diagnosis, integration, and long-term support through a single point of contact.
-
A hospital, not a specialist clinic. Darren likens Horizon AI to a hospital: clients arrive saying only that something is not working, and the team diagnoses the need, assembles the right specialists, and plans a six-month to one-year course of treatment. Horizon AI does not compete with the applications in its ecosystem; it integrates hardware, compute, and applications into a solution enterprises can afford and actually use.
-
The token economy as a new utility. Darren describes tokens as "the new traffic fee of this era, a telecom-scale business model." He is candid that Taiwan's compute is called sufficient not because supply is abundant but because real usage has yet to catch up: overseas, some firms already set engineers a monthly compute-consumption target as a KPI, a culture only now emerging in Taiwan.
-
Power limits pushed the group offshore. New northern-Taiwan sites are capped at around 5MW, while US peers discuss 1GW to 2GW. KONST expanded early into Japan (tax and power incentives, existing trading-house ties), Korea (the strongest AI momentum at international trade shows), Vietnam (thousands of AI engineers), and Malaysia (low-cost power for large facilities). Roughly 70 to 80 percent of the group's compute footprint now sits overseas. Darren closed by naming curiosity as what really drives the company forward, and the goal as taking a key position early as tokens become the era's foundational cost.
Horizon AI is the enterprise AI adoption partner within the KONST group, turning the group's strength in compute and infrastructure into AI that enterprises can actually use, from assessment and custom development to system integration and ongoing operations. Book an AI readiness assessment →
Read the full article in Chinese: link
Start with a free AI readiness assessment
In one hour, we map your data, systems, and use cases and give you an actionable AI adoption blueprint.
Book an assessment