Fundraise

Taiwan's KONST raises $30M Series B to scale Asian AI data centres

What's the deal? Taipei-based AI compute operator KONSTDealroom has a profile for this one. Try Dealroom → has closed a $30 million Series B equity round, led by ADATA TechnologyDealroom has a profile for this one. Try Dealroom →. Strategic participants include Mobit Hexie and PegatronDealroom has a profile for this one. Try Dealroom →-affiliated Hoding Venture Capital, alongside other corporate groups. The money will expand its AI data centre build-out across Asia and accelerate its "Token Factory" strategy.

What's the endgame? KONST treats compute as production equipment and tokens as output, delivered across three layers. Konstra AIDealroom has a profile for this one. Try Dealroom → handles data centre construction and GPU cluster operations; Glows.aiDealroom has a profile for this one. Try Dealroom → breaks cluster capacity into per-second billing units; and Horizon AIDealroom has a profile for this one. Try Dealroom →'s (k) ATP Token governs who uses what, and how it is tracked.

Who backed it? The lead investors come from memory and storage, electrification, and electronics manufacturing, with others spanning energy, semiconductors, and thermal management. KONST said these groups are themselves enterprise AI buyers, valuing their fit with real use cases as much as the capital.

Why now? Enterprise AI demand is shifting from single-team trials to constant cross-department use, turning compute from one-off project spend into daily, fluctuating consumption. As agentic AI matures, workloads demand more performance and reliability, raising the stakes for AI infrastructure.

"This round's significance isn't just the funding size, but that we and our investors see the same thing: AI has to genuinely enter enterprise operations, not stop at proof of concept," said co-founder and chairman Chang Yung-hsu (translated from Chinese).

The governance gap: KONST argues the hardest layer to get right is governance — knowing which department used which models, at what cost. Its (k) ATP Token works on a "one key, every model" principle, letting companies connect to authorised models through a single project key, with budgets, permissions, and request logs built in from day one.

The signal: KONST is betting that Taiwan's hardware and supply-chain strength can extend upward — from making compute to delivering it as a managed service. With large enterprises adopting AI across multiple departments, demand for operators that handle infrastructure, billing, and governance in one contract is set to grow.

Read more: inside.com.tw

Image credit: bugeaters

More top stories