Exascale Labs signs $71.4m GPU compute deal with Dimension AI
Exascale Labs has entered a three-year Compute Services Agreement with Dimension AI Pte. Ltd., a Singapore-based enterprise technology and distribution company, under which Exascale will procure approximately $71.4 million of dedicated GPU compute capacity. The deal is intended to bolster Exascale's GPU-as-a-Service and token factory platform as enterprise demand for high-performance AI compute continues to outpace available supply.
Hoansoo Lee, chief executive of Exascale, said the agreement gives the company long-term access to the capacity it needs to serve customers running large-scale AI workloads, including LLM training, fine-tuning and high-concurrency inference. Lionel Peh, director of Dimension AI, described strong and growing market demand for dedicated GPU infrastructure. The release did not disclose which GPU hardware is covered by the agreement, the pricing structure, or any named customers that will benefit from the expanded capacity.
Deal context
The contract is structured as a procurement agreement under which Exascale sources capacity from Dimension AI for resale through its own platform. Exascale operates an asset-light model: it does not own the underlying data centres but layers cluster management, optimisation software, and related AI infrastructure tooling on top of third-party hardware. The company says it has also developed modular data-centre, high-density cooling, HVDC power and energy storage solutions aimed at addressing deployment bottlenecks in AI infrastructure build-outs.
Worth noting is that Exascale is simultaneously completing a SPAC merger with D. Boral ARC Acquisition I Corp. (BCAR), after which the combined entity is expected to trade on Nasdaq under the ticker "XLAB". The timing of this announcement, shortly before an anticipated public listing, is consistent with a pattern of milestone releases that support a pre-IPO narrative. Investors and analysts will be tracking whether contracted revenue translates into recognised income once the business combination closes.
Market and competitive context
The GPU-as-a-Service market has expanded rapidly as enterprises seek burst capacity for AI workloads without the capital commitment of owned hardware. Hyperscalers including AWS, Microsoft Azure and Google Cloud each offer managed GPU compute, but a tier of specialist providers, among them CoreWeave, Lambda Labs and a number of smaller asset-light operators, compete on price, availability windows and flexibility for dedicated reserved capacity.
Supply constraints on high-end accelerators remain an industry-wide constraint. Export-control restrictions administered by the US Bureau of Industry and Security limit shipments of advanced GPUs to certain jurisdictions, which has shaped how Asian-based intermediaries structure compute procurement and distribution arrangements. Dimension AI operates out of Singapore, a jurisdiction that has so far retained access to advanced compute hardware, though the evolving BIS framework warrants monitoring by enterprise buyers with cross-border infrastructure footprints.
The three-year tenor of the agreement provides Exascale with revenue visibility, but also concentrates supplier risk on a single counterparty. Enterprise customers evaluating GPU-as-a-Service providers typically scrutinise uptime guarantees, SLA structures and the financial stability of the underlying supply chain, none of which were disclosed in this release. As the market matures, transparency on those terms is becoming a competitive differentiator alongside raw capacity availability.