Acryl Brings GPU Operations Software GPUBASE to North American AI Infrastructure

KOSDAQ-listed Acryl presents 1,272-GPU verification at UKC 2026 in Orlando, visits University of Florida HiPerGator and meets NVIDIA co-founder Malachowsky

LAS VEGAS, NV, UNITED STATES, September 2, 2026 /EINPresswire.com/ — Acryl Inc. (KOSDAQ), an AI infrastructure software company, has begun its entry into the North American market with GPUBASE, its GPU operations platform. The company took part in the 39th US-Korea Conference on Science and Technology (UKC 2026), held August 5–8 in Orlando, Florida, as the only Korean company on the panel of the official Korea-US AI cooperation symposium and delivered the anchor presentation at the official sponsor forum.

GPUBASE is software that pools GPUs and NPUs from different manufacturers into a single resource. It partitions a single GPU so that multiple jobs can share it and reduces network bottlenecks, raising the realized performance of equipment already purchased. Rather than buying more hardware, the product is aimed at getting users the performance they paid for. With large AI data center builds under way worldwide, demand for GPU operations and efficiency software is growing alongside them.

University of Florida HiPerGator visit and on-site verification discussions
During the conference Acryl visited HiPerGator, the University of Florida’s supercomputing facility, to discuss cooperation with its operations team, and held a roundtable with SelectFlorida, the state’s economic development organization. HPC operations cooperation, on-site verification, and siting and incentives were the three threads of discussion.
On the morning of August 7, immediately after the plenary session, Acryl’s management met NVIDIA co-founder Chris A. Malachowsky, a plenary speaker, and introduced the company and GPUBASE. Acryl’s entry point in the North American GPU operations software market is heterogeneous accelerator environments rather than single-vendor ones, and Ethernet-based clusters rather than InfiniBand. Universities and public research institutions are the first target; the company plans to use references secured there to extend into data center and cloud operators.
Ko Sam-seok, distinguished professor at Dongguk University and a member of Korea’s National AI Strategy Committee, who gave the keynote on “Co-evolution” at the forum, said: “The center of gravity in AI competition is shifting from securing models to operating compute resources. A Korean software company working through verification together with US universities and research institutions on their real operating problems is a realistic path into the local market.”

1,272 GPUs verified: utilization above 90%, training time cut by up to 96%
The evidence Acryl presented in North America is the result of the K-Scale Evaluation, a performance verification completed in the first half of 2026. In tests run independently on 1,272 GPUs of seven heterogeneous types across the three major global cloud environments, each with a different network architecture, GPU utilization rose above 90%, training time under high load fell by up to 96%, and job queue wait time fell by up to 93%. Six items were checked: network performance, GPU sharing and resource utilization, heterogeneous accelerator integration, scalability, storage integration, and fault recovery. NCCL all-reduce communication bandwidth, a known bottleneck in distributed training, improved by between 141% and 2,375% depending on load level. Fault detection and recovery completed within five minutes. As the next step Acryl plans a performance test at single-cluster scale of 1,000 GPUs and will seek a KOLAS-accredited test report.
Cost effects are presented as the company’s own estimates. On a basis compiled from published papers and commercial cases, network total cost of ownership (TCO) on a 512-GPU NVIDIA H100 cluster comes to 48% of the InfiniBand figure, a saving of about US$2.2 million. Raising average GPU utilization from 55% to around 85% would also cut the GPU purchase cost needed for the same throughput by about 35%.

US$4 million cumulative supply to Hancom’s modular data center project
Commercial supply continues in Korea. In July Acryl signed a contract to supply GPUBASE to the modular data center (MDC) construction project led by Hancom. Following the basic tier, an advanced-tier contract worth KRW 3.3 billion (US$2.4 million) was added on July 29, bringing cumulative supply for the project to KRW 5.5 billion (US$4 million). The advanced tier runs on NVIDIA H200 NVL GPU servers and direct-liquid-cooled racks, with delivery scheduled to complete in the third quarter. On July 24 the company signed a memorandum of understanding with Megazone Cloud to extend its coverage to large AI data centers and distributed AI infrastructure. Earlier MOUs with MangoBoost in May and FuriosaAI in July widened the verification scope to Korean-made DPUs and NPUs.

In June Acryl was selected as lead organization for a project commissioned by Korea’s Institute of Information & Communications Technology Planning & Evaluation (IITP) to develop an Ethernet-based GPU cluster network fabric. With a total budget of about KRW 6.7 billion (US$4.8 million), including KRW 5.5 billion in government funding, the project runs through December 2028 jointly with Yonsei University, Sungkyunkwan University and Ajou University. Its goal is to replace InfiniBand, NVIDIA’s proprietary networking technology, with open Ethernet.

Park Woe-jin, CEO of Acryl, said: “Competition in AI infrastructure is moving from how many GPUs a company has secured to how efficiently and reliably it operates the compute it holds. We will make joint verification and business cooperation concrete on the basis of the operating problems North American universities, research institutes and data centers are actually facing.”

About Acryl Inc.
Acryl is an AX (AI transformation) infrastructure software company founded in March 2011 and listed on the KOSDAQ in December 2025. Starting from emotion recognition and natural language processing technology, it now supplies industry-specific AI models together with the infrastructure software that runs them. Its main products are the GPU operations platform GPUBASE; the integrated AI platform Jonathan, which bundles GPUBASE with ModelBase and AgentBase; the industry-specific foundation model family ALLM; and the healthcare AI solution NADIA. Headquarters: 704 Seolleung-ro, Gangnam-gu, Seoul, Korea. www.acryl.ai
Won-dollar conversions at US$1 = KRW 1,390 (August 2026). Cost-saving figures are Acryl’s own estimates.

Jerry (Jaehyuck) Yang
Acryl Inc
+82 10-9949-0296
jerry2@acryl.ai

Media gallery