South Korea has switched on Hangang, its sixth national supercomputer, at the Korea Institute of Science and Technology Information (KISTI) in Daejeon. The liquid-cooled machine is not just a speed upgrade over its predecessor Nurion-it is purpose-built to eventually work alongside a trapped-ion quantum processor from IonQ, under a national research push backed by billions of dollars in government funding.
A 24-fold jump in raw computing power
Built on HPE Cray EX400 hardware and Nvidia's GH200 "superchips," Hangang reaches a theoretical peak of 614 petaflops, roughly 24 times the capacity of Nurion, the system it replaces. Each GH200 module fuses an Arm-based Grace CPU with a Hopper GPU over a 900 GB/s link, giving researchers a single 576 GB memory pool instead of shuttling data across a slower PCIe bus. In early pilot tests, that design reportedly delivered up to a 41.8-times speedup over older A100 GPU clusters on memory-intensive tasks such as molecular and materials simulations-gains tied to specific workload types rather than a guaranteed across-the-board improvement.
Laying the groundwork for a hybrid quantum link
Hangang's bigger significance lies in what it is designed to connect to next. Under a trilateral agreement among KISTI, Nvidia and IonQ signed in March 2026, the supercomputer is slated to interface on-site with IonQ's Tempo trapped-barium-ion quantum processor by May 2027, using Nvidia's NVQLink and the CUDA-Q software framework. Tempo's architecture allows any qubit to interact directly with any other, which can reduce the overhead of extra "SWAP" operations often needed on hardware with more limited qubit connectivity. In the planned hybrid setup, Hangang would handle classical tasks-preparing data, updating parameters, post-processing results-while Tempo takes on specific quantum subroutines, such as modeling molecular electronic structures or tackling combinatorial optimization problems.
Part of a larger national research bet
Hangang is positioned as infrastructure for South Korea's K-Moonshot initiative, a program the government is supporting with roughly ₩10.1 trillion (about $7.5 billion) in 2026 AI research funding spread across 12 priority areas, including battery materials, clean energy and quantum computing. After pilot trials involving about 190 projects, the system is scheduled to open for full production use on December 14, 2026, with 15% of its computing capacity reserved for small and medium-sized enterprises and startups.
What remains to be seen
The quantum link is still more than a year away, and performance claims so far come from limited pilot benchmarks rather than independent, large-scale verification. How smoothly classical and quantum systems will actually hand off work-and whether the promised speedups generalize beyond early test cases-will become clearer only as the Tempo integration moves from roadmap to operation.