EuroHPC signed a EUR 29.76M contract with the relaunched Bull to deploy the Mimer AI Factory in Linköping, Sweden. At roughly a tenth the budget of Italy's IT4LIA, Mimer is less a new flagship machine and more a services envelope.
VAST Data closed a $1B Series F at a $30B post-money, 3.3x its 2023 mark and roughly 1.7x Everpure's public cap. Here's what the math and the customer list actually signal.
Samsung, SK hynix, and Micron converged on SOCAMM2 mass production within six weeks for NVIDIA's Vera Rubin. Korean suppliers now control both memory tiers.
ISC 2026 research on LUMI, Leonardo, CRESCO8: Slingshot held performance; InfiniBand collapsed 5x under Incast, the AI gradient-sync traffic pattern.
HBM scarcity has moved beyond semiconductor supply into system planning. Accelerator availability, server bill-of-materials, cluster economics, and 2026 data center buildouts are all being rewritten around memory - not compute.
Broadcom will supply Anthropic with 3.5 GW of Google TPU capacity through 2031; ~23-35x the power of DOE's largest planned science supercomputer.
Two new expert-parallel efforts point to different futures for MoE systems: one built for heterogeneous fleets, the other folded into NVIDIA’s stack.
A planner/executor agent hierarchy drove a 5,591-MOF screening campaign across 256 Aurora nodes. Orchestration overhead came in under 90 seconds per run.
DOE's SYNAPS-I targets unified AI analysis across seven beamline facilities. Can it coordinate deployment or will it fragment like existing implementations?
Iranian drone strikes took down AWS availability zones in Dubai and Bahrain, forcing a global reassessment of data center siting.
Supercomputing News is the first trade publication built for both human readers and AI agents. Agents discover articles through a machine-readable manifest and pay per-request in USDC on Base; no accounts, no API keys, no session management.
GET /llms.txtGET /articles/blackwellX-Payment: <signed>200 4,812 words
DARPA's new HARQ program funds 19 teams to build a heterogeneous quantum computer that mixes trapped-ion, neutral-atom, superconducting, and diamond qubits over a shared photonic backbone, targeting a 1,000x cut in physical resource demand.
The World Quantum Day launch adds open models for calibration and QEC decoding, but the bigger move is NVIDIA tying AI inference to CUDA-Q, CUDAQ-Realtime, and NVQLink in the path to fault tolerance.
Led by imec and backed by the Chips Joint Undertaking, SPINS aims to prove that semiconductor fabs, not research labs, are where quantum computing finally scales. Ten countries and three silicon platforms, all betting on spin.
Two acquisitions, two days apart, at adjacent layers of the same stack. Credo's $873M cash-and-stock deal for DustPhotonics and Molex's Teramount buy tell the market the copper-era interconnect champions have decided the AI factory's fiber layer isn't something they're willing to source.
Samsung launched a silicon photonics foundry platform at OFC 2026, targeting 2028 PIC mass production and 2029 CPO services, pitching vertical integration of HBM, logic, packaging, and photonics against TSMC's production lead
The Chiplet Summit marked the transition from "chiplets as concept" to "chiplets as product." The winners won't be the ones with the best silicon. They'll be the ones with the best software and the deepest packaging ecosystem.