Scaleway says targeting 60 to 70 percent GPU cluster utilization is the sweet spot for balancing token costs against infrastructure capacity.
Read articleNews
HPC News Outlets
The daily sweep across the trade press: HPCwire, The Next Platform, insideHPC, Data Center Dynamics and ServeTheHome, rewritten in plain language with a straight link to the original. Lab and community announcements live on Community News.
Nov 20, 202615
- Cloud providers break down AI token pricing and GPU utilization economics
Data Center Dynamics· 2d ago
- Parametrix shared guidelines for GPUaaS service level agreements
Data Center Dynamics· 2d ago
They argue GPU cloud providers need objective telemetry and clear availability metrics to back debt financing against compute fleets.
Read article - Lambda deployed Nvidia co-packaged optics switches for 10,000 GPUs
Data Center Dynamics· 2d ago
Lambda is using Nvidia's 144-port 800Gbps CPO switches to connect GB300 NVL72 racks, claiming a latency drop of up to 10 percent at pod scale.
Read article - AMD plans to lease a new 5C data center in Pennsylvania
Data Center Dynamics· 2d ago
Filings show the planned site will start with a 160MW fuel-cell-powered building using closed-loop liquid cooling.
Read article - Microsoft taps Intel EMIB packaging for custom Azure silicon
Data Center Dynamics· 2d ago
Microsoft signed a multi-year deal to package its in-house silicon with Intel's EMIB tech and secure long-term Xeon supply.
Read article - SK hynix demonstrated CXL memory pooling racks at SC26
HPCwire· 2d ago
They demonstrated sharing AI key-value caches across SSDs and DRAM in a rack using CXL memory pooling to reduce GPU idle time.
Read article - Microsoft and Nvidia expanded Azure preview of Vera Rubin NVL72 systems
HPCwire· 2d ago
Azure opened previews of Nvidia Vera Rubin NVL72 instances and added Nvidia Nemotron model customization tools to Microsoft Foundry.
Read article - Attackers stole 430 private keys from Fugaku supercomputer accounts
HPCwire· 2d ago
RIKEN says intruders stole 430 SSH private keys from home directories and accessed account records for 4,222 Fugaku users.
Read article - Danaher and Microsoft partnered to build an AI platform for scientific instruments
HPCwire· 2d ago
The Azure-based platform uses AI agents to automate lab workflows, which they claim cut experiment times by up to 83% in internal tests.
Read article - Chinese team won Gordon Bell Prize for exascale FeMoco simulation
HPCwire· 2d ago
A 10-member Chinese team won the ACM Gordon Bell Prize at SC26 for developing an exascale classical solution to the FeMoco quantum chemistry problem.
Read article - SC26 panel debated whether AI and traditional HPC can share hardware
HPCwire· 2d ago
Panelists debated whether low-precision AI chips, networks, and schedulers can effectively support double-precision scientific simulation workloads.
Read article - NCSA won four HPCwire awards at SC26
HPCwire· 2d ago
Awards recognized whole-cell simulations on Delta, closed-loop AI nuclear reactor control on DeltaAI, and HPC accessibility efforts.
Read article - AI workloads drove SC26 attendance to a record 23,700
HPCwire· 2d ago
SC26 drew over 23,700 attendees as AI scaling demands and rising component prices put pressure on academic supercomputing centers.
Read article - TACC won two HPCwire Readers' and Editors' Choice awards at SC26
HPCwire· 2d ago
Dan Stanzione received an award for HPC leadership and TACC's ACSC student program won for workforce diversity at SC26.
Read article - E4 and Openchip partner on European RISC-V compute systems
HPCwire· 2d ago
The companies plan to integrate Openchip's RISC-V BERNIA accelerators into E4 systems for large-scale AI inference and HPC workflows.
Read article
Nov 19, 202620
- The November 2026 Top500 added only 28 new systems
ServeTheHome· 3d ago
InfiniBand took 19 of the 28 new system spots while total list performance grew by roughly 5%.
Read article - Parametrix analyzed SLA outage risks for GPU neoclouds
Data Center Dynamics· 3d ago
They report customer downtime penalties can exceed data center compensation by over 4x during neocloud outages.
Read article - OpenAI says its models improved chip designs for AMD and its own silicon
Data Center Dynamics· 3d ago
OpenAI claims its models cut standard-cell area by up to 13 percent on its custom chip and delivered up to 14 percent PPA gains on AMD GPU designs.
Read article - STORM team won the Gordon Bell Climate Prize for AI data assimilation on Frontier
HPCwire· 3d ago
The team scaled their AI data assimilation workflow to 74,400 GPUs on Frontier, achieving up to 6 exaflops of sustained BF16 throughput.
Read article - Chinese researchers won the 2026 Gordon Bell Prize
HPCwire· 3d ago
Their FeMoco quantum chemistry code reached 1.456 sustained double-precision exaflops on 20,480 nodes of the LineShine supercomputer.
Read article - Nscale brought its first Nvidia Vera Rubin NVL72 racks online
HPCwire· 3d ago
Nscale says the liquid-cooled Dell racks pack 72 Rubin GPUs and 36 Vera CPUs interconnected via NVLink and Ethernet.
Read article - Dnotitia demoed its vector search ASIC at SC26
HPCwire· 3d ago
They claim offloading vector search to the chip cuts host CPU utilization by 90% while targeting 10x the throughput of a dual-socket server.
Read article - Luxquanta and partners demo 8,000-km quantum-safe network at SC26
HPCwire· 3d ago
The team says the setup linked Chicago and Europe using QKD on short spans and post-quantum crypto across transatlantic fiber at up to 1.6 Tb/s.
Read article - HPCwire honored its 2026 class of HPC Legends at SC26
HPCwire· 3d ago
The annual luncheon recognized 35 leaders in the field during SC26 in Chicago, where total conference registrations topped 23,700.
Read article - HPE built 10 of the top 15 supercomputers on the November TOP500
HPCwire· 3d ago
HPE says it built 10 of the top 15 machines on the latest TOP500 list, totaling 128 systems and over 8.65 exaflops of combined compute.
Read article - SC26 drew 55 quantum vendors as attendance topped 24,000
HPCwire· 3d ago
Hyperion says the quantum market will grow to $3 billion by 2028, though concerns about another quantum winter are rising.
Read article - Nvidia won 15 HPCwire awards at SC26
HPCwire· 3d ago
The company picked up awards for its Vera Rubin NVL72 platform, Quantum-X800 InfiniBand, and Ising quantum-control tools at SC26 in Chicago.
Read article - Penguin Solutions is building an optical shared-memory pool for HPC and AI
HPCwire· 3d ago
They say the box will pool up to 16 TB of DDR5 memory over Marvell optical links with 57.6 Tbps of bandwidth, targeting samples in late 2027.
Read article - BMO tested ML feature selection on IBM quantum systems
HPCwire· 3d ago
They claim their quantum game approach selects ML features with comparable accuracy while needing 10 to 100 times fewer data samples.
Read article - Bull systems swept the top three spots on the Green500 list
HPCwire· 3d ago
Bull claimed the top three spots on the Green500 list for the third consecutive cycle, led by CALMIP's Kairos system at number one.
Read article - MinIO launched general availability of AIStor Memory at SC26
HPCwire· 3d ago
MinIO announced general availability for AIStor Memory, which they say unifies long-term agent memory, workspaces, and secrets on enterprise infrastructure.
Read article - Penguin Solutions showed a photonic memory appliance for CPUs and GPUs
HPCwire· 3d ago
Penguin says the photonic appliance lets CPUs and GPUs share external memory pools at HBM speeds across multi-rack clusters.
Read article - Arm-based supercomputers kept top spots on TOP500 and Green500
HPCwire· 3d ago
Arm says its architecture powers four of the top 10 fastest supercomputers and the four most energy-efficient machines on the Green500.
Read article - TACC launched Sagebrush for persistent AI inference
HPCwire· 3d ago
Funded by the NSF, the system runs on SambaNova chips to host persistent AI inference and RAG workflows alongside TACC's supercomputing facility.
Read article - Pasqal added native Slurm scheduling for its neutral-atom QPUs
HPCwire· 3d ago
The integration uses Slurm 26.11's hierarchical resources model to let users schedule hybrid CPU, GPU, and QPU jobs natively.
Read article
Nov 18, 20263
- IQM will deploy a 54-qubit quantum computer for Turkey's TÜBİTAK
Data Center Dynamics· 4d ago
The 54-qubit Radiance R3 system will integrate directly into TÜBİTAK's existing HPC environment starting in 2027.
Read article - QuEra and NERSC combined Aquila and Perlmutter to simulate 180 qubits
HPCwire· 4d ago
Researchers used NERSC's Perlmutter to validate analog quantum experiments run on QuEra's Aquila system across arrays of up to 180 qubits.
Read article - EuroHPC plans 2027 arrival for Alice Recoque exascale system
HPCwire· 4d ago
TGCC completed site preparations for Alice Recoque, with Bull scheduled to deliver Europe's second exascale machine in 2027.
Read article
Nov 17, 20266
- Nvidia announced Rubin HPC GPU architecture for 2028
ServeTheHome· 5d ago
Nvidia announced Rubin HPC, a server GPU architecture aimed at scientific computing workloads slated to launch in 2028.
Read article - IBM plans 1,040-qubit quantum system using two linked chips
The Next Platform· 5d ago
IBM plans to link two 520-qubit Nighthawk r3 processors using chip-to-chip couplers to create a 1,040-qubit system in 2027.
Read article - Quantinuum updated its Nexus software roadmap for upcoming quantum hardware
The Next Platform· 5d ago
The company says it is co-designing software and error correction tools ahead of Apollo, its planned fault-tolerant trapped-ion system in 2029.
Read article - Osaka University built a platform to link its supercomputer to quantum chips
Data Center Dynamics· 5d ago
The SEAFOOD platform connects Osaka's 2.3-petaflop NEC OCTOPUS supercomputer with RIKEN-supplied superconducting quantum processors.
Read article - DDN unbundled its EXAScaler file system for third-party servers
Blocks & Files· 5d ago
DDN announced EXAScaler 7 will run on qualified third-party servers starting in 2027 rather than being locked to its proprietary appliances.
Read article - Hammerspace launched an all-flash storage system for AI and HPC
Blocks & Files· 5d ago
They claim their 1U all-flash enclosure hits 250 GB/s read bandwidth and packs up to 9.83 PB of raw capacity.
Read article
Nov 16, 20268
- University of Münster deployed ZutaCore two-phase liquid cooling
Data Center Dynamics· 6d ago
They claim the direct-to-chip system supports up to 2,800 watts per socket and cuts cooling energy use by up to 82 percent.
Read article - Core Scientific filed for a 285MW Texas data center
Data Center Dynamics· 6d ago
The $1.3B facility will support up to 285MW of IT capacity, with AMD lined up to lease 110MW as the company pivots to AI infrastructure.
Read article - New supercomputer additions hit record low on November 2026 Top500 list
The Next Platform· 6d ago
The November 2026 Top500 list added only 27 new systems, marking the lowest churn since the rankings began in 1993.
Read article - NCSA landed $334M from Schmidt Sciences to deploy the Garnet supercomputer
HPCwire· 6d ago
The grant funds the new Garnet system at UIUC's petascale facility to anchor the Schmidt Science Computing Network.
Read article - TOP500 turnover hits historic low with only 27 new systems
HPCwire· 6d ago
The top 10 supercomputers remain unchanged while only 27 new systems entered the latest TOP500 list, the lowest turnover ever recorded.
Read article - Nvidia unveiled its Rubin HPC GPU
Data Center Dynamics· 6d ago
The new chip features over 50,000 CUDA cores and seventh-generation NVLink to speed up scientific simulation workloads.
Read article - Top500 added only 27 new supercomputers in late 2026
The Next Platform· 6d ago
The November 2026 Top500 list included just 27 new systems, marking the lowest churn in the rankings since 1993.
Read article - Dell introduced direct liquid-cooled storage for AI clusters
Blocks & Files· 6d ago
The 2-OU chassis holds up to 32 NVMe SSDs and plugs into the same liquid cooling racks as accelerator nodes.
Read article
Nov 15, 20266
- NASA announced its technical demo schedule for SC26
HPCwire· Nov 15, 2026
NASA's booth sessions will cover spaceflight computing, OpenMP target offload for NAS benchmarks, and modular supercomputing data centers.
Read article - Argonne detailed its research talks and workshops for SC26
HPCwire· Nov 15, 2026
Argonne researchers will present work on AI accelerators, high-performance Python, and low-precision hardware tuning at SC26 in Chicago.
Read article - HPSF detailed its SC26 schedule for tutorials and workshops
HPCwire· Nov 15, 2026
The foundation outlined sessions covering projects like Spack, Kokkos, Chapel, and E4S across the SC26 conference program.
Read article - The Lustre Collective partnered with Do IT Now for software-defined storage
HPCwire· Nov 15, 2026
The companies will deploy software-defined Lustre on standard x86 and Arm servers with node-based pricing instead of capacity-based fees.
Read article - ATTO previewed an Ethernet-connected DRAM storage box
HPCwire· Nov 15, 2026
They claim the NVMe-oF appliance hits 35GB/s aggregate bandwidth and under 600 nanoseconds access latency.
Read article - LineShine supercomputer powered Gordon Bell finalist DFT run
HPCwire· Nov 15, 2026
Researchers say they hit 1.31 exaflops FP64 on 20,480 LineShine nodes while scaling their linear-scaling DFT code to 200 million atoms.
Read article
Nov 14, 20262
- Cerebras brings up WSE-5 chip for its CS-5 system
The Next Platform· Nov 14, 2026
Cerebras says its WSE-5 chip is already running models in system tests and will more than double CS-4 speeds.
Read article - SC26 will showcase 59 quantum exhibitors alongside new cooling tech
HPCwire· Nov 14, 2026
SC26 will host 59 quantum vendors—up from seven in 2022—alongside new rack-mounted liquid cooling pumps and 800 VDC power gear.
Read article
Nov 13, 202613
- Syntys started building a 20MW data center in Qatar
Data Center Dynamics· Nov 13, 2026
Syntys says the carrier-neutral facility will support high-density configurations for GPU-accelerated AI and HPC workloads.
Read article - Microsoft signed a deal to deploy Cerebras AI chips
Data Center Dynamics· Nov 13, 2026
Microsoft signed on as Cerebras' third hyperscaler customer alongside AWS and Oracle, with full deployment details promised in 2027.
Read article - Penguin launched a CXL card that lets AI servers reuse DDR4 memory
HPCwire· Nov 13, 2026
Penguin says the eight-DIMM card uses a Marvell CXL controller to let AI servers repurpose older DDR4 memory.
Read article - Marvell showed optical and CXL interconnects for AI clusters
HPCwire· Nov 13, 2026
They demonstrated optical fabrics, CXL memory expansion, and 1.6T active electrical cables to connect multi-rack AI systems.
Read article - HPE launched hybrid Cray storage and a rack cooling controller
HPCwire· Nov 13, 2026
HPE says the hybrid E2000 arrays cost 82% less than all-flash while pushing up to 116 GB/s of read throughput per rack.
Read article - LANL moved seven HPC clusters to OpenCHAMI management
HPCwire· Nov 13, 2026
Los Alamos moved seven production clusters with over 11,000 compute nodes to the open-source OpenCHAMI systems-management framework.
Read article - Argonne's Rick Stevens pitches DOE Genesis Mission at SC26
HPCwire· Nov 13, 2026
Stevens argues AI agents can boost research productivity tenfold and wants the supercomputing community to accelerate adoption.
Read article - Women in HPC selected 12 Rising Stars for SC26 travel support
HPCwire· Nov 13, 2026
The group is providing travel funding for 12 early-career researchers to present lightning talks at SC26, with full sponsorship for four of them.
Read article - AI data center demand sparks a US-China race for nuclear power
HPCwire· Nov 13, 2026
A Brookings paper estimates that 188 GW of US data center capacity by 2032 could match the entire US residential sector's annual power use.
Read article - BSC outlined its technical lineup and research sessions for SC
HPCwire· Nov 13, 2026
Researchers from Barcelona Supercomputing Center are presenting work on SmartNICs, RISC-V, memory systems, and quantum computing at the conference.
Read article - Staera raised $4.6M to build optical links directly into AI chips
HPCwire· Nov 13, 2026
They claim their on-chip optical links can move 24 TB/s for under 50W, using less than a tenth of the power of current co-packaged optics.
Read article - QuiX Quantum secured Dutch funding to build a 100-qubit photonic computer
HPCwire· Nov 13, 2026
The consortium will develop a 100+ qubit photonic quantum platform by 2028, targeting control ASICs with latencies below 10 ns.
Read article - HPE added hybrid flash and disk options to Cray E2000 storage
Blocks & Files· Nov 13, 2026
HPE claims the new 32TB HDD and SSD configurations deliver up to 116 GB/s read throughput per rack at 82% lower cost than all-flash.
Read article
Nov 12, 202621
- Jensen Huang leaves Nvidia amid misconduct probes and GPU smuggling claims
SemiAnalysis· Nov 12, 2026
SemiAnalysis reports Jensen Huang resigned as CEO amid ethics probes and claims over 1.5GW of restricted Nvidia GPUs reached China.
Read article - CoreWeave hiked GPU rental prices as shortages keep older fleets profitable
The Next Platform· Nov 12, 2026
CoreWeave says H100 rental rates jumped 50 percent in Q3 as tight supplies let them raise prices on fully amortized older chips.
Read article - Jefferson Lab decommissioned its Xeon Phi supercomputer
Data Center Dynamics· Nov 12, 2026
The DOE lab retired its 425.8 teraflops SciPhi-XVI cluster after nearly a decade and moved lattice QCD workloads to newer systems.
Read article - Nvidia board searches for potential CEO replacements
Data Center Dynamics· Nov 12, 2026
The board has hired an outside search firm and is considering an interim CEO for Jensen Huang following personal misconduct allegations.
Read article - Jensen Huang resigns as Nvidia CEO
HPCwire· Nov 12, 2026
Jensen Huang has stepped down as CEO and board member effective immediately, with CFO Colette Kress taking over as interim chief.
Read article - Ace Computers will bundle ThisWay Global's GPU orchestration software
HPCwire· Nov 12, 2026
Ace will resell ThisWay Global's Amalgamy orchestration tool with its custom GPU systems to help manage mixed accelerator fleets.
Read article - AMD partnered with PsiQuantum to co-design hybrid quantum systems
HPCwire· Nov 12, 2026
AMD plans an equity investment in PsiQuantum as both companies tie Instinct GPUs and ROCm into fault-tolerant photonic quantum systems.
Read article - Cornelis announced 800G CN6000 SuperNICs and Omni-Path switches
HPCwire· Nov 12, 2026
They claim the 800G Omni-Path SuperNIC pushes 1.6 billion messages per second with 21% lower MPI latency than Nvidia's 800G InfiniBand.
Read article - Jensen Huang resigned as Nvidia CEO
Data Center Dynamics· Nov 12, 2026
CFO Colette Kress stepped in as interim CEO after Huang resigned from the company following 33 years at the helm.
Read article - Jensen Huang resigned as Nvidia CEO amid workplace scandal
Data Center Dynamics· Nov 12, 2026
Nvidia named CFO Colette Kress interim CEO after founder Jensen Huang stepped down following reports of multiple affairs with staff.
Read article - SpaceX is building custom Nvidia Rubin NVL72 racks
Data Center Dynamics· Nov 12, 2026
SemiAnalysis says the custom rack swaps NVLink cables for an 84-layer PCB midplane for both ground and orbital use.
Read article - AMD expanded its Instinct MI400 GPU lineup for HPC and AI
HPCwire· Nov 12, 2026
The MI430X features 432 GB of HBM4 memory with native FP64, and AMD claims their MI455X beats Nvidia's Vera Rubin in MLPerf inference by up to 30%.
Read article - Axelera AI unveiled its Titania AI inference architecture
HPCwire· Nov 12, 2026
They claim the LPDDR6-based chiplet design delivers up to 7x higher throughput per watt than existing data center servers without needing HBM.
Read article - RAVEL and Sandia collaborate on distributed AI orchestration
HPCwire· Nov 12, 2026
Project Quicksilver tests power-aware scheduling to distribute and manage AI workloads across separate data centers based on energy and capacity limits.
Read article - Researchers simulated an ITER fusion blanket on Aurora and Frontier
HPCwire· Nov 12, 2026
The Gordon Bell finalist team modeled multiphysics in ITER's liquid-metal breeding blanket across the two DOE exascale systems.
Read article - SMU won $4.2M from NIST for HPC-driven manufacturing digital twins
HPCwire· Nov 12, 2026
The university will use its Nvidia DGX SuperPOD cluster to simulate semiconductor manufacturing workflows and target 10% to 20% operational efficiency gains.
Read article - Arena Physica will train physics models on ORNL's Lux supercomputer
HPCwire· Nov 12, 2026
They will train their Heaviside EM model on ORNL's AMD-based Lux system, claiming it runs orders of magnitude faster than traditional solvers.
Read article - Q-CTRL raised $290M to grow its quantum control software
HPCwire· Nov 12, 2026
The company closed a $290 million Series C round led by Salesforce Ventures to expand its quantum infrastructure software for data centers and sensing.
Read article - Qunova ran an 80-qubit hybrid chemistry simulation on Fugaku
HPCwire· Nov 12, 2026
They claim their HI-VQE algorithm reached chemical accuracy on a 4Fe-4S cluster using 2,518 Fugaku node-hours and IBM quantum hardware.
Read article - Jensen Huang stepped down as Nvidia CEO
The Next Platform· Nov 12, 2026
CFO Colette Kress was named interim CEO while the board searches for a successor amid ongoing scrutiny over GPU export controls.
Read article - Jensen Huang resigned as Nvidia CEO
SemiAnalysis· Nov 12, 2026
Nvidia appointed CFO Colette Kress interim CEO after Jensen Huang resigned amid discussions on export controls and accelerator supply chains.
Read article
Nov 11, 202614
- Broadcom shows off Ethernet switches and 800G NICs for AI clusters
HPCwire· Nov 11, 2026
They plan to showcase Tomahawk and Jericho switches alongside a workshop on scale-up Ethernet fabrics for AI and HPC.
Read article - University of Alabama builds new HPC center with Dell and Nvidia
HPCwire· Nov 11, 2026
The cluster pairs Dell PowerEdge servers with Nvidia H200 NVL and RTX PRO 6000 GPUs over Quantum-2 InfiniBand for statewide research.
Read article - Nvidia board weighs ousting Jensen Huang over alleged affairs
Data Center Dynamics· Nov 11, 2026
Nvidia's board is reportedly preparing to oust CEO Jensen Huang after reports surfaced alleging multiple improper relationships.
Read article - Nvidia hires law firm to investigate CEO's affair with suspected spy
Data Center Dynamics· Nov 11, 2026
Nvidia hired Wachtell Lipton to investigate whether trade secrets or national security were compromised following Jensen Huang's affair.
Read article - Quantinuum upgraded Riken's Reimei quantum system to 56 qubits
Data Center Dynamics· Nov 11, 2026
Quantinuum upgraded the on-premise system at Riken from a 20-qubit H1 to its 56-qubit Model H2 and plans to link it to SoftBank AI data centers.
Read article - IBM announced Nighthawk r3 and plans a 1,040-qubit system for 2027
HPCwire· Nov 11, 2026
IBM says the 520-qubit processor will run 10,000 quantum operations, with plans to link two together into a single system in 2027.
Read article - Chatsworth Products introduced OCP Open Rack V3 systems for AI and HPC
HPCwire· Nov 11, 2026
The 44OU and 48OU racks support 48V DC power shelves up to 33 kW and busbars up to 2500 A for high-density clusters.
Read article - Growing data center demands clash with local utility grids and communities
HPCwire· Nov 11, 2026
Communities like Washington's Grant County are limiting data center power draw as large builds clash with local energy and water supplies.
Read article - Supermicro started shipping H15 servers with AMD EPYC 9006 CPUs
HPCwire· Nov 11, 2026
Supermicro claims the new platforms provide 33% more cores and 2.6x more memory bandwidth over prior AMD generations.
Read article - Boeing ran aerodynamic turbulence simulations on Frontier
HPCwire· Nov 11, 2026
Boeing used the Frontier supercomputer to simulate microturbulent airflow and ice buildup on aircraft wings to help reduce cruise drag.
Read article - Classiq will demo hybrid quantum-HPC software at SC26
HPCwire· Nov 11, 2026
The company says its platform lets developers build hybrid quantum-HPC workloads once and deploy them across different quantum hardware.
Read article - Hitachi Vantara updated its VSP One object storage for AI workloads
HPCwire· Nov 11, 2026
They claim the updated system delivers up to 5.5x faster small-object access and queries open data formats directly on storage.
Read article - MGHPCC will present 75 research projects at SC26
HPCwire· Nov 11, 2026
The center plans to feature work spanning AI, climate science, and its shared AI Compute Resource at an in-person booth and virtual exhibit.
Read article - Dell built a 157 PB all-flash storage rack with Kioxia SSDs
Blocks & Files· Nov 11, 2026
They say the 32U setup packs 640 245.76 TB flash drives across 16 PowerEdge servers to target AI data lakes.
Read article
Nov 10, 202614
- UK Met Office finished its move to an Azure cloud supercomputer
Data Center Dynamics· Nov 10, 2026
The Microsoft-managed cloud system gives the agency 1.8 million cores and roughly 60 petaflops of compute for weather modeling.
Read article - AMD launched its Venice Epyc CPUs with up to 256 cores
The Next Platform· Nov 10, 2026
AMD claims the Zen 6 cores deliver a 20 percent IPC uplift over Zen 5, with the 2nm lineup scaling up to 256 cores and 16 DDR5 channels.
Read article - AMD detailed its Zen 6 Venice Epyc processors for agentic AI workloads
The Next Platform· Nov 10, 2026
AMD claims agentic AI workloads will push datacenter GPU-to-CPU ratios from up to 8:1 down to roughly 1:1.
Read article - Gartner projects datacenter hardware spending will hit $1.73 trillion by 2027
The Next Platform· Nov 10, 2026
Gartner forecasts datacenter systems and IaaS spending will reach $1.73 trillion in 2027, surpassing software on massive AI hardware investments.
Read article - Quantinuum upgraded RIKEN's quantum system and expanded hybrid HPC work
HPCwire· Nov 10, 2026
Quantinuum upgraded its Reimei quantum system at RIKEN to Model H2 to connect with the Fugaku supercomputer and GPU clusters for hybrid simulations.
Read article - INFINITIX added Slurm-on-Kubernetes scheduling to AI-Stack
HPCwire· Nov 10, 2026
The software pools GPUs across HPC and AI jobs, adding automatic training failover and disaggregated LLM inference.
Read article - AMD shared architecture details for its Zen 6 Venice EPYC chips
HPCwire· Nov 10, 2026
AMD claims its 2nm Zen 6 EPYC chips deliver 27% higher IPC than Zen 5 and provide up to 1.6 TB/s of memory bandwidth per socket.
Read article - chipStar compiles CUDA and HIP code to SPIR-V for OpenCL backends
HPCwire· Nov 10, 2026
The toolchain compiles CUDA and HIP kernels to SPIR-V, allowing apps like OpenMM to run on Intel GPUs on the Aurora supercomputer without full rewrites.
Read article - Jefferson Lab decommissioned its Xeon Phi supercomputers
HPCwire· Nov 10, 2026
The lab retired its Knights Landing clusters after nearly a decade of running lattice QCD simulations.
Read article - Sparxity raised new funding to build chips for sparse AI models
HPCwire· Nov 10, 2026
They claim their dynamic sparsity chips target up to 44x faster AI inference and 10x faster sparse solvers.
Read article - ISC opened nominations for the 2027 Jack Dongarra Award
HPCwire· Nov 10, 2026
The early-career HPC award carries a 5,000-euro prize and a lecture at ISC 2027, with nominations due by February 12, 2027.
Read article - Fast remote file systems prevent idle GPUs in AI clusters
HPCwire· Nov 10, 2026
The article argues AI training bottlenecks shift to remote storage and NFS architectures as clusters scale to millions of files.
Read article - Xanadu and HyperLight reduce optical switch loss to 7 mdB
HPCwire· Nov 10, 2026
They claim a thin-film lithium niobate switch reached 7 mdB insertion loss, reducing photon loss nearly threefold compared to their previous hardware.
Read article - Quantum X Labs cut quantum decoding errors in tests on IBM hardware
HPCwire· Nov 10, 2026
They claim their AI-driven noise calibration reduced logical decoding failures by up to 20% on experimental data from IBM quantum processors.
Read article
Nov 9, 20268
- IBM deployed a Quantum System Two in Chicago
Data Center Dynamics· Nov 9, 2026
The 156-qubit system is IBM's first in Illinois, and the firm also paired a quantum chip with an AMD-based ORNL cluster for error correction.
Read article - Digital Realty opened a liquid cooling lab in São Paulo
Data Center Dynamics· Nov 9, 2026
The site supports liquid-to-liquid and air-to-liquid cooling tests for high-density AI racks drawing 150kW and higher.
Read article - Google modified its TPU v8 accelerators for AI inference workloads
The Next Platform· Nov 9, 2026
The TPU 8tx swaps the 3D torus for a lower-latency boardfly fabric and replaces Google's Axion Arm CPUs with AMD Epyc processors.
Read article - IBM deployed its Quantum System Two in Chicago
HPCwire· Nov 9, 2026
The system runs on IBM's Nighthawk chip at Hyde Park Labs before moving to the Illinois Quantum and Microelectronics Park.
Read article - Fermaca deployed DDN storage for its AI cloud in Mexico
Blocks & Files· Nov 9, 2026
The provider says it will pair DDN storage with Nvidia B300 NVL72 systems at its 60 MW Mexico City facility.
Read article - Diraq deployed an 8-qubit quantum computer in an Equinix data center
HPCwire· Nov 9, 2026
The silicon spin system draws under 20 kW at 170 square feet, and Diraq targets expanding it to 256 qubits within a year.
Read article - UKRI committed over £51 million to new quantum projects
HPCwire· Nov 9, 2026
UKRI announced over £51 million in quantum funding, including £35 million from Innovate UK to support quantum networking testbeds.
Read article - DAOS hit 30 TBps bandwidth across 8,192 nodes on Aurora
Blocks & Files· Nov 9, 2026
Argonne says its 800-node DAOS setup delivered up to 30 TB/s of I/O bandwidth across 8,192 Aurora compute nodes.
Read article
Nov 8, 20261
- Nvidia CEO admits affair with executive at firm under GPU smuggling probe
Data Center Dynamics· Nov 8, 2026
Nvidia's board is investigating after CEO Jensen Huang admitted to an affair with an executive at a cloud firm accused of smuggling GPUs to China.
Read article
Nov 7, 20261
- ASRock Rack displayed a 16-slot motherboard for Intel Diamond Rapids
ServeTheHome· Nov 7, 2026
The DMRD16-2L+ server board features 16 DDR5 memory slots designed for Intel Diamond Rapids processors.
Read article
Nov 6, 202611
- China stalls import approvals for 600,000 Nvidia H200 GPUs
Data Center Dynamics· Nov 6, 2026
Roughly 600,000 Nvidia H200 AI chips are stranded in Hong Kong while awaiting Chinese government import approval, according to reports.
Read article - IBM, AMD, and Oak Ridge co-locate quantum and HPC systems for error correction
The Next Platform· Nov 6, 2026
They put IBM's Nighthawk QPU alongside Oak Ridge's AMD-powered Umbra cluster in one facility to speed up hybrid quantum error correction research.
Read article - EuroHPC ordered bidders to resubmit AI Gigafactory tenders
HPCwire· Nov 6, 2026
A technical glitch forced EuroHPC to scrap existing AI Gigafactory bids and push the medium-scale deadline to February 2027.
Read article - AWS boosted metadata IOPS on FSx for Lustre Intelligent-Tiering
HPCwire· Nov 6, 2026
AWS says FSx for Lustre Intelligent-Tiering now hits up to 192,000 metadata IOPS, a 16x increase over the previous 12,000 limit.
Read article - UChicago secured $10M to develop an open-source AI weather modeling platform
HPCwire· Nov 6, 2026
The team is building an open-source AI Almanac platform designed to blend physics and AI weather models for regions with modest compute resources.
Read article - Cerabyte verified S3 object storage on ceramic media with CERN
HPCwire· Nov 6, 2026
CERN openlab tested writing and reading sample physics data on Cerabyte's ceramic media using standard S3 tools.
Read article - Women in HPC selected 12 researchers for its SC26 Rising Stars cohort
HPCwire· Nov 6, 2026
The 12 early-career researchers will receive mentorship and travel funding to present lightning talks at the SC26 conference.
Read article - Argonne engineer Kevin Hayden will serve as SC26 general chair
HPCwire· Nov 6, 2026
Argonne network engineer Kevin Hayden was named general chair for SC26, which organizers expect to draw over 20,000 attendees to Chicago.
Read article - Smartbird partners with Supermicro to offer managed AI clusters
HPCwire· Nov 6, 2026
Smartbird and Supermicro are teaming up to deploy and operate dedicated, rack-scale AI clusters for enterprise training and inference.
Read article - Analysis highlights edge chips and supply-chain security in AI race
HPCwire· Nov 6, 2026
An analysis argues low-power edge chips, hardware supply-chain security, and data center backlash will shape AI competition beyond frontier training clusters.
Read article - NQCC and STFC funded 13 UK quantum computing projects
HPCwire· Nov 6, 2026
The grants support three- and six-month industry projects exploring quantum applications across logistics, telecoms, and finance.
Read article
Nov 5, 20264
- Chemours launched two-phase cooling fluids for data centers
Data Center Dynamics· Nov 5, 2026
Chemours claims its Faizeon 200 fluid keeps GPUs up to 19°C cooler than single-phase water cooling while using lower flow rates.
Read article - Quint Computing raised €9.1M to build fault-tolerant quantum chips
Data Center Dynamics· Nov 5, 2026
They claim their superconducting chip design suppresses errors at the circuit level to cut hardware overhead and power use by 100x.
Read article - Arista plans Ethernet scale-up trials for AI clusters in 2027
The Next Platform· Nov 5, 2026
Arista expects ESUN scale-up network trials for up to 1,024 accelerators to start in 2027, with production rollouts following in 2028.
Read article - ORNL, IBM, and AMD paired quantum chips with HPC for error correction
HPCwire· Nov 5, 2026
The project links IBM's quantum chips with ORNL's AMD MI355X-based Umbra cluster to speed up error-correction decoding research.
Read article
Nov 4, 202610
- Nvidia told cloud partners to limit compute sales to China
Data Center Dynamics· Nov 4, 2026
Reports say Nvidia told cloud partners to sell at least 51 percent of their capacity to non-Chinese clients on future deals.
Read article - NCSA hosted its industry conference on supercomputing and AI
HPCwire· Nov 4, 2026
Speakers focused on full-stack AI infrastructure tradeoffs beyond peak accelerator FLOPS, including memory, networking, and power.
Read article - Argonne added Nvidia GH200 and B200 clusters to expand its AI inference service
HPCwire· Nov 4, 2026
The expansion adds 2,688 Nvidia GH200s on HPE Cray hardware and 64 B200s, which ALCF says gives the service 100 times more GPU memory.
Read article - Universal Quantum launched its iQPU trapped-ion processor
HPCwire· Nov 4, 2026
The trapped-ion chip features an active area over 400mm² and is fabricated on a commercial 200mm foundry line to support modular multi-chip scaling.
Read article - QuantumCT bought 150 hours of cloud access on IonQ quantum systems
HPCwire· Nov 4, 2026
Under the agreement, Connecticut researchers and companies get over 150 hours of cloud runtime on IonQ's trapped-ion hardware.
Read article - Synopsys brings EDA tools to the European Chips Design Platform
HPCwire· Nov 4, 2026
Synopsys partnered with the EU Chips JU to offer its semiconductor design software to European startups and research teams through EuroCDP.
Read article - Supermicro will build AI inference servers combining GPUs and SambaNova RDUs
HPCwire· Nov 4, 2026
Supermicro plans to deploy 10,000 SambaNova SN50 chips in systems that use Nvidia GPUs for prefill and RDUs for decode.
Read article - D-Wave updated its Stride hybrid solver
HPCwire· Nov 4, 2026
D-Wave claims its updated hybrid solver finds better solutions in 15 seconds than the previous version did in 120 seconds on benchmark tests.
Read article - AMD hits HBM supply limits in datacenter GPU race with Nvidia
The Next Platform· Nov 4, 2026
Analysis argues HBM supply constraints and heavy memory configurations keep AMD from closing the datacenter GPU gap with Nvidia.
Read article - xAI plans 1.2 million Blackwell GPUs and multi-exabyte storage for Colossus 2
Blocks & Files· Nov 4, 2026
Analysts estimate xAI's planned 1.2 million Blackwell GPU cluster will need over 5.2 EB of storage across its five sites.
Read article
Nov 3, 202613
- Setara launched a daily index tracking on-demand GPU prices
HPCwire· Nov 3, 2026
The index tracks list prices across seven GPU models from independent cloud providers in dollars per GPU-hour.
Read article - Foxconn aims to manufacture quantum computers for hardware vendors
Data Center Dynamics· Nov 3, 2026
Foxconn plans to build quantum systems for third-party designers, though its chairman says commercial viability is at least two years away.
Read article - Canada joined EuroHPC as its 39th member
Data Center Dynamics· Nov 3, 2026
The partnership gives Canadian researchers access to European supercomputing, AI, and quantum compute resources.
Read article - Marvell acquires Lumentum in $80 billion optics deal
The Next Platform· Nov 3, 2026
Marvell is buying Lumentum for $80 billion in stock to secure laser and silicon photonics supply for AI datacenter interconnects.
Read article - Accelsius brings two-phase liquid cooling rack to Villanova for NSF study
HPCwire· Nov 3, 2026
The company provided its two-phase liquid cooling rack to help an NSF consortium aiming to cut data center energy use by 20 to 35 percent.
Read article - Jülich Supercomputing Centre hosts the open EVE Earth science model
HPCwire· Nov 3, 2026
JSC is now hosting ESA's open EVE-Instruct LLM for Earth science on its Blablador infrastructure via the JUPITER AI Factory.
Read article - Hyperion reported on-prem HPC and AI server spending rose 20% in early 2026
HPCwire· Nov 3, 2026
Hyperion says spending on HPC and AI for science reached $16.5 billion in 1H 2026, with Dell overtaking HPE for the top market-share spot.
Read article - ARIA deploys Lumai's optical AI inference servers in UK test lab
HPCwire· Nov 3, 2026
The UK agency is testing Lumai's 3D optical compute hardware, which they claim can cut AI inference data center power by 90%.
Read article - Dell passed HPE in the HPC server market as revenues jumped 20%
HPCwire· Nov 3, 2026
Hyperion says 1H26 HPC server sales hit $16.5B while unit growth was just 4.4%, driven by rising GPU and memory costs.
Read article - Penguin Solutions outlines AI and memory demos for SC26
HPCwire· Nov 3, 2026
Penguin Solutions will demo its cluster management software, CXL memory hardware, and AI compute systems at SC26 in Chicago.
Read article - Quint Computing raised €9.1M for hardware-level quantum error reduction
HPCwire· Nov 3, 2026
The German startup claims its superconducting chip design can cut quantum hardware overhead and energy use by 100x.
Read article - Bull and Schwarz Digits partner on European AI infrastructure
HPCwire· Nov 3, 2026
The companies plan to offer BullSequana hardware on STACKIT cloud and co-design liquid-cooled OCP servers for European AI and HPC workloads.
Read article - Sandisk and Kioxia split on High-Bandwidth Flash for GPUs
Blocks & Files· Nov 3, 2026
Sandisk is backing the SK Hynix High-Bandwidth Flash standard for GPUs, while joint-venture fab partner Kioxia is pursuing CXL and alternative flash designs.
Read article
Nov 2, 202617
- Qualcomm validated stacked cache memory for Dragonfly AI250
ServeTheHome· Nov 2, 2026
Qualcomm validated High Bandwidth Cache stacked memory ahead of sampling its Dragonfly AI250 accelerator in 2027.
Read article - US investigates Megaspeed executive for alleged spying on Nvidia CEO
Data Center Dynamics· Nov 2, 2026
Officials in the US, Singapore, and Taiwan are probing Megaspeed's Alice Huang over alleged intelligence gathering targeting Nvidia CEO Jensen Huang.
Read article - Industry shifts AI infrastructure focus from raw capacity to cost per token
Data Center Dynamics· Nov 2, 2026
GPT-4-class inference costs fell 98% since 2023 to under $0.50 per million tokens while average enterprise AI budgets grew nearly sixfold.
Read article - Nvidia considers $5B credit guarantee for Nscale AI data center
Data Center Dynamics· Nov 2, 2026
Reports say Nvidia may backstop debt for Nscale's 460MW facility in West Virginia, which will run Vera Rubin hardware for Anthropic.
Read article - Google ran its first AI workload on an orbital TPU satellite
Data Center Dynamics· Nov 2, 2026
Google tested inference on a 1kW, four-TPU prototype in orbit, though the company notes space compute currently costs up to 100x more than on Earth.
Read article - Envision deployed an 800VDC power system for an AI data center in China
Data Center Dynamics· Nov 2, 2026
Envision claims its Ulanqab data center uses solid-state transformers to supply 800VDC directly to AI hardware racks.
Read article - SETI deployed Nvidia GPUs and AI agents to process live telescope data
HPCwire· Nov 2, 2026
The system uses Holoscan and Blackwell GPUs to stream 2.4 terabits per second into memory for real-time radio astronomy signal processing.
Read article - Penguin Solutions launched DDR5-12800 MRDIMMs for AI servers
HPCwire· Nov 2, 2026
They claim the new modules multiplex two physical ranks into one logical rank to boost bandwidth at DDR5-12800 speeds.
Read article - FAU researchers tested quantum ML methods on D-Wave hardware
HPCwire· Nov 2, 2026
A D-Wave quantum support vector machine hit 91.9% accuracy on hardware when trained on only 108 lung cancer sample records.
Read article - CESGA and IGFAE join EU HADRON-2030 to provide quantum compute access
HPCwire· Nov 2, 2026
The project gives 49 European organizations remote access to CESGA's quantum computer and IGFAE's analogue quantum simulator.
Read article - Marvell agreed to acquire optical specialist Lumentum for $115B
HPCwire· Nov 2, 2026
Marvell plans to combine Lumentum's lasers and optical switches with its networking silicon to build optical interconnects for AI infrastructure.
Read article - China unveiled Sugon 8000, its fourth exascale supercomputer
HPCwire· Nov 2, 2026
The 100,000-card cluster uses Hygon x86-compatible chips, proprietary liquid cooling, and a custom RDMA network.
Read article - NetApp completed its acquisition of PEAK:AIO
HPCwire· Nov 2, 2026
NetApp plans to integrate PEAK:AIO's parallel file system and metadata architecture into ONTAP to scale storage for AI clouds and GPU clusters.
Read article - BSC leads a €3M European and Latin American supercomputing network
HPCwire· Nov 2, 2026
The EU is funding a €3 million project to federate supercomputing and AI infrastructure across Europe and seven Latin American countries.
Read article - Qualcomm pitched near-memory compute chips for datacenter AI inference
The Next Platform· Nov 2, 2026
They place an AI accelerator die directly under an LPDDR stack to cut data-movement power and avoid costly HBM packaging.
Read article - Canada joins EuroHPC as 39th participating state
HPCwire· Nov 2, 2026
The agreement gives Canadian researchers access to European supercomputers, AI systems, and quantum hardware for joint research projects.
Read article - ASRock Rack showed liquid-cooled Nvidia GPU servers at OCP
ServeTheHome· Nov 2, 2026
ASRock Rack demonstrated direct-to-chip liquid-cooled server platforms intended for high-density Nvidia GPU systems.
Read article
Nov 1, 20261
- QuEra outlines a 2028 roadmap targeting 256 logical qubits
Data Center Dynamics· Nov 1, 2026
QuEra says its planned 2028 Libra system will use 10,000 physical qubits to deliver 256 logical qubits and one million reliable operations.
Read article
Oct 30, 20265
- HPCwire tracked October 2026 community career moves and awards
HPCwire· Oct 30, 2026
Michael Garland and Nathan Bell won the SC26 Test of Time Award for their foundational 2009 paper on sparse matrix computation on GPUs.
Read article - HPCwire compared US and China AI hardware and model progress
HPCwire· Oct 30, 2026
The US maintains a large AI chip production lead, but Huawei is deploying a 256,000-card cluster as China pushes open-weight models.
Read article - BSC and AstraZeneca partner on biomedical AI and supercomputing
HPCwire· Oct 30, 2026
AstraZeneca researchers will use BSC's MareNostrum 5 system for biomedical simulations, clinical trial modeling, and AI drug discovery.
Read article - South Korean startups push domestic AI chips for sovereign compute
The Next Platform· Oct 30, 2026
Furiosa raised $620 million to scale RNGD chip production to 20,000 units this year while developing a 2 nm part with Broadcom.
Read article - Sparxity installed a sparse computing cluster at Jülich
Data Center Dynamics· Oct 30, 2026
They claim the Jubis system runs sparse solvers 10x faster and cuts energy use by 18x for event-based ML workloads.
Read article
Oct 29, 202612
- CIQ launched an open source Kubernetes distribution for AI and HPC
HPCwire· Oct 29, 2026
They packaged upstream components into a single signed installer with built-in MCP agent tools for managing clusters.
Read article - BSC launched an energy planning tool powered by MareNostrum 5
HPCwire· Oct 29, 2026
The tool runs on MareNostrum 5 to calculate 5 km resolution global renewable energy metrics in real time during climate simulations.
Read article - QuantrolOx raised €12M to automate quantum hardware tuning
HPCwire· Oct 29, 2026
The startup secured new funding to scale its automated QPU calibration software and build out its US presence, bringing total funding to about $31 million.
Read article - IQM demonstrated full-chip entanglement on its 54-qubit processor
HPCwire· Oct 29, 2026
IQM claims it fully entangled all 54 qubits on its Emerald processor with a fidelity of 0.516, keeping it above the classical threshold.
Read article - FuriosaAI raised $620M for its inference accelerators
HPCwire· Oct 29, 2026
The startup plans to use the Series D funding to ramp production of its RNGD AI chips and develop a third-generation part.
Read article - Arm proposed a benchmark method to evaluate CPUs for agentic AI
The Next Platform· Oct 29, 2026
Arm argues datacenter CPU buyers must combine component tests with end-to-end task metrics to measure agentic AI workloads.
Read article - Synopsys joined a Singapore research testbed for tropical data center cooling
HPCwire· Oct 29, 2026
The team will use simulation and testbeds to tackle tropical cooling demands that can consume up to 40% of a data center's energy.
Read article - Alice & Bob lowered physical qubit estimates to break RSA-2048
HPCwire· Oct 29, 2026
They estimate cat-qubit hardware would need 22,593 physical qubits and 3.14 days to factor RSA-2048 using Shor's algorithm.
Read article - Rival.io launched GraphX to run DeepSeek V4 Flash on eight Nvidia L4 GPUs
HPCwire· Oct 29, 2026
They claim their kernel and placement optimizer cuts open-weight model hosting costs by up to 95%.
Read article - Quantum DC canceled a community meeting for a planned 1.2GW Australian data center
Data Center Dynamics· Oct 29, 2026
The developer canceled a public forum for its proposed 1.2GW campus near Adelaide after citing a bomb threat amid local opposition.
Read article - Nvidia pitched efficiency and safety tech for AI data centers at GTC Berlin
Data Center Dynamics· Oct 29, 2026
They claim their DSX platform cuts cooling water use by up to 100% and uses 30% less power for a 2GW data center deployment.
Read article - DCD discussed data center grid regulation with Allison Clements
Data Center Dynamics· Oct 29, 2026
A video interview covers evolving FERC rules, large load management, and safeguards against speculative data center power demand.
Read article
Oct 28, 202616
- LiquidStack makes the case for sustaining engineering in liquid cooling
Data Center Dynamics· Oct 28, 2026
LiquidStack argues cooling vendors must actively manage component obsolescence so hardware survives across multiple server generations.
Read article - Oracle will host Google TPUs in multi-gigawatt OCI rollout
Data Center Dynamics· Oct 28, 2026
Oracle plans a multi-gigawatt deployment of Google's TPUs on OCI, becoming the first hyperscaler to offer the chips outside Google Cloud.
Read article - Blockmate buys eight Nvidia B200 GPUs to enter the AI cloud market
Data Center Dynamics· Oct 28, 2026
They spent $635,000 on the single eight-GPU node and claim the investment could pay for itself in about two years.
Read article - Genius Group and FingerMotion plan a 2.4MW AI data center in Alberta
Data Center Dynamics· Oct 28, 2026
The companies plan to deploy 432 GPU servers across a 2.4MW site in Canada, targeting initial compute availability in mid-2027.
Read article - Intel will acquire quantum software company Classiq for $2.3B
The Next Platform· Oct 28, 2026
Intel plans to buy the Israeli software maker to build out a commercial full-stack quantum computing business.
Read article - SLAC created SPINE to reconstruct neutrino detector data with ML
HPCwire· Oct 28, 2026
The ML framework reconstructs 3D particle tracks from liquid argon detectors for neutrino experiments like DUNE and ICARUS.
Read article - BUZZ HPC landed $129M to deploy Nvidia GB200 systems in Canada
HPCwire· Oct 28, 2026
The three-year debt facility will fund Nvidia GB200 NVL72 clusters to support a sovereign AI contract with Cohere.
Read article - Monterey Data Conference focused on DOE scientific AI and computing
HPCwire· Oct 28, 2026
DOE labs and researchers met to discuss integrated platforms for scientific AI, quantum computing, and the Genesis Mission.
Read article - UbiQD won a $200K grant to build single-photon sources for quantum computers
HPCwire· Oct 28, 2026
The Los Alamos company received state funding to prototype fiber-coupled quantum dot light sources that operate at room temperature.
Read article - NVision and HyperLight partner to put molecular qubits on photonic chips
HPCwire· Oct 28, 2026
They plan to deposit thin-film organic qubits directly onto photonic chips, targeting 1,000 logical qubits on chip by 2030.
Read article - Cowboy Space designed a rocket stage to run as an orbital data center
HPCwire· Oct 28, 2026
They claim each rocket upper stage will host five racks of Nvidia GPUs in orbit and operate as a megawatt-class data center.
Read article - Jülich installed Sparxity's 650k-core neuromorphic system for sparse workloads
HPCwire· Oct 28, 2026
JSC installed Sparxity's 656,640-core JUBIS system alongside Jupiter Booster to handle sparse ML tasks that GPUs struggle with.
Read article - D-Wave and LG CNS partner on quantum annealing optimization
HPCwire· Oct 28, 2026
The companies signed an MOU to use D-Wave's quantum annealing systems for industrial optimization applications in South Korea.
Read article - Intel will buy quantum software maker Classiq for $2.3 billion
The Register· Oct 28, 2026
Intel is buying quantum software startup Classiq for $2.3 billion as it moves its decade of spin-qubit chip research into a commercial business.
Read article - Fujifilm plans to increase LTO tape production fivefold by 2028
Blocks & Files· Oct 28, 2026
They are investing $63 million to quintuple tape manufacturing output to meet rising demand for AI training archives.
Read article - Meta used larger memory and AI agents to match Nvidia B200 speeds on AMD GPUs
Supercomputing News· Oct 28, 2026
Meta says MI350X GPUs reached 101.2% of Nvidia B200 training throughput across ten recommendation models after model tweaks and agent-based kernel porting.
Read article
Oct 27, 202616
- NVIDIA argues AI cluster network failover must run in hardware
Supercomputing News· Oct 27, 2026
NVIDIA claims hardware load balancing recovers from dropped links in under 3 ms, compared to over a second for software balancers running above NCCL.
Read article - Oracle will offer Cerebras wafer-scale AI inference on its cloud
HPCwire· Oct 27, 2026
Oracle plans to deploy Cerebras wafer-scale processor hardware on Oracle Cloud Infrastructure starting in 2027.
Read article - Microsoft showed its Maia 300 AI chip at OCP Summit
ServeTheHome· Oct 27, 2026
Microsoft showcased its custom Maia 300 AI accelerator silicon at OCP Summit 2026.
Read article - SoleCrypt expands planned AI data center in Tunisia to 40MW
Data Center Dynamics· Oct 27, 2026
SoleCrypt plans a 40MW campus in Bizerte with an initial 7MW phase deploying 1,728 Nvidia Rubin GPUs by early 2028.
Read article - Quantinuum and GE Vernova partner on quantum superconductivity simulation
HPCwire· Oct 27, 2026
GE Vernova will use Quantinuum's Helios quantum hardware to simulate Fermi-Hubbard models for high-temperature superconductivity research.
Read article - Oracle added Google TPUs, Cerebras, and SambaNova chips to OCI
The Next Platform· Oct 27, 2026
Oracle plans to offer Google TPUs, Cerebras CS-4 systems, and SambaNova RDUs in its cloud starting next year.
Read article - Broadcom arranged $63B in bank loans for Anthropic to buy TPUs
Data Center Dynamics· Oct 27, 2026
A group of banks agreed to loan $63 billion to finance Anthropic's purchase of Broadcom-developed Google TPUs starting in 2027.
Read article - Oracle will offer Google TPUs on OCI
Data Center Dynamics· Oct 27, 2026
Oracle plans to deploy Google's TPUs starting in 2027 through a multi-gigawatt cloud compute partnership.
Read article - MIT deployed Ferveret's in-rack immersion cooling on its HPC systems
Data Center Dynamics· Oct 27, 2026
Ferveret claims its self-contained immersion boxes deliver up to 35 percent more compute from the same power envelope.
Read article - PNNL used AI agents to accelerate radiation-hard chip testing
HPCwire· Oct 27, 2026
PNNL deployed a hierarchy of AI agents via SciLink to coordinate simulations and analyze ion beam damage data on radiation-hardened microchips.
Read article - UNSW Sydney chose DDN to replace 11 legacy storage systems
HPCwire· Oct 27, 2026
DDN and Arcitecta will deploy EXAScaler and Mediaflux to unify on-campus flash and central disk storage for the university's research workloads.
Read article - ChipAgents added analog and digital workflows to its chip design AI
HPCwire· Oct 27, 2026
The startup says its platform uses autonomous AI agents and domain models to handle tasks from RTL design to analog layout generation.
Read article - Spectra Logic launched Storage Reclaim to move cold data off flash
HPCwire· Oct 27, 2026
Spectra Logic says the tool analyzes unstructured data and moves inactive files to tape or object storage to free up primary capacity.
Read article - Parallel Works promotes engineering leaders for HPC and AI orchestration
HPCwire· Oct 27, 2026
Parallel Works promoted Michael McQuade to VP of Engineering to oversee development of its ACTIVATE hybrid cloud platform.
Read article - CERN partnered with Mistral to build a physics foundation model
HPCwire· Oct 27, 2026
They plan to train a physics foundation model called Flare on billions of simulated collisions from the Large Hadron Collider.
Read article - Oracle and Nvidia plan multi-gigawatt Vera Rubin clusters on OCI
HPCwire· Oct 27, 2026
They plan to scale OCI Superclusters with Nvidia Vera Rubin NVL72 racks and introduce HGX Rubin NVL8 systems with Intel Xeon 6 CPUs.
Read article
Oct 26, 202617
- Micron and Liqid pooled CXL memory to share KV caches across GPUs
Blocks & Files· Oct 26, 2026
They claim pooling KV caches over CXL boosted Llama 3 70B inference throughput by 2.8x while running within 99 percent of host DRAM speeds.
Read article - Intel agreed to acquire quantum software maker Classiq
HPCwire· Oct 26, 2026
Intel plans to pair Classiq's algorithm synthesis tools with its silicon spin-qubit hardware to build a full-stack quantum platform.
Read article - Nvidia simulated quantum error mitigation on the Jupiter supercomputer
The Next Platform· Oct 26, 2026
Nvidia ran a 104-qubit simulation on Europe's Jupiter supercomputer, claiming the tested method cuts qubit overhead by 90 percent over prior approaches.
Read article - Cadence and Nvidia teamed up on AI models for chip design
The Next Platform· Oct 26, 2026
Cadence claims its Nvidia Nemotron-tuned Maestro models deliver up to 30 percent better output quality and 5x token efficiency in EDA tasks.
Read article - DapuStor detailed how to offload KV caches to NVMe SSDs
Blocks & Files· Oct 26, 2026
They claim packing KV cache tokens enables sequential NVMe SSD I/O, providing a cheaper storage tier that avoids recomputing evicted context.
Read article - Nvidia bypasses JEDEC with custom NVHBM memory for GPUs
Blocks & Files· Oct 26, 2026
Nvidia moved the memory controller into the HBM base die, claiming up to 30 percent higher bandwidth and 25 percent more room for GPU compute cores.
Read article - Generative Bionics plans AI data center in Italy using AMD Helios racks
Data Center Dynamics· Oct 26, 2026
The company plans to use AMD Helios rack-scale systems to train robot action models and run digital-twin simulations by late 2027.
Read article - Arteco evaluated coolant additives for direct-to-chip data center cooling
Data Center Dynamics· Oct 26, 2026
Arteco says additive chemistry creates major differences in corrosion and stability across PG25 liquid coolants with the same base fluid.
Read article - DKRZ awarded Bull a 40M euro contract for its HLRE-5 supercomputer
HPCwire· Oct 26, 2026
The BullSequana XH3500 system uses Nvidia GB200 and Vera chips, which DKRZ says will cut energy per simulated year by about 75%.
Read article - Nagare Labs raised $5M to develop capacitor-less memory for AI
HPCwire· Oct 26, 2026
The startup claims its Z2RAM architecture combines SRAM speed with DRAM density and can stack directly on top of logic dies using standard fab processes.
Read article - SWI Compute ordered $1 billion in Nvidia systems for European AI data centers
HPCwire· Oct 26, 2026
The startup says it plans to build a 30 MW UK site with 7,000 Rubin GPUs and 3,500 Vera CPUs online by late 2027.
Read article - Researchers ran the Lumina cosmological simulation on Frontier
HPCwire· Oct 26, 2026
The simulation covers 100 times the volume of previous work to model the universe's first 2 billion years.
Read article - UChicago spinoff is designing an accelerator for graph and sparse workloads
HPCwire· Oct 26, 2026
They claim early tests show their accelerator delivers a 350x speedup over GPUs for graph and sparse matrix workloads.
Read article - Quantum Art published a quantum algorithm for wave-propagation simulations
HPCwire· Oct 26, 2026
They claim their physics-guided method avoids brute-force equation solving to model large-scale wave problems across acoustics, radar, and seismology.
Read article - SC26 marks 20 years of the Student Cluster Competition
HPCwire· Oct 26, 2026
The competition will host 10 on-site teams and 29 remote teams from 10 countries this November at SC26 in Chicago.
Read article - TACC joins $9M NSF project to scale National Science Data Fabric
HPCwire· Oct 26, 2026
TACC will handle user support for the $9 million NSF project connecting scientific instruments and supercomputers nationwide.
Read article - Cadence launched an Nvidia-powered AI data center in Munich
Data Center Dynamics· Oct 26, 2026
The facility hosts Cadence's Millennium supercomputer alongside Nvidia Blackwell GPUs and Grace and Vera CPUs for EDA simulations.
Read article
Oct 25, 20262
- NVIDIA launched Confidential Bare Metal for Grace Blackwell racks
Supercomputing News· Oct 25, 2026
They claim it lets tenants run custom operating systems while keeping decrypted model weights away from cloud operators.
Read article - Aivres showed server designs for Nvidia Vera Rubin and HGX B300
ServeTheHome· Oct 25, 2026
The vendor showcased server architectures built for Nvidia's Vera Rubin and HGX B300 platforms at the OCP Summit.
Read article
Oct 24, 20262
- Silk Cloud and Fuse Energy will deploy Nvidia Blackwell Ultra GPUs
Data Center Dynamics· Oct 24, 2026
Silk Cloud plans to deploy Nvidia HGX B300 systems in Georgia by 2027, while UK startup Fuse Energy says it will bring 128 GPUs online this December.
Read article - GF launched an integrated polymer control valve for liquid-cooled racks
Data Center Dynamics· Oct 24, 2026
GF says its polymer valve combines flow, pressure, and temperature sensing in one unit that is 70% lighter than metal setups.
Read article
Oct 23, 20261
- New Jersey launched a $125M shared supercomputer initiative built by SHI
Supercomputing News· Oct 23, 2026
The $125M system will sit in SHI's data center, with a nonprofit managing access for New Jersey research universities starting in 2027.
Read article
