At the SC26 supercomputing conference in Chicago on November 12, 2026, Hewlett Packard Enterprise announced high-density hybrid configurations for its HPE Cray Supercomputing Storage Systems E2000 alongside an intelligent Rack Management Controller for liquid-cooled clusters. The announcements introduce 32-terabyte hard disk drives paired with flash in 74- and 100-drive enclosures to tackle soaring solid-state storage costs for university high-performance computing centers and national laboratories facing fixed infrastructure budgets.
What HPE announced at SC26
Hewlett Packard Enterprise (HPE) introduced two major infrastructure solutions at SC26. The primary release is an updated hybrid tier for the HPE Cray Supercomputing Storage Systems E2000, pairing enterprise NVMe flash with 32-terabyte (TB) SAS-4 spinning drives in dense enclosures. In parallel, HPE launched the HPE Rack Management Controller, a dedicated telemetry and automation system designed to regulate power shelves and detect leaks in direct liquid-cooled supercomputing racks.
HPE rates the hybrid storage configuration at up to 116 GB/s read throughput and 92 GB/s write throughput per rack under the Lustre parallel file system, delivering up to 32% more usable capacity than earlier hybrid E2000 models. As reported by StorageReview, Marten Terpstra, HPE’s vice president of product management for HPC and AI Infrastructure Solutions, confirmed that headline ratings assume a baseline ratio of roughly 10% flash and 90% hard drive capacity, adjusted to customer workloads.
Terpstra stated that sharp price increases for high-capacity solid-state drives (SSDs) over the past 12 to 18 months prompted the release. With data-intensive AI and simulation datasets expanding, fixed institutional grants often force universities to trade compute nodes for flash storage. “Budgets are not keeping up with the cost of the infrastructure that people want to go and buy,” Terpstra noted.
Technical specifications of the hybrid E2000 system
The E2000 platform builds on the open-source Lustre parallel file system. The updated hybrid design moves beyond previous 24TB limits by integrating 32TB SAS-4 HDDs in top-loading enclosures holding 74 or 100 drives. Dual active-active storage server nodes powered by AMD processors connect over a 24 Gb/s SAS-4 fabric, providing redundant data paths across drive channels.
External connectivity supports HPE Slingshot 400, NDR InfiniBand (400 Gb/s), and 400GbE networks. System telemetry and administrative controls run on an independent System Management Unit (SMU) with dedicated SSDs, exposing Redfish and Swordfish APIs for cluster-level management.
| Metric | Previous Hybrid Baseline | SC26 Hybrid Configuration | Operational Role |
|---|---|---|---|
| Maximum HDD Capacity | 24TB SAS-3 / SAS-4 HDDs | 32TB SAS-4 HDDs | Expands secondary capacity for checkpoints and datasets |
| Enclosure Density | Up to 84 drives per shelf | 74 or 100 drives per enclosure | Increases usable petabytes per standard rack footprint |
| Rack Read Throughput | Sub-90 GB/s per rack | Up to 116 GB/s per rack (HPE rating) | Feeds parallel read streams across compute nodes |
| Rack Write Throughput | Sub-75 GB/s per rack | Up to 92 GB/s per rack (HPE rating) | Absorbs periodic checkpoint dumps from large jobs |
| Flash-to-Disk Ratio | Manual tiering profiles | ~10% Flash / 90% HDD baseline | Combines flash burst buffers with economical disk media |
The E2000 software stack includes Lustre data movers that monitor file activity and migrate aged data to spinning disks. HPE indicated future work will evaluate integrating these movers with schedulers like Slurm to pre-stage files to flash before jobs start.
Comparative claims and storage economics
HPE presented two key comparative figures during the announcement, reported by StreetInsider and StorageNewsletter. First, HPE claims up to 41% higher throughput in GB/s per rack than the closest hybrid competitor, identified as DataDirect Networks (DDN). Second, HPE claimed up to 82% lower acquisition costs than comparable all-flash systems, calculated against prevailing prices for 15PB of usable storage meeting NVIDIA specifications for a 1,024-GPU cluster.
These figures reflect HPE’s internal sizing models and have not been independently audited. HPE did not publish exact competitor bills-of-materials. Nonetheless, the cost argument addresses a genuine pressure point: in academic grants with strict funding caps, all-flash storage often forces reductions in GPU or CPU nodes. Hybrid architectures aim to provide required I/O bandwidth while keeping petabyte-scale storage affordable.
Liquid cooling and the HPE Rack Management Controller
To support high-density compute, HPE introduced the HPE Rack Management Controller. With modern GPU clusters drawing 300 kW to 500 kW per rack, liquid cooling is essential to prevent thermal throttling and equipment failure.
The controller provides real-time telemetry, power-shelf monitoring, and automated leak detection. If coolant loss occurs, it can isolate fluid valves, power down individual blades, or execute an orderly rack shutdown before fluid damages electronics. The unit adapts controller designs refined over seven years on Cray EX systems into a modular component for broad deployment.
The controller debuts on the HPE Cray Supercomputing GX5000, HPE’s next-generation platform. It reports into the HPE Supercomputing Management Software suite (Foundation, Composable, and Observability) and provides open APIs for integration with institutional facility systems.
What it means for researchers, universities, and research offices
The announcements have direct implications for university computing centers and research administration:
- Grant proposals and procurement: Principal investigators applying for major instrumentation grants (such as NSF MRI or NIH S10 awards) can balance compute and storage budgets. Sizing hybrid storage preserves funding for compute nodes without sacrificing petascale capacity.
- Data lifecycle and repository integration: Slower 32TB drive tiers offer economical on-premises staging for datasets before submission to an external data repository, aligning with requirements tracked through the DMP lifecycle.
- Laboratory workflow pipelines: Automated data movers in parallel storage simplify handoffs from instrument clusters into an institutional scientific data management system (SDMS) without manual file migration.
- Facilities and sustainability: Liquid cooling leak mitigation protects campus datacenter facilities. Telemetry enables universities to monitor their compute carbon footprint and assess the efficiency of cold-storage energy in research operations.
Implementation timelines and what happens next
HPE confirmed commercial availability across early 2027. Hybrid Cray E2000 configurations with 32TB drives will ship in early 2027, with pricing determined per configuration based on drive counts and flash ratios.
The HPE Rack Management Controller is scheduled for March 2027 on Cray Supercomputing GX5000 systems. Support for AMD and NVIDIA rack-scale platforms will follow in June 2027, before expanding to ProLiant DL server racks. Research institutions planning 2027 deployments should factor these delivery schedules and liquid-cooling requirements into facility retrofit plans.
Sources
- Hewlett Packard Enterprise: HPE Addresses Industry Demand for Scalable HPC Solutions with World’s Fastest Hybrid Supercomputing Storage and Intelligent Rack Management — November 12, 2026. Checked 2026-11-15.
- StorageReview: HPE Cray E2000 Hybrid Storage Moves to 32TB SAS-4 HDDs, Rated for 116GB/s Reads per Rack, by Brian Beeler — November 12, 2026. Checked 2026-11-15.
- StreetInsider: HPE unveils new supercomputing storage and rack management tools — November 12, 2026. Checked 2026-11-15.
- StorageNewsletter: SC26: HPE Addresses Industry Demand for Scalable HPC Solutions with World’s Fastest Hybrid Supercomputing Storage and Intelligent Rack Management — November 13, 2026. Checked 2026-11-15.







