Achieving the Exabyte Milestone

WEKA has officially introduced the WEKApod 3, a series of storage appliances engineered to maximize density for its latest NeuralMesh 6 software. The premier configuration, known as Prime Max, achieves a total effective capacity of 1.1 exabytes within a standard 56U rack, marking a significant industry milestone in storage density.


However, the transition to exabyte-scale is largely driven by intelligent software rather than physical hardware alone. While the system utilizes approximately 1,800 Micron 245.76 TB 6600 ION SSDs, the raw hardware capacity stands at 441.5 PB. This discrepancy highlights the critical role of WEKA's data reduction technologies.


Advanced Data Reduction and Performance

To reach these capacity goals, WEKApod 3 employs a suite of native features, including:

  • Fingerprinting and similarity hashing
  • Automated data deduplication
  • Advanced compression algorithms

According to the company, these processes operate with a write overhead of less than 5%. WEKA further asserts that for AI training workloads, the system can achieve up to 6x capacity savings, a claim supported by contractual guarantees regarding both performance and data reduction ratios.


Tiered Storage and Hardware Efficiency

The synergy between NeuralMesh 6 and the hardware is best demonstrated by the AlloyFlash tiering technology. This system intelligently directs latency-sensitive tasks to TLC flash, while bulk data is offloaded to more cost-effective QLC drives. The resulting configuration delivers 10.2 TB/s of throughput per rack alongside 210 million IOPS.


Jeremy Werner of Micron highlighted the physical efficiency of the design, noting, "The architecture paired with Micron's 245TB SSDs delivers 15.8 petabytes within a single 2U footprint."


Addressing Data Center Constraints

WEKA's focus on density is a strategic response to current infrastructure challenges, including:

  • A significant reduction in data center construction projects forecasted for 2025.
  • Long lead times for grid connections in major markets.
  • A projected 49-gigawatt power shortfall in the United States through 2028.

By optimizing storage to be as compact as possible, WEKA aims to ensure that storage hardware does not compete unnecessarily with GPUs for limited rack space and power budgets. Furthermore, by managing its own hardware supply chain, the company intends to offer customers greater predictability regarding pricing and delivery schedules.


The WEKApod Nitro, Prime, and Prime Max units are currently available for order, with shipments scheduled for autumn 2026. NeuralMesh 6 is expected to be commercially available in the second half of 2026, with free upgrades planned for existing clients.