- WEKApod 3 reaches exabyte capacity via software coupled with dense Micron storage
- NeuralMesh 6 converts 441.5 PB of flash memory into exabyte-scale storage capacity
- WEKApod 3 relies on compression to push the limits of raw storage hardware
WEKA has launched WEKApod 3, a line of storage devices built specifically to run its new NeuralMesh 6 software at rack-scale density.
The flagship Prime Max configuration claims to have 1.1 exabytes of effective capacity within a single 56U rack, positioning it as the first system to cross that threshold.
Using Micron’s 245.76 TB 6600 ION SSD in approximately 1,800 drives per rack produces just 441.5 PB of raw hardware capacity, much less than half an exabyte.
Latest videos ofTechnologyRadar
Software Tiering Closes the Capability Gap
This means that advertised exabyte-scale capacity is highly dependent on data reduction technologies and not solely on installed flash.
That configuration includes 70 NVMe drives in a compact two-rack unit, two-node chassis for maximum density.
Fingerprinting, similarity hashing, deduplication, and compression are run by default in each implementation, with write overhead kept below 5%.
WEKA claims to save up to six times capacity specifically on AI training data, backed by a contractual guarantee covering both reduction rate and performance.
This is where NeuralMesh 6 and WEKApod 3 really work as an integrated product rather than separate ads.
AlloyFlash tiering automatically routes latency-sensitive operations to TLC flash memory while sending high-capacity data to QLC drives, which run approximately 30-40% cheaper per terabyte.
Performance per rack reaches 10.2 TB/s with 210 million IOPS, figures that WEKA says outperformed the next best publicly available alternative by 114% in performance density.
Jeremy Werner, senior vice president of Micron’s core data center business unit, said the architecture combined with Micron’s 245TB SSDs offers 15.8 petabytes in a single 2U footprint.
Steve McDowell, chief analyst at NAND Research, argued that production-scale inference workloads demand different metrics than training, pointing to tokens per rack and cost per inference as the numbers buyers should now track.
Limitations Force Storage Vendors to Completely Rethink Hardware
WEKA points to several structural pressures behind this hardware focus, starting with a documented drop in US data center construction in 2025.
Grid connection queues in major markets now stretch between four and seven years before new capacity comes online.
Morgan Stanley has forecast a 49 gigawatt power shortfall across the United States that will extend into 2028, exacerbating existing NAND supply constraints.
WEKA argues that inefficient storage competes directly with GPUs for the same limited rack space and power budget.
By managing its own hardware supply chain rather than relying on third-party OEM channels, the company says it can offer more predictable pricing and delivery times.
NeuralMesh 6 will hit shelves in the second half of 2026 and existing customers will upgrade at no additional cost.
WEKApod Nitro, Prime and Prime Max can now be ordered through the WEKA dealer network and deliveries will begin in autumn 2026.
Via Storage Review
Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds.
