• 2 min read
WEKA claims 1.1 exabytes in one storage rack
WEKA claims its WEKApod 3 Prime Max can deliver 1.1EB in a 56U rack using compression, deduplication, and 441.5PB of raw flash.
Image: TechRadar
WEKA is claiming 1.1 exabytes of effective capacity from a single 56U rack with its new WEKApod 3 Prime Max storage appliance. The figure depends on software-driven data reduction: the rack contains roughly 1,800 Micron 245.76TB 6600 ION SSDs, providing 441.5PB of raw flash capacity.
How WEKApod 3 reaches exabyte-scale capacity
WEKApod 3 is designed to run WEKA’s NeuralMesh 6 software at rack scale. Fingerprinting, similarity hashing, deduplication, and compression are enabled by default, while write overhead is kept below 5%.
WEKA claims up to 6x capacity savings for AI training data, backed by a contractual guarantee covering both the reduction ratio and performance. The system places 70 NVMe drives in a compact two-rack-unit, two-node chassis.
Recommended reading
Zeiss expands EUV optics production for 2nm chips
Its AlloyFlash tiering automatically directs latency-sensitive workloads to TLC flash and moves bulk-capacity data to QLC drives, which WEKA says cost roughly 30% to 40% less per terabyte. Per-rack performance reaches 10.2TB/s and 210 million IOPS. WEKA says that represents a 114% advantage in throughput density over the next-best publicly available alternative.
Jeremy Werner, senior vice president at Micron’s Core Data Center Business Unit, said the architecture combined with Micron’s 245TB SSDs delivers 15.8PB in a single 2U footprint. Steve McDowell, chief analyst at NAND Research, said production-scale inference requires different buying metrics from training, including tokens per rack and cost per inference.
Availability and data center constraints
WEKA attributes the design to mounting infrastructure constraints. US data center construction fell in 2025, while grid-connection queues in major markets now take four to seven years. Morgan Stanley has forecast a 49-gigawatt US power shortfall through 2028, alongside continuing NAND supply constraints.
The company argues that inefficient storage competes directly with GPUs for limited rack space and power. By managing its own hardware supply chain instead of depending on third-party OEM channels, WEKA says it can provide more predictable pricing and delivery timelines.
NeuralMesh 6 is scheduled to launch in the second half of 2026, with existing customers receiving upgrades at no additional cost. WEKApod Nitro, Prime, and Prime Max are already orderable through WEKA’s distributor network, with deliveries beginning in fall 2026.
Enterprise Editor
Marcus follows the money. He covers enterprise software, cloud architecture, and the tectonic shifts in Big Tech strategy. He translates dense earnings calls and complex M&A activity into actionable insights about where the industry is actually heading. If a tech giant makes a silent pivot, Marcus is usually the first to notice.
via TechRadar
