AI infrastructure has a storage density problem. Here's how to solve it.

I’ve spent a lot of time talking about the importance of using the right tool for the right job. It's a principle that runs deep at Micron, and it applies not just to our products but to the close, engineering-level collaboration we build across the ecosystem. The best outcomes happen when two companies independently solve different parts of the same problem and then bring those solutions together to create something greater than either could deliver alone.

15.8 petabytes of storage in 2U of rack space. That is what happens when two companies work together toward a common goal. On July 21, WEKA launched an all-new line-up of WEKApod™ systems to meet the capacity and performance density AI data centers require. The highest capacity-dense system, WEKApod Prime Max, fits up to 64 Micron 6600 ION 245TB SSDs. The performance-optimized WEKApod Nitro carries up to 56 Micron TLC drives (e.g. Micron 9550) for maximum performance density. Here's why that matters.

Data center space is precious

If you follow AI infrastructure, you've noticed the conversation has shifted to a simple question. How do I get more out of my physical data center? Proximity to power grids, limited available power, cooling limitations, and even government regulations in some areas are holding back data center construction. WEKA notes that US data center construction growth slowed in 2025 for the first time since 2020 (CBRE.com).¹ Connecting to electric grids in major markets often takes more than four years (Berkeley Lab).² That means every square foot of data center real estate is precious.

Storage has historically occupied a large share of the data center, sometimes inside servers and at other times in large storage arrays. Every rack unit that needs high-capacity storage but isn't operating at peak density is one that could hold compute. So, what’s the answer? Store more data in less space to free up room for more AI infrastructure.

A quarter petabyte in a single drive

Recently, we announced that the 245TB Micron^® 6600 ION SSD is now shipping in production. It's built on the Micron G9 QLC NAND that is at least one generation ahead of any competing QLC NAND used in data center SSDs today. Available in E3.L or U.2 form factors, a single 245TB Micron 6600 ION drive delivers a quarter petabyte of capacity while consuming only 30 watts at peak power. Compared to equivalent HDD-based deployments, customers need 82% fewer racks to achieve the same raw storage capacity.³

That's not a roadmap item. That's a drive you can order and deploy today. And if you've read my previous blogs, you know how strongly we feel about announcing products when they're actually ready.

Better together: WEKApod and the Micron 6600 ION SSD

WEKA explains that it designed the latest WEKApod from the ground up as a purpose-built AI storage appliance. They engineered a custom 2U chassis around a PCIe Gen6 internal fabric, a cable-based backplane-free drive interconnect, and a software-managed thermal architecture that keeps the system running at 35°C ambient temperature without triggering shutdown. It is an impressive engineering design, and the results speak for themselves.⁴

When you pair that chassis with the 245TB Micron 6600 ION, the density numbers become extraordinary. A fully populated WEKApod Prime Max holds up to 64 Micron 6600 drives in 2U, delivering over 15.8 petabytes of storage capacity.⁵ Scale that across a full 56U rack, and you reach over 441 petabytes, or more than 1.1 exabytes of effective capacity with WEKA's NeuralMesh™ data reduction (assuming a 3:1 data reduction ratio). WEKA notes that this makes WEKApod Prime Max the first single-rack system to break the exabyte barrier.⁶ WEKApod Nitro delivers unmatched performance density with throughput reaching 10.2 TB/s per rack. IOPS hit 210 million per rack.⁴ For AI cloud providers and hyperscalers building out inference infrastructure, this can translate to more GPUs per rack, more tenants per rack, and more tokens generated from the same physical footprint. For enterprise AI teams, it means scaling production inference without expanding the data center budget or waiting years for new capacity to come online.

It's a small detail that's easy to miss, but WEKApod's architecture is built on PCIe^® Gen6. That readies the platform for drives like the Micron 9650, the world's first PCIe Gen6 data center SSD.⁷

A collaboration that keeps growing

WEKA selected the Micron 6600 ION as the primary SSD for WEKApod Prime based on capacity leadership, time to market, and energy efficiency. Our teams are working together to ensure the drive performs optimally within the platform. That is ecosystem collaboration done right. It brings together two companies focused on their respective strengths, with the customer benefiting from both.

It’s awesome to see this amazing platform from WEKA come to life. What Micron brings is the storage foundation that helps unlock the full density and performance potential of that design. When you put this incredibly dense AI storage appliance together with the world's highest-capacity SSD, customers win. That's the “better together” story, and it's one we're proud to be part of.

That's the value of collaboration. Customers don't have to wonder whether the pieces will fit together and work together. The foundation is already in place.

Congratulations to the WEKA team on the launch of WEKApod. We are proud to be part of this story, and I'm looking forward to what we build together next.