The transition to PCIe 6.0 storage is gathering pace as controller makers and drive vendors outline products that combine very high throughput with unprecedented capacities. Recent announcements point to sequential reads as high as 28 GB per second and design envelopes that scale to as much as 512 terabytes per drive. Early adoption will focus on cloud and AI infrastructure, with hyperscalers looking to reduce latency and power per bit while feeding compute clusters more efficiently.

Silicon Motion detailed its MonTitan SM8466 enterprise controller during the Future of Memory and Storage events this summer. The company targets a PCIe 6.0 x4 interface and 16 NAND channels, enabling up to 28 GB per second of sequential read bandwidth and around 7 million random IOPS. The controller architecture supports NVMe 2.0 class features and is designed for capacities up to 512 TB, which places it squarely in hyperscale and cloud storage territory rather than consumer desktops. Analysts expect silicon partners to bring SSDs based on this controller to market over the next product cycles for AI servers and storage nodes.

Micron has already shown working PCIe 6.0 hardware on the drive side. Its data center grade 9650 SSD was introduced as the first PCIe Gen6 unit with up to 28 GB per second sequential reads and up to 14 GB per second sequential writes. Micron positions the 9650 for AI training and inference back ends and pairs the drive with networking and retimer components to validate end to end PCIe 6 interoperability. The company’s product pages and investor materials corroborate the claimed performance and target markets, and independent trade coverage notes that the initial capacities top out well below 512 TB since Micron’s device is focused on performance and latency.

A second thread involves South Korean controller vendor FADU, which unveiled its Sierra FC6161 PCIe Gen6 controller with headline specifications that mirror the 28 GB per second read target while keeping power draw under roughly 9 watts and enabling theoretical capacities up to 512 TB depending on the NAND configuration. Press and industry coverage around FADU’s tenth anniversary and conference keynotes indicates the firm has secured mass production agreements with two of the largest hyperscalers. Although FADU did not publicly name those customers, reporting and event appearances have led to speculation that Meta could be among the first to deploy the technology. VideocardzTechRadar

For end users, PCIe 6.0 SSDs will not arrive in consumer PCs soon. Multiple outlets note that controller cost, thermals, and platform readiness make PCIe 6.0 a data center story for the near term. Client systems will continue to standardize on PCIe 4.0 and 5.0 while server boards and AI accelerators absorb the first wave of PCIe 6.0 storage to alleviate I/O bottlenecks.

The broader context is familiar. Every PCIe generation has doubled per-lane bandwidth, and Gen6 uses PAM4 signaling to achieve 64 gigatransfers per second per lane, which allows an x4 drive to approach the 28 GB per second class speeds vendors are now quoting. Pairing that bandwidth with larger channel counts and more advanced error correction is what enables controller makers to talk about extremely high capacities, even if initial shipping products will prioritize balanced performance points for AI and cloud workloads. Micron’s early lead with a shipping Gen6 device and FADU and Silicon Motion’s roadmaps suggest that real deployments will scale through 2026 and beyond inside large data centers, with consumer adoption remaining a longer term prospect.