Ronni Shendar

August 05, 2025

[8 min read]

Inside UltraQLC: The Enterprise SSD Platform Engineered for AI

Ronni Shendar

August 05, 2025

[8 min read]

When top engineers at Sandisk set out to architect a new platform from the ground up, they drew on decades of expertise with a clear goal: create an entirely new architecture that would streamline the next generation of enterprise storage solutions.

That vision became reality at the Future Memory and Storage (FMS) event this week, where Sandisk previewed its groundbreaking 256TB NVMe™ Enterprise SSD, built on the new UltraQLC™ platform.

Behind this breakthrough lay a multi-year journey that began with an ambitious challenge.

A bold mandate

In 2021, Sandisk’s Chief Product Officer, Khurram Ismail, approached Vice President of Engineering, Tsiko Shohat Rozenfeld, with a bold mandate: to break from current product lines and build a best-in-class next-generation enterprise SSD platform. The sense of possibility was undeniable, but so was the weight of responsibility.

In the client market, the company’s platform had supported nearly a decade of successful product derivatives—all built on a single, solid architectural foundation. But applying that same approach to the enterprise space would require entirely new thinking.

Trying to build something new from scratch in the enterprise space was a huge and ambitious goal; monumental from an engineering accountability perspective,” Shohat said about taking on his leadership role.

Shohat went on to hand-pick the top subject matter experts from across the company. He wanted to apply the learnings across client, consumer, and enterprise products—to leverage every success but push beyond the limits of past thinking.

It’s rare to pull engineers from their day-to-day work and fence them off in a future-facing project. And it’s a high-stakes move that carries risk to revenue-generating products. But if they got it right, a foundational platform could provide a long-term return on investment well beyond a single product.

Building a platform from scratch is a little like that first shot in golf: you want to get as close to the goal and as far as possible with that first attempt,” he said. “Any degree of deviation or misalignment could potentially create a disaster down the line.”

Building is believing

For the next few years, that core team explored, invented, and debated fiercely how to build the foundation for the best enterprise SSD platform on the market.

Once the silicon was in place, the focus shifted to productization. Turning the platform into a working, validated product meant sweating the details and shaping it into something customers could trust, test, and build on.

Coordinating the process was Ilya Gusev, Senior Director of Systems Design Engineering at Sandisk and UltraQLC’s Product Development Team lead.

Gusev isn’t shy about the challenges of the undertaking. After all, it demands cross-functional collaboration between several hundred engineers across more than a dozen groups—ASIC design, hardware validation, mass production, firmware, system design, memory systems and memory health, packaging, marketing, and more.

Building a platform from scratch is, by definition, painful and challenging,” he said. “It’s not just about solving technical challenges for the company’s most ambitious platform; it’s also about changing a mindset to break from legacy thinking.”

For Hyuk-Il Kwon, Senior Director of Firmware Engineering at Sandisk, that mindset shift meant reshaping how teams worked. He reorganized groups to mix key talent and draw from experience across the company’s portfolio. His large team is fully focused on the efforts to bring the first UltraQLC drives into customers’ hands.

It's a rare opportunity in a career to have a chance to work on something new and build it from scratch. It comes with a lot of excitement, but it also comes with challenges,” he said.

Architecting ultra-high capacity

One of the critical turning points in the process was the decision to focus efforts on ultra-high capacity SSDs. With some of the world’s AI data centers now managing over an exabyte of data, hyperscalers are increasingly viewing high-capacity QLC SSDs as a compelling alternative to hard drives for AI data lakes.

The enterprise market is changing, and focusing our efforts on QLC ultra-high capacity was a tough moment, but the most impactful decision in the process,” Kwon said.

With that clarity, the teams zeroed in on what it would take to deliver an exceptional product, focusing on features customers cared about most.

AI data lakes need fast, high-throughput storage access, so it was clear we want to fully utilize the PCIe Gen 5 interface,” Kwon said. “On the platform end, we’ve automated critical data paths into the hardware section. But for the SSD storage architecture, which includes the ASIC and the firmware, the challenge was how to optimize performance and maximize the interface bandwidth.”

There were also unexpected challenges that emerged due to the sheer capacity, explained Mike James, Senior Director of Enterprise SSD Systems Architecture at Sandisk.

When you're getting into higher and higher capacities of SSDs, interesting challenges emerge. One example is the challenge of data recycling on NAND at this scale,” James said. “You can’t overwrite 128 terabytes every few days—it’s not effective or efficient. So, we’re constantly finding new ways to truncate how much needs to be recycled and inventing novel methods to reduce the impact of recycling in the background.”

James leads a global team of architects, distinguished engineers, and technologists who negotiate the system-level trade-offs within an SSD. They guide a product from its platform inception all the way through to the customers’ hands, collaborating on feature and performance solutions.

He explained that the innovation in UltraQLC lies in fundamentally rethinking existing processes, down to the core philosophy of how individual I/Os are managed.

Instead of trying to brute force a solution, we took a step back to see the forest instead of the trees. You can work to change a single algorithm, or you can change the entire approach to how you get there,” he said.

Bringing the possibilities into customers’ hands

James and his team are already busy at work with the company’s UltraQLC roadmap to a one petabyte SSD. What he is most excited about, though, is the shift he sees in how customers are thinking about ultra-high-capacity SSDs.

Building on that momentum is Shai Tubul, Senior Director of Program Management Engineering at Sandisk. He and his team are leading the execution of the first ready-to-ship SSD built on the new UltraQLC platform. While the Sandisk UltraQLC 256TB NVMe SSD will be available in U.2 form factor in early 2026, Tubul is preparing to put the first 128TB drives into customers’ hands for testing in just a few weeks. It’s a major milestone that reflects the work of hundreds of engineers who made it possible.

For Tubul, UltraQLC comes down to how Sandisk can deliver better customized, best-in-class solutions to hyperscale customers more quickly.

As customers deploy hundreds, and soon thousands, of SSDs in a system, the performance to power ratio has become a defining metric and it’s where our UltraQLC platform delivers game-changing results,” he said. “But beyond that, it’s about serving a fragmented market, where each customer has customized needs. UltraQLC gives us a robust and flexible architecture so we can deliver tailored products much faster into customers’ hands.”

As Ilya Gusev looks back at the journey of productization, he can’t help but feel immense pride. “It is nothing shy of amazing that we've been able to deal with all the complexity of delivering a new, customized product for a customer in such a short time,” he said. “That’s the power of the platform—being agile and flexible. I’m incredibly proud to be part of this team.”

For Tsiko Shohat Rozenfeld, who kickstarted that initial team, the bold bet to reshape the company’s future is finally unfolding.

If you look at major technology companies’ attempts to repivot their engineering organizations toward AI, success isn’t guaranteed. The odds are against you,” he said about the undertaking. “But we were able to do it. It’s a testament to our culture and what I’m most proud of: the people.”