Powering AI at scale: Dropbox qualifies Seagate’s Exos 32TB drives

Dropbox and Seagate took the next step in scalable, sustainable infrastructure this June when Dropbox successfully qualified Seagate's HAMR-enabled Exos® 32TB hard drives — powered by the Mozaic™ platform — for deployment in Dropbox data centers. This marks a milestone in Dropbox’s ability to scale for AI workloads while advancing its sustainability goals. The mass-capacity Mozaic platform enables Dropbox to expand capacity without rearchitecting its infrastructure, and to do so without compromising on power, space or carbon footprint

To explore the significance of this qualification and the broader industry context, we chatted with Dropbox’s infrastructure team, Ali Dasdan, CTO, Jesse Lee, Tech Lead, and Vishal Manannal, Storage Technologist.

How are evolving AI workloads reshaping your data infrastructure priorities — especially around storage capacity, performance and efficiency?

Dropbox: AI is fundamentally changing how we think about infrastructure. It’s not just about storing more data — it’s about storing smarter. The focus is shifting from raw capacity to holistic efficiency. As AI models grow more complex and data-hungry, at Dropbox, we recognize that performance, scalability and efficiency are no longer trade-offs but mutually aligned.

AI data is dynamic — constantly ingested, generated and processed across diverse formats like text, images and audio. Each stage in the AI lifecycle imposes distinct demands on storage as larger and richer data sets grow. Selecting the right storage architecture — which is a strategic mix of hard drives and SSDs — has become a critical differentiator for AI success, from application speed to cost optimization and efficiency.

How is AI at scale shaping storage infrastructure decisions?

AI workloads span hot, warm and cold data stages, each with distinct performance and cost requirements. For example, offline document ingestion and archival workloads benefit from hard drives’ cost efficiency and scale, while SSDs accelerate operational workflows where low latency is paramount.

Our architecture reflects this:

• HBM/DRAM powers compute-side throughput for real-time training and inference.
• SSDs handle hot data — latency-sensitive tasks, model training and real-time inference — where speed and IOPS are critical.
• Hard drives anchor warm and cold data tiers. They support scalable ingestion, long-term retention and intermittent access — ideal for AI data lakes, historical training sets and inference archives. As hard drive technology like HAMR advances, they will only deliver better density, durability and energy efficiency needed to support AI growth while advancing sustainability goals.

Why did Dropbox choose to integrate Seagate’s Mozaic platform? How does innovation at the device level — specifically, the leap with Mozaic — enable Dropbox’s data ambitions or solve technical and operational needs that the PMR technology couldn’t?

As AI workloads intensify, we prioritize storage density, latency control and cost management without compromising performance. That’s why we integrated Seagate’s Mozaic platform. This next-generation solution delivers a leap in areal density through HAMR technology, supported by Seagate’s advanced media, redesigned read/write heads, evolved firmware and seamless host integration — all decisive factors in our adoption.

Seagate and Dropbox also co-engineered a high-density chassis optimized for acoustic, vibration and thermal dynamics that ensured low-latency access even as we scale.

The result? A win-win partnership that enables Dropbox to future-proof our infrastructure while advancing our AI and data ambitions with precision and confidence.

How was Dropbox able to adopt HAMR technology without major retooling of your architecture?

Our forward-looking infrastructure strategy enabled frictionless integration of HAMR drives — no new racks, no complex redesign and no major infrastructure overhauls of existing architecture.

Dropbox’s storage architecture was intentionally designed for adaptability. By previously adopting shingled magnetic recording (SMR) for its Magic Pocket platform, the groundwork has been laid for future innovations like HAMR. Mozaic drives aligned with Dropbox’s existing software stack, leveraging familiar APIs and zone-based write patterns, allowing frictionless adoption of advanced technologies and seamless scalability without rearchitecting. We've been able to scale to higher-density storage with minimal disruption — delivering operational excellence and long-term business value, from improved cost efficiency to a reduced energy footprint, all without compromising reliability or performance.

Dropbox has long focused on operational sustainability. As AI drives up storage needs, to what extent does business efficiency equal operational sustainability — or do they sometimes conflict?

We are consolidating workloads onto fewer, denser platforms and have seen early wins that deliver a dual impact: both operational efficiency and sustainability gains. Higher storage density per server means fewer servers — directly reducing power, cooling and maintenance overhead. At the same time, it optimizes data center space, minimizing e-waste and raw material consumption.

The early investment pays off as the hardware upgrade significantly lowers TCO across the lifecycle, aligning financial prudence with long-term sustainability.

What early efficiency or sustainability gains has Dropbox achieved with higher-capacity drives, and how do these early results scale to long-term impact?

Adopting Seagate’s Mozaic platform has been a game-changer in building Dropbox’s most cost-efficient data centers to date. By significantly increasing drive areal density, we’ve optimized our TCO on multiple fronts — reducing the number of drives and racks and the amount of network infrastructure needed to achieve the same storage targets.

The immediate gains are clear: moving from our 4th to 7th generation hardware, we’ve cut power consumption by 50% and reduced physical footprint by 75% per exabyte of storage. This translates into tangible OpEx savings across power, cooling and drive maintenance, while lowering CapEx by minimizing hardware and infrastructure requirements.

Beyond the operational efficiencies, these early wins stack up to a larger, long-term sustainability impact — shrinking our carbon footprint while enhancing our ability to scale seamlessly as data demands grow. The strategic payoff is a data center ecosystem that’s leaner, greener and built for the AI-driven workloads of tomorrow.

Looking ahead, how does investing in high-capacity, energy-optimized platforms like Mozaic help Dropbox scale future workloads while meeting ESG goals?

Investing in major platforms like Mozaic allows us to advance two critical priorities in parallel: scalable performance and environmental responsibility.

Mozaic solves challenges that legacy PMR drives couldn’t. By deploying mass-capacity hard drives in high-density architectures, we gain access to unprecedented areal density of more than 3TB per disk. Mozaic-enabled drives deliver optimal performance at scale for a vast majority of modern workload tasks.

At the same time, Mozaic helps us consolidate server hardware and use fewer drives, which in turn lowers energy consumption per petabyte stored, shrinks our physical infrastructure footprint and reduces both cooling needs and raw material usage. This translates to a continuous year-over-year reduction in our watts-per-petabyte (W/PB) metric — a direct driver of both operational efficiency and carbon impact reduction.

As data continues to grow, platforms like Mozaic will be essential in future-proofing our infrastructure and creating an agile, resilient and climate-conscious data foundation.

We’re seeing a shift across the industry from general-purpose systems to infrastructure optimized for specific workloads. How is Dropbox approaching this trend and serving the needs around emerging AI opportunities?

Dropbox is advancing its workload-optimized infrastructure strategy to balance performance with sustainability — critical for AI scalability.

As AI workloads intensify, Dropbox is moving decisively away from legacy general-purpose systems, which are energy-inefficient and unsuited for modern inference demands. Instead, we're deploying high-efficiency compute platforms purpose-built for AI, right-sizing workloads to reduce over provisioning and continuously improving power usage effectiveness (PUE) across our co-location facilities.

Operating as a hybrid cloud, Dropbox also strategically offloads select AI tasks to optimize resource utilization and reduce internal energy burden — while maintaining control over cost and sustainability trade-offs.

This infrastructure-first approach reflects Dropbox’s commitment to responsible innovation and positions us to scale AI capabilities without compromising environmental or operational integrity.

How do Dropbox’s infrastructure decisions today help future-proof its platform for AI models that will only get larger, more disaggregated and more power-hungry?

The infrastructure decisions we make today will continue to shape the future. We are architecting for scale, precision and adaptability — building our infrastructure strategy on three pillars: platform integration, supplier alignment and operational flexibility.

As demand for AI solutions grows, Dropbox is introducing dedicated hardware tiers, exploring liquid cooling and refining acoustic and thermal management — especially as HAMR-based storage scales and becomes a cornerstone of our future-ready infrastructure.

This mindset of continuous innovation will be the focus area for Dropbox.

Why was partnering with Seagate the right strategic choice for Dropbox?

Dropbox and Seagate share a truly collaborative, innovation-led and multi-faceted partnership that goes far beyond a buy-and-sell engagement.

Together, we’ve co-engineered and deployed advanced storage technologies — from enabling SMR to jointly characterizing HAMR-enabled server systems, and integrating Seagate’s Mozaic platform into Dropbox’s production environment. This collaboration has allowed Dropbox to scale high-density storage with precision and efficiency.

Beyond infrastructure, our go-to-market alignment includes Seagate’s Ultra Touch product line, where customers benefit from automatic backup to Dropbox via external hard drives. Dropbox also created a unique customer profile as Seagate’s fully enabled storage partner by leveraging Seagate hard drives on-premises and through Lyve Cloud.

This partnership exemplifies how strategic alignment and shared innovation can unlock new possibilities for data platforms at scale.

Finally, for peers wrestling with where to invest to keep up with AI-driven data demands while staying sustainable, what’s your advice on preparing storage and compute infrastructure to handle the next wave?

The future of infrastructure is about scaling smart — not just scaling big. Our three-point strategy for future-ready infrastructure is as follows: first, build an innovation engine. Second, unify infrastructure strategy. Third, invest in strategic partnerships.

To stay ahead in performance, efficiency and sustainability, it's crucial to build an innovation engine by embracing emerging technologies — platforms like Seagate’s Mozaic are great examples, offering gains in both density and energy savings. At the same time, it’s important to unify your infrastructure strategy by thinking holistically about storage and compute. Especially with AI workloads, tight integration between the two helps maximize throughput and reduce waste. And finally, scale isn’t something you achieve alone. Strategic partnerships — like the one between Dropbox and Seagate — play a key role in accelerating capabilities, co-engineering solutions and future-proofing operations.