In 2016, NVIDIA introduced DGX-1 and changed the trajectory of accelerated computing. For DDN, it marked the beginning of a new chapter. As NVIDIA celebrates the 10th anniversary of DGX, that chapter shows how DDN became the company it is today. The problem that connected them was simple to state and hard to solve: graphics processing units (GPUs) deliver their speed only if data reaches them fast enough.
Much of DDN’s position today was shaped long before the current AI boom. The company spent decades solving data problems in some of the world’s most demanding computing environments, experience that became increasingly relevant as accelerated computing changed the demands on modern infrastructure.
NVIDIA was at the center of that change. As NVIDIA pushed accelerated computing forward, DDN brought high-performance data infrastructure expertise to increasingly powerful GPU environments. Over the decade that followed, the two companies solved problems at the intersection of compute and data that helped shape DDN’s transition from supercomputing into AI infrastructure.
Neither company started out building what it builds today. NVIDIA began in graphics, betting that accelerated computing would fundamentally change what computers could do. DDN came from high-performance computing (HPC), building data infrastructure for national laboratories, research institutions, and organizations tackling some of the world’s largest computational problems.
Different companies, markets, and starting points, but with something in common: a tendency to follow difficult problems and push beyond the limits of existing technology. Innovative companies have a way of finding each other. This is the story of what happened when their paths converged.
DDN Before AI
Long before generative AI became a boardroom conversation, DDN built its reputation solving data problems at scale.
For nearly two decades, DDN worked alongside leading supercomputing organizations, building expertise in parallel data access and in what happens when compute outpaces the infrastructure feeding it. At extreme scale, infrastructure couldn’t simply become bigger. Performance had to hold as systems scaled.
That experience became the foundation for DDN’s move into AI.
Accelerated computing increased the density and speed of processing, placing new demands on the infrastructure required to keep pace. GPU computing introduced different workload behaviors and performance requirements, and eventually a new model for building computing systems.
DDN already knew how to feed some of the fastest computers in the world. DGX gave that expertise a new problem to solve.
DGX and a New Direction for DDN
The arrival of DGX-1 in 2016 marked a shift. NVIDIA packaged accelerated computing into a system designed specifically for deep learning, making significant computational capability easier to deploy and use.
It also exposed an infrastructure challenge: more powerful GPUs needed data fast enough to remain productive. As compute performance increased, the ability to keep GPUs fed with data became increasingly important to the performance of the entire system.
DDN began applying decades of HPC experience to GPU-accelerated workloads. By 2018, that work had become DDN A³I (Accelerated, Any-Scale AI), data infrastructure engineered specifically for AI and deep learning environments, including NVIDIA DGX.
A Collaboration That Made Both Companies Better
As DDN moved deeper into AI, its relationship with NVIDIA deepened as well. NVIDIA continued to increase the power and density of accelerated compute. DDN focused on ensuring data did not become the constraint, bringing decades of experience moving enormous amounts of data in parallel to keep GPU environments supplied, productive, and scalable.
The benefit flowed in both directions. Working closely with NVIDIA gave DDN an early view into infrastructure challenges that would eventually confront the broader AI market. Each advance in accelerated computing created new demands on the data layer and new problems for DDN engineers to solve at scale.
The relationship was never exclusive. NVIDIA was building a broad ecosystem around accelerated computing, while DDN continued to work across supercomputing and technology communities. The significance of the collaboration came from the work itself and the opportunity to solve emerging problems together.
Selene: Putting the Strategy to the Test
By 2020, DDN’s move into AI had progressed from strategy to execution at significant scale, and Selene immediately tested that progression.
NVIDIA built Selene using DGX A100 systems backed at launch by 7 petabytes of DDN A³I AI400X all-flash storage. It debuted at No. 7 on the June 2020 TOP500 list of the world’s fastest supercomputers, climbed to No. 5 that November, and was a demanding environment in which to prove infrastructure designed specifically for AI.
For DDN, Selene demonstrated that its HPC expertise could meet the demands of a new generation of accelerated computing. It also reinforced the importance of designing compute and data infrastructure as parts of a complete system: fast GPUs could deliver their potential only if the rest of the infrastructure kept pace.
Those lessons helped inform repeatable architectures for NVIDIA DGX SuperPOD, and DDN became its first certified storage solution. Capabilities that historically required the expertise and resources to build bespoke supercomputers were becoming a foundation for organizations to deploy, scale, and continue building on as their AI requirements evolved.
From Supercomputing to AI Infrastructure
The work continued with NVIDIA’s Cambridge-1 supercomputer and later Eos, which combined 576 NVIDIA DGX H100 systems with DDN AI400X2 storage. Each generation exposed DDN’s teams to new levels of accelerated computing and more demanding AI workloads.
The collaboration also expanded beyond individual systems. Lessons learned in these environments informed reference architectures designed to make high-performance AI infrastructure easier to deploy at scale. Each generation built on the one before it, allowing DDN to evolve alongside increasingly demanding AI workloads.
DDN evolved with those demands. The AI400X line advanced to AI400X2 and AI400X3, and the work moved from training infrastructure into production AI environments. Today, it spans the NVIDIA Vera Rubin platform, NVIDIA Cloud Partners, and emerging areas such as distributed key-value (KV) cache acceleration with NVIDIA Dynamo. The NVIDIA Vera Rubin platform is a POD-scale platform designed for agentic AI, unifying multiple rack-scale systems into a single AI supercomputer to fuel token generation.
DDN brought its supercomputing heritage to AI. A decade solving AI infrastructure problems, many alongside NVIDIA, helped shape the company in return.
The Next Problem to Solve
10 years after DGX-1, the infrastructure challenge has changed again.
Accelerated computing has established its role in AI. The challenge now is deploying AI at exabyte scale while improving the economics, efficiency, and sustainability of the infrastructure behind it. Keeping GPUs productive remains fundamental, but AI infrastructure must move growing volumes of data without allowing cost, power consumption, and complexity to grow at the same rate. and complexity to grow at the same rate.
AI factories need to produce more useful work from every GPU, watt, rack, and dollar invested while remaining adaptable to a technology landscape changing quickly. These are data challenges as much as compute challenges, shaping infrastructure designed to sustain performance as AI scales while improving the economics of that growth. DDN continues to work on many of them alongside NVIDIA.
The technologies, workloads and scale have changed, but a consistent thread runs through the company’s evolution: as advances in compute create new demands on the data layer, DDN applies what it knows to the next infrastructure problem.
10 Years of DGX and What Comes Next
The 10th anniversary of DGX is a milestone for NVIDIA and for the accelerated computing industry it helped create.
DGX helped turn accelerated computing into infrastructure organizations could deploy, repeat, and scale. Over the decade that followed, systems grew larger, models became more sophisticated, and an extensive ecosystem developed around them. What began with DGX-1 helped lay the foundation for the AI factories being built around the world today.
DDN’s work with NVIDIA contributed significantly to that evolution. NVIDIA gave DDN engineers early exposure to new infrastructure challenges and pushed the pace at which those challenges emerged. DDN brought the data expertise required to help increasingly powerful GPU environments perform efficiently and scale.
This anniversary celebrates an NVIDIA milestone and a partnership that helped shape DDN. It also illustrates something fundamental about DDN: its history has been defined less by any single generation of technology than by a willingness to keep solving the infrastructure problems created by what comes next.
Congratulations to our partners at NVIDIA on 10 years of DGX. Innovative companies find each other, and we are proud of what we have accomplished together.