From Paris Eccentric to AI Billionaire: Inside DDN's Data Storage Boom

How Alex Bouzari's DDN Quietly Powered the AI Revolution

Alex Bouzari does not look like a typical tech CEO. The 65-year-old Franco-Iranian founder of data storage firm DDN designs his own clothing — imagine Regency-era suits, towering collars, and crocodile-skin shoes — and once shared a shirtmaker with Karl Lagerfeld. Yet behind the eccentricity is a relentless business logic that has quietly turned a once-niche supercomputing supplier into a billion-dollar engine of the AI revolution.

DDN was founded in 1998 to solve a problem that has only become more acute: feeding data fast enough to the world’s most powerful computers. Its software slices enormous files into pieces and distributes them across thousands of disks in parallel, keeping GPU-filled supercomputers from idling. The approach was vital for national labs and physicists, but the customer base was tiny. That changed when Nvidia began building its own GPU supercomputers for AI and discovered that conventional storage could not keep up. “The day we turned on the supercomputer, the storage system failed,” recalls Nvidia vice president Marc Hamilton. DDN’s technology became essential to Nvidia’s AI infrastructure.

Even with that win, growth was slow until the AI explosion. Revenues hit roughly $400 million in 2024, $500 million in 2025, and are on track to double to $1 billion in 2026, propelled by deals with Elon Musk’s xAI and national governments building sovereign AI machines. In early 2025, private-equity giant Blackstone bought a $300 million stake that valued DDN at $5 billion, instantly making Bouzari and co-founder Paul Bloch billionaires on paper — each holds about 40%.

The real commercial breakthrough, however, came when DDN began selling its software independently of its hardware racks in 2023. The software-only business carries gross margins above 90%, compared with 60% for hardware. The shift arrived as competitors like Vast Data, Weka and Hammerspace scrambled for the same AI storage market. DDN is now planning an even bolder move: a 1.3-gigawatt data center in Wheeler, Texas, that could cost more than $60 billion, aiming to offer not just storage but the entire infrastructure AI workloads demand.

The Business Logic Behind DDN's Software Pivot and Blackstone's Bet

Why the Software Pivot Was the Real Turning Point

For years, DDN insisted on bundling its file-system software with its own hardware, limiting its addressable market. By unbundling in 2023, the company could sell to customers who wanted to use their own commodity disks while keeping DDN’s parallel-file-system intelligence. The result was a surge in software-only deals and a jump in profitability — gross margins now exceed 90% on the software side. This mirrors the path taken by the broader storage industry, where value has shifted from spinning disks to the data-management layer.

Blackstone’s $300 Million Key to an Empire of Data Centers

Blackstone’s investment is about more than capital; it gives DDN privileged access to the private-equity firm’s $325 billion portfolio of data center assets and clients. John Watson, a senior executive at Blackstone, described DDN as “a business that hit puberty in its late twenties.” With that network, DDN can embed its software into colocation and cloud facilities, locking in recurring revenue and expanding its footprint far beyond the laboratory supercomputers that originally defined the company.

Vast Data and a Crowded Competitive Field

DDN’s biggest rival is New York-based Vast Data, which raised $1 billion at a staggering $30 billion valuation in April 2026 and counts CoreWeave, Nebius and Mistral among its clients. Other challengers like Weka and Hammerspace are also winning business. The competitive gap has narrowed because customers no longer need to buy a complete proprietary stack; they can mix and match. DDN’s advantage remains its long track record of extreme parallel throughput — the very thing that caught Nvidia’s attention — but sustaining margins against well-funded rivals will require continued innovation and a fast-scaling data center offering.

The Texas Gigawatt Gamble

DDN’s plan to build a 1.3 GW data center in Wheeler, Texas, signals an ambition to move from component supplier to full-stack infrastructure provider. The price tag, potentially north of $60 billion, dwarfs DDN’s current balance sheet and will almost certainly require significant external project financing. Success would transform DDN into a vertically integrated operator serving the largest AI clusters; failure could distract management and strain its core storage business at a moment when competitors are flush with cash.

What DDN's Rise Means for AI Infrastructure Buyers and Rivals

• For AI infrastructure buyers: Evaluate whether software-defined storage from DDN, Vast, or Weka can improve GPU utilization. DDN’s unbundled model shows that 90%-plus software margins are achievable without proprietary hardware lock-in, making performance comparisons directly against hardware-plus-software bundles essential before large-scale purchases.
• For competitors: Vast Data’s $30 billion valuation and CoreWeave partnership have set a high bar; DDN’s response should be watched. The Texas data center project, if executed, could bundle storage with compute and power, reshaping the competitive dynamic beyond the software layer alone.
• For investors watching the AI infrastructure space: DDN’s revenue trajectory from $400 million to a projected $1 billion within two years underscores the blistering pace of AI hardware demand. The Blackstone stake values the company at 5x forward revenue, a benchmark that could influence future funding rounds or a potential IPO.
• For data center operators and colocation providers: Integrating high-throughput parallel storage into AI-optimized facilities is becoming a differentiator. DDN’s software could be a bolt-on to existing footprints, but the imminent competition from standalone appliance vendors and storage-as-a-service models means operators must define clear performance SLAs tied to GPU throughput, not just capacity.