Tech Disruptors: DDN CEO Bouzari on Solving AI Data Bottlenecks

The article discusses how AI infrastructure demand is shifting the focus beyond GPUs to the data layer, emphasizing the need to move, manage, protect and deliver data at AI speeds. It features Bloomberg Intelligence analyst Woo Jin Ho in conversation with DDN CEO Alex Bouzari about the company’s role in AI data infrastructure. The piece is informational and contains no earnings, guidance, or quantified financial updates.

Analysis

The marginal bottleneck in AI is shifting from FLOPs to data throughput, which changes where pricing power accrues. That is structurally bullish for infrastructure layers that sit closest to the workload scheduler: high-performance storage, networking, and data management vendors should see stickier attach rates as customers are forced to overbuild for latency and checkpointing resilience. The second-order effect is that GPU suppliers lose some monopoly-like leverage over the AI stack as buyers increasingly optimize for system-level utilization, not just accelerator count.

The key winner set is likely broader than the obvious storage names: OEMs and integrators that can bundle compute, interconnect, and data movement into a single procurement decision should gain share versus point solutions. Conversely, commodity storage arrays and generic enterprise software vendors risk margin compression if AI buyers migrate spend toward purpose-built architectures with performance SLAs. Expect procurement cycles to lengthen but contract values to rise, because customers will pay up only after failing to hit model training and inference efficiency targets in production.

This theme is underappreciated as a capex reallocation story, not just an incremental TAM expansion. If GPU capex slows even modestly over the next 6-12 months, vendors exposed to the data path may still compound because their revenue is tied to cluster utilization and data gravity rather than the next accelerator refresh. The contrarian risk is that hyperscalers and large model builders continue internalizing these capabilities, which would cap third-party monetization and push value back to in-house platforms over a 2-3 year horizon.

Near term, the cleanest catalyst is earnings commentary from AI-infrastructure suppliers: any mention of tighter attach rates, larger deal sizes, or elevated gross margins from AI workloads would confirm the trade. The reversal trigger is evidence that customers are solving the bottleneck with software optimization instead of hardware spend, or that AI utilization rates plateau before new data-layer capacity gets deployed.

Key Decisions for Investors

• Go long DDN-adjacent AI infrastructure beneficiaries through public comps or suppliers with similar exposure; use a 3-6 month horizon and look for 15-25% upside if management teams confirm rising attach rates and deal sizes.
• Pair long AI data-path infrastructure / short a legacy enterprise storage basket (e.g., long ANET or NTAP-adjacent winners vs short slower-growth storage names) to isolate the shift from generic storage to AI-optimized architectures; target a 1-2 quarter catalyst window.
• Add to semis/networking exposure only where names monetize cluster efficiency, not just accelerator shipments; prefer a basket tilt toward networking and interconnect over pure GPU beta for better risk/reward if AI capex broadens.
• If hyperscaler capex commentary softens, fade the most expensive AI hardware beta and rotate into data-layer enablers; that trade has better downside protection because it benefits from utilization, not just buildout pace.