SHANGHAI, June 10, 2026—As "Physical AI" and robotics take center stage in NVIDIA's strategic expansion—highlighted by its recent high-profile R&D partnerships with leading South Korean enterprises—the surging global demand for AI applications is putting computing infrastructure to the ultimate test.
Against this backdrop of rapid infrastructure transformation, Vincent Chen, Managing Director at CIC, shared his exclusive analysis with the Securities Times on NVIDIA's recent, highly discussed configuration adjustments for its next-generation Rubin NVL72 AI server clusters.
The Rubin NVL72 Pivot: A Strategic Delivery Play, Not a Downgrade
Recent industry reports reveal that NVIDIA has significantly revised the memory configuration of its upcoming Rubin NVL72 cabinets amid tight global supply chains. Total memory capacity per cabinet has been scaled back from the initially planned 55TB to 28TB, with 192GB SOCAMM modules replaced by 96GB alternatives.
While some market observers viewed this as a compromise, Vincent Chen offers a deeper, more bullish structural perspective:
"The core driver here is purely a supply shortage of SOCAMM components. NVIDIA is intentionally prioritizing speed-to-market and rapid computing deployment over rigid specifications. Crucially, they have retained 100% of the core specifications for GPUs, HBM, and interconnection modules. This is a masterful delivery strategy, not a cost-cutting exercise."
Moreover, this shift does not signal a cooling market for storage. In fact, while orders for 192GB versions have softened, demand for 96GB modules has surged, driving the total demand for SOCAMM upwards. The broader market demand for AI storage products remains fundamentally robust.
Modular Flexibility: The Technical Advantage Behind the Adjustment
Vincent Chen emphasizes that the modular architecture of the Rubin platform provides enterprise clients with a highly flexible upgrade path. Unlike HBM3e, which is packaged directly onto the GPU substrate and requires a complete chip redesign for upgrades, SOCAMM features a modular pluggable design that is physically decoupled from the CPU motherboard.
"This modularity is a game-changer for enterprise lifecycles," Vincent Chen notes. "Once the 96GB-module servers are deployed, hyperscalers and enterprises can effortlessly scale up cabinet memory capacity in the future by simply swapping modules or adding slots—zero GPU repackaging required. This adjustment is a temporary supply-chain adjustment, mirroring NVIDIA's previous successful capacity ramp-up patterns seen during the Blackwell generation."
Client Sentiments: Lowering Barriers to Entry for Hyperscalers
Will top-tier clients balk at the modified specs? Vincent Chen believes the impact is minimal. While the downsized specification inadvertently lowers the procurement threshold—potentially accelerating the market penetration of the new Vera Rubin platform among a broader tier of clients—core Cloud Service Providers (CSPs) remain fiercely committed.
The marginal performance trade-off caused by the SOCAMM adjustment will have an insulated limited impact on the content storage capabilities required for advanced Agentic AI. Rubin remains the undisputed gold standard for top-tier AI computing, and purchasing intent among global tech giants remains ironclad.
The Macro Shift: Storage as a Fundamental AI Infrastructure Asset
Beyond NVIDIA's immediate tactical moves, Vincent Chen highlights a profound paradigm shift sweeping the global storage sector: the decoupling of storage from traditional hardware cyclicality.
In the AI era, data transmission efficiency has emerged as the primary bottleneck preventing large-scale computing clusters from achieving maximum theoretical performance. Consequently, both DRAM and NAND products are shedding their identities as "cyclical commodities" and evolving into mission-critical fundamental infrastructure assets. As AI model parameters and agentic applications explode, the strategic valuation of the storage industry will only continue its upward trajectory.
Note: Portions of Vincent Chen's insights were originally featured in the Securities Times report, "NVIDIA Expands Its Physical AI Circle and Cooperates with South Korean Enterprises to Accelerate Physical AI Implementation."
About CIC
CIC is a professional consulting firm offering tailored end-to-end support across the full investment and financing lifecycle. The firm boasts a world-leading track record in guiding landmark first-in-sector IPOs across global markets, alongside unrivaled reach and in-depth coverage capabilities across specialized niche market segments.
CIC helps enterprises refine scalable business models and craft compelling capital narratives to enable seamless access to global capital markets, while serving as a trusted due diligence partner to investment institutions. It delivers granular industry insights and direct access to subject matter experts, empowering clients to identify high-value opportunities and mitigate critical risks effectively.
Media Contact
marketing@cninsights.com
