NVIDIA’s ‘Cutting-Edge’ Vera Rubin AI Servers Are Already in Development by Foxconn
NVIDIA’s next-generation Vera Rubin AI server lineup has officially entered the development phase, with long-time manufacturing partner Foxconn taking the lead, according to a report from Taiwan Economic Daily. The servers, designated as the Vera Rubin NVL144 MGX, are already under construction, with mass production expected to begin in the second half of 2026.
Named after the pioneering astrophysicist Vera Rubin, this upcoming server platform represents NVIDIA’s most ambitious AI infrastructure project to date, integrating new-generation Rubin AI processors designed to deliver unprecedented computing power for data centers, generative AI, and large-scale model training. The Rubin architecture is expected to push the limits of both performance-per-watt efficiency and scalability, positioning NVIDIA to maintain its dominance in the AI hardware market as demand continues to surge.
For now, the company’s GB200 Blackwell AI servers remain the primary revenue driver for NVIDIA and its manufacturing partners, including Foxconn. These systems are already in mass production and will continue to ship throughout late 2025 and the first half of 2026. The Blackwell Ultra variant, which offers improved throughput and enhanced interconnect performance, is expected to lead shipment volumes during this period.
The anticipated release window of the Vera Rubin lineup, roughly six to eight months after the GB300 ramp-up, highlights NVIDIA’s aggressive roadmap cadence. This tight schedule reflects CEO Jensen Huang’s strategy of continuous innovation cycles to stay ahead of the competition and sustain NVIDIA’s leadership in AI computing.
Foxconn’s involvement further reinforces its growing role in the AI hardware ecosystem. According to market estimates, Foxconn now accounts for nearly 60% of NVIDIA’s AI server manufacturing, underscoring its critical partnership with Team Green. The Taiwanese manufacturing giant has also been expanding its global footprint by establishing US-based production facilities, aligning closely with NVIDIA’s long-term vision of localized manufacturing to support American and European data center infrastructure.
This collaboration also signifies a broader shift in the semiconductor and AI supply chain landscape. With NVIDIA’s Blackwell series already dominating hyperscale data center deployments, the arrival of the Vera Rubin architecture is poised to drive a new wave of adoption across the AI industry. Early reports suggest that major players, including OpenAI, have already expressed interest, with multi-gigawatt-scale (multi-GW) deals reportedly in discussion for early adoption of Rubin-based systems.
If current timelines hold, NVIDIA’s Vera Rubin AI servers could reach mainstream availability as early as late 2026 or early 2027, coinciding with the next major upgrade cycle for hyperscale data centers. With massive improvements expected in both energy efficiency and compute density, Rubin may become the defining platform for next-generation large language models, simulation workloads, and advanced generative AI systems.
NVIDIA’s rapid iteration between architectures, Ampere, Hopper, Blackwell, and soon Rubin, demonstrates the company’s determination to lead not only in AI performance but also in hardware ecosystem integration, ensuring end-to-end control from GPU design to large-scale deployment. As Foxconn continues to expand production capacity, the partnership between these two giants will likely set the tone for the next era of AI infrastructure manufacturing.
Do you think NVIDIA’s aggressive roadmap will help it maintain its dominance, or will competition from AMD and Intel catch up by 2026? Share your perspective below.
