NVIDIA’s next technological milestone is already taking shape. The company’s upcoming Vera Rubin AI server platform – named after the pioneering astronomer – is now under development by long-time partner Foxconn, with full-scale production expected to begin in the second half of 2026. 
This follows the rapid ramp-up of the Blackwell Ultra GB300 series, showing just how aggressively NVIDIA is pushing its AI roadmap.
According to Taiwan’s Economic Daily, Foxconn has begun engineering work on the Vera Rubin NVL 144 MGX servers, signaling that the design phase is well underway. These servers will mark a major leap in AI computing, integrating NVIDIA’s most advanced Rubin architecture chips. Unlike minor refreshes of the past, Vera Rubin is expected to redefine AI infrastructure from rack layout to chip design, enabling greater power efficiency, faster interconnects, and more scalable compute clusters optimized for hyperscale data centers.
Currently, NVIDIA and its ecosystem partners are seeing record profits from GB200 Blackwell AI servers. The upcoming Blackwell Ultra refresh remains the main driver through early 2026, with suppliers already in mass production. Vera Rubin’s release timeline – just six to eight months after the GB300 expansion – illustrates NVIDIA’s remarkably short innovation cycle, an almost yearly cadence that leaves rivals scrambling to keep pace.
Foxconn’s role is particularly crucial. The Taiwanese manufacturing giant reportedly holds around 60% market share of NVIDIA’s next-gen AI server builds. Beyond engineering, Foxconn is investing heavily in US-based manufacturing facilities, aligning with NVIDIA’s goal of strengthening domestic production capacity amid global supply chain shifts. This could represent a subtle but meaningful relocation of AI hardware production from Asia to North America, a move with both political and logistical implications.
Industry insiders suggest the Vera Rubin systems have already drawn early interest from OpenAI and other major AI labs, with preliminary multi-gigawatt deployment deals reportedly in discussion. That scale of infrastructure investment underscores the hunger for cutting-edge compute – and NVIDIA’s dominance in providing it. As 2026 approaches, the tech world will be watching closely to see whether Team Green can maintain its momentum or if competitors like AMD’s MI450 series will finally manage to challenge NVIDIA’s reign.
Either way, with Rubin’s rumored N3P process and unprecedented performance targets, the next AI arms race has already begun.
3 comments
AMD won already, Rubin on N3P won’t save them this time 💅
DOA trash, calling it now 🤷
Whole AI chip market is starting to look like a shovel sale before the crash 😬