According to reports, NVIDIA's recent unveiling of NVLink Fusion has reignited industry debate over its rivalry with the emerging UALink standard. While NVIDIA remains the dominant force in AI compute interconnects thanks to its mature product ecosystem, UALink's open-standard approach is increasingly drawing attention from developers and major industry players.
Industry insiders reveal that although UALink is still in its early stages—with version 1.0 only officially released a few months ago—its first silicon tape-out is rumored to arrive by late 2025, with broader product deployments expected in 2026. Currently, there are said to be “dozens of projects” underway based on UALink, signaling its potential to challenge NVIDIA's dominance.
NVIDIA's NVLink is already deployed in commercial products and widely used across AI infrastructure. However, concerns have surfaced regarding the actual openness of NVLink Fusion, with reports suggesting it may be more closed than initially advertised. In contrast, UALink's open design is seen as more attractive by many companies seeking flexibility and long-term scalability, potentially accelerating its adoption.
In parallel, AMD's latest AI chips and integrated rack-scale systems are gaining traction. Sources indicate that major hyperscalers—including Microsoft and OpenAI—may significantly ramp up their adoption of AMD-based solutions around 2026. This shift could further boost the influence of the UALink standard and drive broader development of ASIC-based AI accelerators.
Amid these developments, speculation is growing that NVIDIA may adopt a more conservative outlook for its post-2026 operations, with potential reductions in wafer orders. Meanwhile, ASIC-focused chipmakers are revising their market forecasts upward, reflecting a possible redistribution of AI hardware momentum. Collectively, these trends suggest that NVIDIA's dominance in cloud AI infrastructure may be entering a more competitive phase, with UALink poised to expand its ecosystem footprint.