NVIDIA chief executive Jensen Huang says ASICs cannot catch GPUs, arguing that attempting to chase AI advances with rigid silicon is illogical. In a Taipei interview, he dismissed the notion that ASICs could supplant GPUs as a fundamental threat, while revealing a soaring R&D budget of $45 billion to keep NVIDIA at the forefront of chip architecture, interconnects, and the CUDA software ecosystem.
According to Digitimes, Huang contends that the rapid evolution of AI algorithms makes specialised hardware obsolete too quickly for an ASIC’s lifecycle. The article notes that as large cloud providers pursue ASICs such as Google’s TPU, Microsoft’s Maia, and Meta’s MTIA to cut costs, NVIDIA maintains a saturation-style investment strategy to preserve its universal advantage.
It also contrasts the programmability of GPUs with the rigidity of ASICs, arguing that NVIDIA’s multi-layered approach ensures competitors struggle to match its aggregate performance and development velocity.