Nvidia was close to developing an x86 CPU to compete with AMD and Intel, but had to switch to Arm due to 'certain legal issues'.

During a technical session at the SC24 event, Dave Ditzel, founder of Esperanto Technologies, shared interesting insights about Nvidia's early server processor initiatives. According to his statements, Nvidia initially launched Project Denver as an x86 architecture CPU but had to adapt it to Arm due to legal constraints.

Ditzel, who was previously the CEO of Transmeta, explained that the shift to Arm was influenced by the licensing of Transmeta's Tokamak technology, which allowed for the conversion of x86 code to a RISC instruction set. In his words, "Nvidia presented a product called Denver, which originally started as an x86 CPU, but through certain legal issues, it had to become an Arm CPU." This decision was crucial for Nvidia to align with the Arm architecture.

Tokamak, the technology developed by Transmeta, was envisioned as a third x86 chip, following the Crusoe and Efficeon processors. However, this project was never officially launched, and the technology was licensed to companies such as Intel and Nvidia. Although Intel acquired the design, it also did not release a product based on it. Ditzel humorously noted that one can imagine the reasons behind this.

Nvidia officially launched Project Denver as an Arm-based CPU in 2011 and later integrated it into its Tegra product line. While there was significant enthusiasm for Arm-based servers initially, adoption was hindered by challenges in the software ecosystem. Since then, Nvidia has developed its Grace CPU and abandoned its ambitions to acquire Arm following regulatory opposition.

Ditzel, who founded the chip design company Esperanto about seven years ago, chose to work with RISC-V, as it offered a more cost-effective alternative without the legal issues associated with x86. According to him, "at least we have a testing area where we can experiment without some lawyer constantly supervising us."