Nvidia's CUDA no longer has protective moat thanks to DeepSeek

Nvidia's once-unassailable competitive edge—CUDA—is losing its dominance thanks to DeepSeek. While some see this as a groundbreaking change, others believe it marks the commoditization of AI technologies, leveling the playing field for AI’s newest developments. 

We spoke with Manil Vasantha, AI Industry Expert, to understand why CUDA might be giving way to competitors. Vasantha is the CEO/COO of customer service consultancy Custx.ai and a VP at one of the major Enterprise cloud software providers.

The moat: "DeepSeek has clearly pointed out that CUDA is no longer Nvidia's moat, which no one is talking about," Vasantha says. Nvidia released their NIM DeepSeek architecture, only underscoring that further." He elaborates, "I don’t think it’s really a competition anymore. AI as a technology is evolving into a commoditized technology stack faster than we expected. The real differentiation will come from AGI. The question is—who’s going to cross the finish line first?"

Integration race: Actual integration of AI into daily life is one of the main barrier to entries that AI technologies face. Accessibility is the other key determining factor into which technologies take off. 

Pointing to a further erosion of CUDA as a leader, Vasantha says, "The reason I talk about commoditization is simple—Nvidia took 15 years to fully saturate CUDA, and now we have 4.5 to 5 million trained users. But AI is evolving at an unprecedented speed. DeepSeek models can already run efficiently on Nvidia’s GeForce RTX GPUs, making AI more accessible than ever—even for a kid with a gaming PC."

Mixture of Experts (MOE): DeepSeek’s particular MOE model also sets it apart: "OpenAI is a very strong generalist with a breadth of knowledge. It can diagnose as much as possible and then refer to the specialist, which could be another LLM." explains Vasantha. "Whereas DeepSeek is that one stop shop. You not only have the generalist, which is the first entry point into DeepSeek, you have the specialist, which is the Mixture of Experts. 

Competitive advantage: Ultimately, performance and speed may be the decisive factor for which AI emerges as the winner. "In a side-by-side comparison, DeepSeek will be much faster in my opinion. Hypothetically, DeepSeek's benchmark will be much faster because they tuned it for performance. I would come close to calling it hard-coded for performance."