After 2021, NVIDIA has been among the top 1~2 companies where American college graduates most want to work. In the past, the best engineers would flock to Apple or Google, but now the company that top engineers most want to join is NVIDIA. Through scholarship programs and industry-university cooperation projects with well-known universities such as Stanford University and MIT, NVIDIA has also ensured the replenishment of top research manpower in the academic world.
Moreover, NVIDIA has DevTech (computing expert team), so that these researches are not just theoretical, but can also apply the research results.
After all, most scientific and technological researchers are busy publishing technologies and theories as papers, but rarely have the ability to transfer content to business departments (because they focus on research and have no time to commercialize research), but NVIDIA's "Research→Optimization→ Commercialization→Marketing", the whole process is very complete, from decision-making to execution can be completed in a short time.
CUDA must be mentioned
In the process of developing a gaming graphics card equipped with a GPU in 2006, NVIDIA realized that a GPU that can efficiently perform parallel operations can be used for a variety of purposes, so it invested US$ 10 billion to develop CUDA. As a result, developers in various fields such as AI, autonomous driving, and smart factories that need to process large amounts of data have begun to use CUDA.
Since CUDA and NVIDIA's GPU are bound to each other, the more widely used CUDA is, the higher the sales of GPU will be. CUDA has actually become an industry standard. When CUDA was launched, it published a manual for developers and created an active community, using marketing engineers to promote and support customers. Some major computer science departments have included CUDA in required courses.
Over the past 20 years, many elites in the field have been using CUDA, and it is not easy to break CUDA's monopoly. In Korea, if you have the opportunity to talk to engineers, you will find that their reactions are similar. They said that if they want to use AMD's GPU, they must first adjust the software originally applied to CUDA, which gives them a huge headache.
In South Korea, AMD is considered the "second choice" for GPUs, but in fact, the gap between NVIDIA and AMD is very large. In terms of market value, NVIDIA is about US$ 3 trillion, while AMD's market value is US$ 262.1 billion. The former is 11.4 times the latter. Therefore, some engineers even directly suggested that the company should not use AMD products and wait until NVIDIA's new chips come out.
When engineers say this, it means that they have a high degree of loyalty to NVIDIA, and it also shows that there are currently no competitors that can challenge the strong moat of CUDA.
NVIDIA's position will be difficult to shake in the short term
Jensen Huang said at the 2024 SIEPR (Stanford Institute for Economic Policy Research) Economic Summit: "Even if competitors provide chips for free, they cannot beat NVIDIA. Even if NVIDIA's GPUs are more expensive than competitors, from the perspective of data center infrastructure From an operational perspective such as construction costs and management costs, it is difficult for competitors' product performance to catch up with NVIDIA's competitiveness." This sentence appropriately expresses NVIDIA's current position in the AI ecosystem.
NVIDIA is currently trying to build AI infrastructure in Southeast Asia. And it is not limited to the field of AI, they are also expanding their business to other industries, including robotics, autonomous driving, logistics and even biotechnology, not only cooperating with various enterprises, but also cooperating with government agencies. NVIDIA's future changes and development are worth looking forward to.
