GPU large manufacturer NVIDIA (NVIDIA) recently demonstrated its Blackwell Ultra AI server's outstanding ability in investment rate (ROI), which means that its investment in funds to buy NVIDIA's solution is more financially beneficial, and has certified NVIDIA executive president Huang Rensheng's famous saying "The more you buy, the more you save."
NVIDIA in the AI market claims that its computing power is not only in terms of performance, but also in terms of revenue that can be generated. The famous saying of the executive chief Huang Rensheng in the past, "The more you buy, the more you save", is to emphasize the return on investment that can be brought by using NVIDIAAI servers. Recently, Ian Buck, Vice President of NVIDIA's ultra-large and high-performance computing business, revealed the investment return status of the GB200 NVL72 AI server in a theme speech, and the results showed that NVIDIA's products can produce significantly higher investment return efficiency.
Based on the projection videos displayed by NVIDIA, the company generously shows off the financial efficiency of its AI system, saying that investing US$3 million in a GB200 NVL72 server rack can generate a revenue of up to US$30 million through AI token inference, which represents a 10-fold increase in investment rate. NVIDIA calls this phenomenon "AI Factory ROI". In contrast, the revenue generated by other AI chips is very meager, which clearly shows that for large cloud service providers, adopting NVIDIA systems is the correct choice.
NVIDIA emphasizes that its hardware can not only execute AI recommendation tasks on a large scale, but it is obviously a profit multiplier. Therefore, this also explains why NVIDIA charges such high prices for its rack-level solutions.
Although NVIDIA has not yet disclosed the specific indicators behind this acquisition analysis, it can be reasonably deduced that other AI chips referred to here may be far behind in terms of power efficiency, which is also one of the reasons for their lower revenue. More importantly, with NVIDIA's shifting development focus from data centers to so-called AI factories, it is important for modern systems to have an impressive data on efficiency-per-dollar to ensure that the AI development wave can continue and stably advance.