According to a recent social media post Barron’s senior writer Tae Kim, Nvidia is making massive profits from the sale of its H100 GPU accelerators. Kim estimates that the company is generating profits of up to 1,000% for each H100 sold. Even with a street-price of $25,000 to $30,000 per accelerator, the cost per chip and peripheral components is estimated to be around $3,320. This means that Nvidia is comfortably covering its production costs and generating substantial profits.
It is important to note that the estimated cost per chip does not include factors such as research and development (R&D) expenses and salaries of engineers involved in product development. Nvidia’s average salary for an electronics hardware engineer is around $202,000 per year. Considering the number of specialized workers and hours required for the development of chips like the H100, these additional costs need to be taken into account.
Despite the expenses involved, Nvidia is experiencing a surge in demand for its AI-accelerating products. Orders for these products are reported to be sold out until 2024. With the AI accelerator market projected to be worth $150 billion 2027, Nvidia is well-positioned to benefit from this growth.
However, the boom in AI acceleration comes with consequences for other sectors. The increasing interest in AI servers has led DDR5 manufacturers to revise their expectations on the adoption of new memory products. It is now predicted that DDR5 will only reach parity with DDR4 the third quarter of 2024.
While Nvidia’s profits are soaring, investing in AI acceleration at current prices may limit opportunities for other investments or research and development pursuits. Budget constraints and the opportunity cost of focusing on AI acceleration could impact the choices of some players in the industry.
In conclusion, Nvidia is reaping the benefits of its early investments in AI and its dominance in the AI accelerator market. With strong sales and projected growth, the company’s bottom line is thriving.