Graphics cards as the basis for AI technology - Why Nvidia in particular benefits so much

Nobody could have imagined that an Nvidia supercomputer would one day run the most famous AI program - ChatGPT - in the year it was founded in 1993. After Nvidia revolutionized the gaming sector in 1999 with the invention of the graphics card, this sector was initially the main focus of the company. With the start of Bitcoin mining, the first big run on graphics cards began. Especially during the height of the crypto boom, graphics cards (GPU) became hard to come by and if it was possible at all, it was only at very high prices. The next big boost in sales came when it became apparent that graphics cards were much better suited to computationally intensive processes such as machine learning and artificial intelligence than traditional processors (CPU). Nvidia in particular is benefiting immensely from this development and this is also reflected in its record-breaking quarterly figures. With a turnover of around 13.5 billion US dollars, this has increased by 88 percent compared to the previous quarter. Compared to the previous year, it was even 101 percent. The majority of this revenue is attributable to the Data Center segment. This includes solutions for corporate data centers. Specific applications include AI, data analysis, high-performance computing and rendering. Another part of the revenue is generated in the gaming division. Here, there was a 22 percent increase in sales compared to the previous year. In concrete figures, this division brought in just under 2.5 billion US dollars. Looking to the future, the company is once again optimistic. For the third quarter, Nvidia issued a forecast of around $16 billion for total revenue.

According to Business Insider, Nvidia currently has immense market power in the AI chip market with a market share of over 80 percent. Such strong pricing power also explains the gross margin of around 70 percent in the second quarter. This becomes even clearer when you consider that, as Business Insider reports with reference to Srini Pajjuri, the estimated production cost for the high-end AI model (H100) is just over $3,000 and the product sells for around $25,000. However, Advanced Micro Devices (AMD) is a competitor waiting in the starting blocks, which has already shown in the CPU market that it can knock top dogs like Intel off the throne.

AI software requires an enormous amount of computing power. This is mainly due to the fact that a large amount of data is needed to train an AI. The AI tries to recognize patterns and regularities in order to draw logical conclusions. To do this, you first need training data, which is elementary for the later success of the artificial intelligence. Humans evaluate the data to show the AI what the correct answer would be in each case. If this step was successful, the AI is fed with larger amounts of data. To evaluate these mountains of data and further optimize the AI, the AI needs high computing power, which is why graphics cards are best suited.

Due to the ever-increasing implementation of AI in business processes and private life, the coming period will be exceedingly exciting for chip companies like Nvidia, which provide the infrastructure for this technology. Only the future will show whether Nvidia can prevail against competitors like AMD and Co. in the long run and thus retain its strong market position.