According to the latest report from Reuters on July 26th, six Amazon AWS engineers tested a new server that was closely protected in a chip laboratory in Austin, Texas. It is worth mentioning that this server is equipped with an AI chip that Amazon AWS is developing independently. Although Amazon did not disclose specific information such as the name and parameters of the chip mounted on the test server, AWS Computing and Network Vice President David Brown revealed that 'in some cases, the performance of the chip can be improved by 40% to 50% compared to Nvidia, so the cost should be half of the same type of Nvidia operation.'

The purpose of testing this server is simple. Rami Sinno, a company executive, mentioned during the laboratory visit that it was 'to compete with the market leader Nvidia's chips.' Considering Amazon's business matrix, the AI-specific chips developed by the company's AWS cloud computing service platform include Inferentia and Trainium. According to Amazon, during the recent Prime Day (Amazon's member day, the promotional day on the platform), the company deployed 80,000 AI chips for cloud computing to cope with the surge in activities on its platform. Therefore, it is obvious that as the main driving force, Amazon has a continuous demand for price reduction and efficiency improvement on its cloud computing business in terms of cost and calculation power. Since Nvidia's dominance in the AI chip field, many manufacturers have taken action to get rid of their dependence and high 'Nvidia tax'.

It is not difficult to understand that such a large investment comes from profit-driven by Amazon's AWS business. According to data, AWS’s sales accounted for nearly one-fifth of Amazon’s overall revenue in the first quarter of this year, with sales of 25 billion US dollars, an increase of 17% year-on-year. Currently, AWS occupies nearly one-third of the cloud computing market.

CFRA analysts believe that large technology companies are facing cost pressures in chips and need to develop their own chips to relieve them. In fact, for large technology companies with strong financial resources, self-developed AI chips can not only get rid of the monopoly effect of leading manufacturers such as Nvidia but also can customize hardware that is more in line with their own needs according to their business requirements.

Today's weather in Japanese. Today's weather in Japanese.

Nvidia becomes the target of public criticism.

Sinno pointed out that Amazon's customers are increasingly hoping for an alternative product that is cheaper and can handle complex calculations similar to Nvidia AI chips in terms of performance. In fact, since Nvidia won the 'throne' of the AI chip field, many manufacturers have taken action to get rid of their dependence and high 'Nvidia tax'.

Pioneers like Microsoft have been already researching a new NIC to improve the performance of the Maia AI server chip as early as February this year. The banner under which the project was flagged was 'to reduce the time and cost of training models on OpenAI servers.'

Soon, more and more vendors chose to join the self-sufficient 'go without Nvidia' camp. In April of this year, Google launched the Arm chip Axion, which supports search engine operation and AI-related tasks. In addition, Meta has also released the latest version of its self-developed AI chip MTIA and said that its goal is to reduce its dependence on companies like Nvidia.

Market research firm CFRA analysts believe that large technology companies are facing cost pressures in chips and need to develop their own chips to relieve them. In fact, for large technology companies with strong financial resources, self-developed AI chips can not only get rid of the monopoly effect of leading manufacturers such as Nvidia but also can customize hardware that is more in line with their own needs according to their business requirements.

Behind the movment of 'go without Nvidia' is the pressure on chip costs faced by large technology companies, who need to develop self-made chips to relieve them. In fact, for large technology companies with strong financial resources, self-developed AI chips can not only get rid of the monopoly effect of leading manufacturers such as Nvidia but also can customize hardware that is more in line with their own needs according to their business requirements. As other manufacturers like Broadcom move forward in their plans to become the 'second Nvidia', for example in their Q1 financial report, Broadcom expects AI revenue to reach over 10 billion US dollars in Q1 2024, of which the custom DPU chip accounts for about 7 billion US dollars.

Editor/Lambor