ON OCTOBER 2ND, ACCORDING TO CNBCMicrosoftChief Technical Officer Kevin Scott said on Wednesday, local time, that Microsoft's future plan was to use self-study in its data centreschipI don't know. It's about to take down MicrosoftNvidia,AMD We'll wait for mainstream chip manufacturers。
SEMICONDUCTORS, AS WELL AS SERVERS LOCATED INSIDE DATA CENTRES, ARE CENTRAL TO THE DEVELOPMENT OF ARTIFICIAL INTELLIGENCE MODELS AND APPLICATIONS. AT PRESENT, THE TECHNOLOGY GIANT, YOUNG WEIDAR, IS DOMINANT IN THIS AREA WITH ITS GRAPHIC PROCESSOR (GPU), WITH A RELATIVELY SMALL MARKET SHARE OF THE COMPETITION AMD。
However, major cloud computing manufacturers, including Microsoft, have also started to design customized chips for data centres。
DURING ITALIAN SCIENCE WEEK, SCOTT DESCRIBED THE COMPANY ' S STRATEGIC CONFIGURATION IN THE AREA OF ARTIFICIAL INTELLIGENCE CHIPS IN A FIRESIDE CONVERSATION HOSTED BY CNBC。
CURRENTLY, MICROSOFT DATA CENTRES ARE MAINLY BASED ON BRITISH WIDA AND AMD CHIPS, THE CORE CONSIDERATION OF WHICH IS TO SELECT THE MOST PERFORMING CHIP TO ACHIEVE THE SINGLE CHIP “BEST VALUE FOR MONEY”。
Scott says, "We don't have to choose the chip for a specific brand. For many years, the British Wida chip has been the most expensive solution. In order to ensure that there is sufficient capacity to meet demand, we will consider actively any viable options.”
At the same time, Microsoft has started to use some self-study chips。
In 2023, Microsoft introduced the Azure Maia artificial intelligence accelerator, designed for artificial intelligence loads, and the Cobalt Central Processor (CPU). In addition, there are reports that the company is developing the next generation of semiconductor products. Last week, this American technology giant also released a new cooling technology based on “microflow” to address chip overheating。
Asked whether Microsoft had a long-term plan to make self-research chips the backbone of its data centre, Scott said, “Yes, of course” and added that self-research chips were currently being used extensively。
Scott noted that the focus on chips was an important part of Microsoft ' s strategy to build a full data centre system. He said: “The key is in the design of the entire system, covering the network and cooling systems. We need an autonomous decision-making space in order to optimize our ability to calculate in accordance with the workload needs.”
MICROSOFT, ALONG WITH ITS COMPETITORS GOOGLE AND AMAZON, IS PROMOTING SELF-RESEARCH CHIPS, NOT ONLY TO REDUCE RELIANCE ON WEEDA AND AMD, BUT ALSO TO MAKE PRODUCTS MORE RELEVANT TO THEIR SPECIFIC NEEDS AND MORE EFFICIENT。
Science and technology giants such as Meta, Amazon, Alphabet and Microsoft have pledged to invest more than $30 billion this year (note: the current exchange rate is about 2.14 trillion yuan) in capital spending. Most of this money will be invested in artificial intelligence to meet the rapidly growing AI needs。
Scott pointed out that the current deficit remained critical. He said: “It is likely that the current calculator situation is still conservative to describe it as a `serious shortage'. Since the launch of ChatGPT, we have been faced with the difficulty of keeping pace with demand at the pace of capacity-building.”
The Chief Technical Officer warned that although Microsoft had been improving its ability to calculate through the expansion of the data centre, it was still unable to meet market demand. Scott said, “Even if we make the most optimistic predictions, the end result will often fall short of demand. Over the past year, we have deployed a significant amount of computing resources, and we will be further investing in the coming years.”