In the future, Microsoft wants that in their data centers there have been integrated manufacturing and design chipsas assured by his CTO, Kevin Scott, in an event held in Italy a few days ago. With this step, according to CNBC, the company wants to show its intention to reduce the dependence of external companies. Above all, of great technological ones.
Microsoft has already designed its own custom chips, specifically for data centers. So last Wednesday, Kevin Scott offered several details about Redmond’s strategy with respect to chips for AI in a talk during the Italian Techweek.
Until now, Microsoft, as we have mentioned, has used AMD and NVIDIA chips in their data centers. The main criterion for its selection was to use the semiconductor that offered the best relationship between performance and Price. Scott has assured that they are not faithful only to a manufacturer, and that they have been mentioned to choose them for many years. He also stressed that the chips that have offered the best relationship between performance and price for years have been those of Nvidia.
But while I went ahead with this chips selection policy, Microsoft has begun to use some of its chips. In 2023, the company launched its Azure Maia AI accelerator, designed for AI workloads, as well as the CPU Cobalt. In addition, the company is working in next -generation own semiconductor lines. Only a few days ago the company has made public a new cooling technology that uses microfluids to solve the usual problems with the chips that overheat.
The amount of own chips that Microsoft is using in its data centers has been increasing progressively, and now, according to Scott, the company already uses a large number of own chips. This commitment to design and manufacturing semiconductors is part of a strategy to possibly design a complete system for data centers, which goes from networks to refrigeration, for «have the freedom to make the necessary decisions to really optimize computing for workload«.
In addition to Microsoft, other great technological ones, such as Google and Amazon, are designing chips to reduce their third party dependence and to make their products more efficient for the fulfillment of the requirements and needs of these companies.
On the other hand, despite the millionaire investments in infrastructure to be able to meet the needs caused by AI, Scott has recognized that there is still a shortage of computing capacityand that technological ones carry From the appearance of chatgpt without being able to develop this capacity quickly enough to meet the demand. Despite having already built several data centers for AI, they have not yet managed to reach the level they need. But neither Microsoft nor the rest of great technological ones.
