Today, Google, Facebook, Microsoft, Amazon and China's Baidu and other Internet giants are exploring a variety of chip technology that can promote the development of AI, their choice will change the Intel and NVIDIA manufacturers such as the fate of chip makers. But at this point, even if the Internet giants inside the computer scientists do not know what the future will be.
Less than 12 hours, there are three people who can pay to ask me to spend an hour and talk to a stranger.
The three men said they were interested in writing an article about Google that I was doing for a new type of artificial intelligence computer chip, and they urged me to discuss the matter with one of their clients. They said the client was a manager of a large hedge fund but did not want to reveal the name of the fund.
These requests come from the so-called expert network - a research institution that associates investors with people who can help them understand specific markets and provide information on competitive advantage (which sometimes appears to be an internal message). These experts want me to explain how Google's AI processor will have an impact on the chip market. But they asked me to sign a confidentiality agreement. I refused.
These untreated, very specific and high-pressure requests were about three weeks ago, highlighting the fundamental changes that were taking place in the huge profitability of the computer chip market, which was triggered by the rise of artificial intelligence. Those hedge fund managers see these changes, but are not sure what they will lead to.
Of course, no one can be sure what results they will bring.
Today, Google, Facebook, Microsoft, Amazon and China's Baidu and other Internet giants are exploring a variety of chip technology that can drive the development of AI, and their choice will change the fate of Intel and NVIDIA's chip makers. But at this point, even if the Internet giants inside the computer scientists do not know what the future will be.
Going Deep
The online services of these companies are based on the data centers in which they are equipped with thousands of servers, each of which is driven by a chip called a central processing unit (CPU). But when they apply an AI form called a deep neural network, these companies need to add other processors as a supplement to the CPU. Neural networks can learn tasks by analyzing large amounts of data, including face and object recognition in images, translation of different languages, and so on, so only the CPU is not enough.
So, Google made the Tensor processor (Tensor Processing Unit, TPU). Microsoft is using a processor called a Field Programmable Gate Array (FPGA). They are looking for a new chip that can accelerate the AI of smartphones and other devices.
Any choice of these companies will have a great impact, because their online business is so great. They are more than any other company on Earth to buy or operate more computer hardware, and with the importance of cloud computing continues to increase, this gap will only grow. If Google chooses a kind of processor instead of another, it will fundamentally change the chip industry.
The TPU poses a threat to Intel and NVIDIA's chipmakers because the chip is made by Google itself. But the GPU on Google and other similar companies also have a huge role, and NVIDIA is the main chip manufacturers of these special. At the same time, Intel has acquired itself through the acquisition of Altera (the company sells FPGA to Microsoft). The $ 16.7 billion acquisition is Intel's biggest acquisition, highlighting the dramatic changes in the chip market.
