In some areas of the semiconductor industry, growth has become somewhat unpredictable, triggering a wave of industry consolidation. Last week, Broadcom, a networked technology company, proposed to buy Qualcomm, a leader in smart phone chips, for a whopping $ 105 billion. If the deal finally arrives, it will be by far the largest deal in the industry.
IDC estimates that global spending on AI-related hardware and software may increase from $ 12 billion this year to $ 57.6 billion by 2021. The research institute estimates that a significant portion of the spending will flow into the data center and by 2020 the data center estimates that a quarter of capacity will be spent on AI-related computing.
In recent years, specific AI technologies have become an integral part of the products and services of various major technology companies. For example, with this type of technology,AmazonEcho smart speakers can understand the voice commands; Nest parental control camera from Google's parent company Alphabet can distinguish between acquaintances and strangers, which in turn can give users a warning alert; Facebook also be able to social media posts with the most likely to cause the poster Interest ads match.
Today's largest technology companies - Google, Facebook, Amazon, IBM,Microsoft, And China's Internet giants - have each configured their own data center with specialized hardware to speed up AI software training, such as document translation capabilities for such software.
Internet giants are using AI technology, known as deep learning, to enable software to distribute digital documents such as images, sound recordings and documents. Such programs can take some time to find meaningful patterns in training data. Internet giants want to upgrade their algorithms so they do not have to wait for a few weeks to know if training is working.
Chip makers are trying to help them speed up the work.
Since its founding 24 years ago, NVIDIA has spent most of its time developing high-end graphics chips for personal computers. In recent years, its products have proven to be faster than traditional processors in training AI software.
NVIDIA today reported third-quarter earnings, revenue and profit during the period exceeded Wall Street's expectations. In the past 12 months, its data center business sales almost tripled to about 1.6 billion U.S. dollars. Driven by this fast-growing business, the company's stock has soared nearly 7-fold, closing at US $ 205.32 as of Thursday's U.S. time.
AI chip startups Mythic CEO Mike Henry (Mike Henry) pointed out that in recent years around NVIDIA's optimism that "everyone in the industry to be able to feel." he said, "the explosive growth of interest in helping Mythic received $15 million in venture capital investment, including the well-known Silicon Valley venture capital firm DFJ.
According to PitchBook Data, a venture capital database this year, private equity investors almost doubled their share of AI hardware to $ 252 million.
Nvidia's main rival did not rest on its laurels. Last year, Intel brought Nervana Systems back under the door and the transaction amount was not announced. The chip giant is teamed up with Facebook and other companies to build Nervana-based chips by the end of the year to outperform NVIDIA in terms of AI computing power.
With the exception of Nervana, Intel places the highest priority on AI performance across its entire data center offerings. Intel announced last month's third-quarter earnings report,serverRevenue from the processor and programmable chip business increased by 7% and 10% respectively year-on-year.
"We will not divulge our total investment in these businesses, but it's a big project as important as we are leading the way in traditional processing technologies," said Naveen Rao, who leads Intel's AI division.
AMD, which competes with NVIDIA in gaming systems, recently introduced an AI graphics processor called the Radeon Instinct. Baidu is the company's customer, AMD also disclosed in the recent earnings conference callcloud serviceThe provider is also its client.
Some companies are not waiting for large chip makers to make excellent products. For example, Google opted to design its own proprietary AI accelerator and seek to gain a competitive edge by tailoring chips to its software.
"This field has just started," NVIDIA accelerated computing business director Ian Barker (Ian Buck) said, "it feels every quarter in creating and re creating"