Home > News content

How many Chinese AI chip startups can survive?

via:雷锋网     time:2019/8/8 7:00:39     readed:325

BAT is the top three companies in China investing in artificial intelligence startups, most of which go to companies that focus on application layers or algorithms. The three technology giants are currently only making strategic investments, but the sector has recently received funding from government and venture capital. It is worth noting that Baidu and Ali have entered the field of chip design.

These chip companies are generally divided into three categories, fabless mode startups (outsourcing manufacturing), start-up companies that enter the chip design field from the algorithm field, and Internet companies that are exploring chip design.

In general, there are at least 20 new chip companies in China that meet these categories, and they are designing AI chips. More and more traditional fabless companies are also beginning to design their own AI chips, which makes the market more competitive.

The phenomenon of non-chip companies like Baidu and Ali entering the chip design field is not only in China. In the US, many listed companies, including Tesla, Google, Amazon,Microsoft, Facebook andappleAI chips are also being designed to specifically target the applications they need. If AI chip startups in China and the US face these giants with technical and economic strength, how can they survive?

This vertical integration poses a threat not only to start-ups, but also to traditional chip companies that previously viewed Internet companies as customers. This change will change both industries in the next few years.

These startups have quickly moved from burning money to architectural innovation, and now their mission is to find customers in a highly competitive market. Industry insiders say they are finding it harder to distinguish themselves, and any startup will try to bring out any unique, new or different things.

Companies that innovate and pursue emerging technologies such as analog computing, in-memory computing, and neuromorphic computing are worthy of attention, but commercializing them takes longer.

From the perspective of hardware design, these Chinese AI chips are not necessarily designed independently. Since most companies work in tight time, they tend to buy IP instead of developing it themselves. Most people connect the purchased IP together, called the on-chip interconnect. They also rely on compiler tools for chip development, not to mention high-end chips that are primarily manufactured using TSMC. Of course, a large part of the design also uses the Arm core, and some companies have developed self-developed GPUs or custom neural network cores, but most companies do not. However, VC hopes to see the return quickly!

The returns of other industries are faster. Since the typical semiconductor design process takes 18 months and requires tens of millions of dollars to support the design, it is necessary to find customers from the start.

China also has some venture capital companies that understand this feature, are experienced and more patient, and invest in chips. But it can be expected that with the pressure of venture capital, these startups will have some disappearance in the next few years. However, there will be some acquisitions, and industry leaders are expected to snap up the most promising startups. Xilinx, an FPGA company, has seen this trend when it acquired the Beijing chip unicorn Shenjian Technology last year. Other participants will use the acquisition as a means of entering the market, just as Ali acquired C-Sky (Zhongtianwei).

At this point, I think it is necessary to use dedicated AI hardware. Nowadays, most chips have some kind of AI function, and the future AI function will be enhanced. Some design teams need to decide how much effort they put into AI, but given that almost all major players are involved, it's hard to see that these features won't be a key feature of future design.

For Chinese AI chip startups, there is a need to balance the market, innovation, and patient investors. The most successful will get or get more money. On the other hand, artificial intelligence and semiconductors are two areas that the Chinese government is focusing on. The combination of these two areas will receive more attention. Still, the market is still near saturation, and only those companies that have achieved the right balance can survive.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments