However, the chip is not entirely without opportunities. The promotion of a hot topic will allow a certain area to rise rapidly. For example, artificial intelligence will drive the AI chip business startup tide. Shenzhen Jian Technology, Horizon, and the Cambrian are also standing in IC Ventures. The gusty, and favored by the VC. In the fifth road show of the Anchuang growth camp, the author also saw several chip startups, their technical capabilities and team members have been unanimously affirmed by the on-site judges.
Chip entrepreneurs want to choose the right ticket
Last year, the AI chip set fire to a complete mess, but now there is no chance to enter the market. It is estimated that this is a question that many startup companies are thinking about. For startup companies, only after closely following hot spots can they be concerned by venture capitalists. To financing. Of course, if the team is enough cattle is another matter, such as Shenzhen Dapu Microelectronics, which performed a road show at the Anchuang Growth Camp, they chose a unique startup project called the Data Storage Processor Chip (DPU).
As to why this entrepreneurial point was chosen, Yang Yafei, chief executive officer of Dapu Microelectronics explained, 'As the technology evolves, the operating speed of the main control chip is getting faster and faster, and the storage capacity of the memory is also getting bigger and bigger, but it is needed for data processing. The data is put into the memory from the memory and processed again. The main control chip and memory are rapidly progressing, but the intermediate storage management and control technology is still relatively backward. From the perspective of the current enterprise SSD market, mainly Samsung, Western Digital, and Intel In the control, most domestic data centers use these products. Our DPU has done a lot of innovation, put computing and many smart elements on the chip, and realized the integration of computing, storage and intelligence. And storage is separate, storage and display are external devices of the processor. We think that in the age of artificial intelligence and big data, storage and computing are converged, so for the first time, some calculations are started on the storage disk. Of course, we started very Be cautious, to make the most basic, general-purpose calculations well. For example, data format algorithm, unstructured data query in disk, and number The sorting tray up done, we feel that this is the future of a revolution in storage. '
In the selection of entrepreneurial projects, Yang Yuxin, co-founder of Anchuang Space, said, 'We are looking for applications related to the scene. The previous artificial intelligence and the Internet of Things were basic technologies. Now the basic technology is more integrated with the actual scene, and we Also look at the development trend of basic technologies. Last year we made AI chips and IoT chips. This year is storage. Because you can see that in the process of technology development, the amount of data is increasing, and storage is a bottleneck. So we mainly Look at two trends, one is the direction of technology, and the other is the direction of the industry.
Sell IP or sell chips?
Can a startup company get financing even if it succeeds? In fact, getting financing is just the beginning of a business. The next thing they have to consider is how to get the product to market and how to make a profit. Almost all startup companies will struggle to make a difference in their lives and deaths. There are basically two choices for chip entrepreneurs: selling IP, or selling chips on their own. Arm is a good example of IP licensing, a typical example of success, is it suitable for entrepreneurs who are entering the field? In this interview , Non-web reporters heard two sounds.
Yang Yuxin believes that IP-only is difficult to survive in the perspective of venture capital. He must do his own chip. He analyzed that 'ARM's success is due to a specific historical period. The chip's early development cycle is very long, ARM's licensing fees From hundreds of thousands of US dollars to tens of millions of US dollars, each chip must wait for two years after shipment, and each tablet will probably take a few cents. IP becomes a chip to prove to the customer that the chip can work, and no one dares to With unverified IP, it takes about half a year to obtain verification. Moreover, IP licensing gives small companies less than the licensing fee, which may be as low as hundreds of thousands and hundreds of thousands of dollars. The smaller the IP is, the less the core revenue is. After waiting a year and a half for users to make chips, startups may have been dragged down. For example, AI chip companies would have problems with ecosystems if they only had chips and no algorithms and scenarios, because only neural network acceleration was needed. With specific acceleration capabilities, it is necessary to do enough development of algorithms and applications. I don't think a startup company can launch an ecosystem, so it must have its own scenario, own algorithm, own chip. AI chip Scheduled to do vertical integration, with scenes and algorithms on their own to do, so start-up companies only IP is not true, at least to sell chips do receive money directly. '
As an entrepreneur, Dr. Yang Yafei shared another example. Dr. Yang once worked in Qualcomm for eight years. As the world’s largest manufacturer of smart phone chips, Qualcomm began developing neural network IP in 2010 and formed a team of 20 people. At that time, the goal was to put it in the mobile phone chip SOC. However, due to the power consumption issue, it had not been able to achieve this goal until the team left to set up Nervana Systems in 2014, and nearly 400 million U.S. dollars was acquired by Intel. Intelligence is still in a melee state. It is easy to copy and plagiarize only the algorithm. Therefore, the entrepreneur has got the financing instinct and said that they need to protect themselves. After that, big chip companies may look for such an IP, which is equivalent to making a module. As a carrier, Huawei Hass uses a third-party IP as the model. As an entrepreneur, if the capital is sufficient, the chip can be boldly designed. If the funds are limited, the chip's landing problem must be taken into consideration. Starting IP is also a development path.
The strategy of Dapu Microelectronics is walking on two legs. Dr. Yang introduced, 'The DPU is still at an early stage. We don't have a big step for each generation of products. For the consumer sector, it belongs to the Red Sea. It's a very tough fight. We use technology. Authorized methods are used by consumer-grade SSD hard drives, notebook computer manufacturers, so that we can extract from the inside; for the enterprise-class market, we will do our own products. Qualcomm also makes money through two channels: one is the product, one is Patent. We have been thinking about how to put the technology into the value. It can be seen that we have a very strong business scenario. Now we just want to optimize this business scenario. We thought of these technologies using machine learning and deep neural networks. Innovation in this area. '
It is also a good place for start-up companies to be acquired by large companies
Unicorns are a vision of most start-up companies. However, after all, only one or two unicorns are common. Ordinary startups are also looking for ways to survive. They can get a stable business to survive. Is the direction of the efforts of all companies. In the field of chips, acquisitions in previous years are very frequent, big fish eat small fish, small fish eat shrimp. Is the start-up company acquired a good destination?
Yang Yuxin analyzes, 'Large companies may have insufficient incentives to be an innovative technology. The company’s small and medium-sized teams do not necessarily have access to resources. The startup companies are desperate to be a technological innovation and do well in the future. May choose, if you are optimistic about this market it is possible to acquire, which is why small companies will be bought. For a startup company, the founder is the company's largest ceiling, how strong the founder, the company can develop to how much Therefore, the founder must clearly recognize his own limit, or where he can break the upper limit. Entrepreneurship is actually related to life. The more early the project is to look at people first, the person decides the development of this project. The acquisition of a company by a large company is also a good destination.
Dr. Yang Yafei expressed his own opinions based on his own experience. 'The difference between China and the United States in the acquisition of companies is very large. After a certain degree of development, U.S. companies will be acquired and acquired. This is a high-probability event, and after mergers and acquisitions All major companies are doing technology integration. After two years, they began to integrate the original system; Domestic is very lacking in this regard. The domestic so-called acquisitions are based on revenue in many cases. Listed companies want to acquire. Or control is mainly for consolidation considerations, that is, to consider how much the company needs to increase its revenue in a new field. Because the original business has no new stories, it is necessary to find a new topic. This is often the case at home. There are many companies that are demanding. Relatively speaking, these big companies in the United States are more lenient and do not actually require the acquisition companies to create more revenues. I think Huawei will continue to take such a path in the future and continue to find The company's mergers and acquisitions, and then integrated within the department.
Chip entrepreneurship is doomed to be more difficult. Fortunately, the government is strongly supporting it. There are also organizations like Anchuang Growth Camp that provide resource matching support for startup companies. It is a good time for domestic entrepreneurs. We will see more chip startups in the future. The company grew up.