According to relevant statistics, the number of networked devices will increase 20 times in the next decade, which will bring a new market of up to $7 trillion, followed by a large number of chip opportunities. The report released by IC insights, a market research institution, pointed out that benefiting from the strong demand of the market, the revenue of the overall chip market is expected to increase by 24% this year and break through the first $500 billion mark in history.
New trend of chip industry in the era of Internet of things
They further pointed out that this growth trend is expected to continue until 2023. During the forecast period (2020-2025), the annual compound growth rate of the chip market will reach 10.7%. By 2023, the global chip market revenue will exceed US $600 billion. With the global deployment of Internet of things devices, this proportion will increase significantly, but the chip R & D demand in the era of intelligent Internet of things is relatively different from that in the past, and the development of the whole industry also needs more talent investment.
In addition, the Internet of things will also bring different challenges to the chip industry.
New demand for chips brought by the Internet of things
Looking back at the development of the chip industry over the past years, whether it is the mainframe era, the PC era or the mobile phone era, they all show a characteristic, that is, they are single variety and large-scale products. Especially in the mobile phone era, the shipment of single products can reach the level of 1 billion. That means developers can focus on making chips for these single product markets. It is precisely because of such a huge shipment support that developers can have enough financial resources to invest in the R & D of the next generation of products.
However, when we come to the Internet of things market, as we all know, the Internet of things is a market superimposed by countless subdivided and fragmented applications. Although the scale is huge, the TAM of each subdivided market is relatively small (few more than 100 million). At the same time, because of the miniaturization characteristics of most Internet of things products, This makes it more sensitive to the cost of the chip.
For chip manufacturers, they also need to face customers in the industry from the top 10 mobile phone manufacturers and top 10 computer manufacturers to thousands of small and medium-sized intelligent hardware IOT hardware manufacturers. The diversification of customers also makes chip manufacturers lose the opportunity to define future product specifications with customers. R & D needs people who can hear the fire to make decisions. Diversified customers also hope to improve the differentiation and competitiveness of their products through differentiated chips, and put forward a variety of customization needs. On the one hand, such customization increases R & D investment; On the other hand, it also limits the application scope of the product, further shredding the market and reducing the shipment of each custom chip.
However, even so, the demand of end customers for chip processing capacity has not weakened. At the moment when Moore’s law slows down, it brings them new problems. Statistics show that after 2015, it will take 10 years to double the performance of the chip, which is different from buying back double computing power with the same money every 18 months in the past 50 years.
However, the Internet of things will not reduce the demand for performance, and customized design and advanced nodes mean higher R & D investment, IP and streaming costs. As mentioned above, the fragmented market reduces the shipment and profit of a single product. With more and more investment and less money, this contradiction has become the bottleneck restricting the popularity of Internet of things chips.
Finally, how to effectively seize this fragmented market is another huge challenge to the profit model of traditional chips by the Internet of things. There is no doubt that the chip industry has encountered unprecedented new challenges in the era of the Internet of things.
Countermeasures of chip Enterprises
Facing the current competitive situation of the Internet of things, chip manufacturers also have their new countermeasures. For example, in the pursuit of performance, we no longer blindly seek speed and innovation, but need to find the appropriate optimization direction and the most appropriate process technology, which has become a common trend in the industry. Because although the research and development of chips become more and more refined, these customizations can improve the performance a hundred times under the same process conditions.
At the same time, in order to embrace the new market demand of the Internet of things in the future, chip companies have to make multiple products at the same time, and the shipments, sales and profits superimposed by multiple application directions can maintain the normal development of a company.
In order to solve the problem of chip profitability, customized chiplet has become a new path for manufacturers. In their view, this new method solves the contradiction between the increasingly high chip R & D cost and the difficulty of sharing the cost of low shipments encountered by the chip industry in the Internet of things world.
It is understood that by splitting an SOC module into several key small chips, each small chip can be shipped to 10 or even 100 applications at the same time to balance the R & D cost. Adding up the amount of 10-100 million, 10-100 applications can also reach more than 1 billion.
This makes the problems mentioned above easily solved.
In addition, in the past, it became infeasible for large companies to develop chips according to Moore’s law and verify product R & D planning with core key customers. It is difficult for large companies to use the “top-level planning + saturated investment” of the large group army to develop these shipped products for the Internet of things less than 100 million. This requires people who can hear the fire to make a decision. Because only entrepreneurial teams who can hear the gunfire can recognize and grasp the advantages of the fragmented market. New subdivided tracks are constantly emerging. A team must be able to grasp more than 10 subdivided tracks at the same time, and the shipment can reach the order of 1 billion. At the same time, it must have efficient R & D and supply chain capabilities, so that the profits of the company or team can reach more than 100 million yuan.
To this end, the company has created many amoeba style innovation departments to explore new directions. However, this raises new problems. That is, after years of exploration, these new departments think that after clarifying the direction, they can not get the support of the company level. After all, their potential revenue and shipment are 1 ~ 2 orders of magnitude different from the company’s mainstream products. A large number of heads of innovation departments come out to start businesses with the support of capital, which gives birth to a large number of small and medium-sized companies.
In this case, the chip supply chain is required to complete a round of “upgrading” from top to bottom.
“Upgrading” of chip supply chain
In the past, no matter EDA, IP, wafer foundry, packaging and testing, semiconductor suppliers were prepared for mass production. Hundreds of millions of single products were shipped or even 1 billion. Shipping is the access threshold of this supply chain system. However, with the advent of the Internet of things, the original system has become a thing of the past. Everyone is based on the “28” principle or even the “19” principle. The top 10% – 20% of suppliers can account for more than 80% of the total revenue.
In the face of a large number of amoeba like small and medium-sized organizations and companies emerging in the era of Internet of things, efficient and flexible industrial chain support is needed. In this situation, chip companies first need to enhance the awareness of cooperation and collaboration;
In the era of Internet of things, companies and teams also need to adjust themselves. They can’t wait for several years of research and development as before, and expect products to be a blockbuster. Instead, they should learn how to cooperate with other companies in the industrial chain in the process of research and development, so as to complete product R & D and verification as quickly, low-cost and efficiently as possible, Work with partners in the industrial chain to win as many sub categories as possible as soon as possible. In the era of Internet of things, there is no chance of a generation of champion, and chip companies cannot break out with the direct performance of a product developed for many years.
Secondly, EDA / IP should also realize that the later sales sharing of products is the direction of EDA / IP reform in the future. This is mainly related to the uncertainty of the success of the application direction of the Internet of things;
Nowadays, we hope to try 5-10 applications in different directions at the same time, but these applications need different IP addresses. In this case, neither the departments of large companies nor small and medium-sized start-ups have enough funds to buy these EDA and IP at the same time. Therefore, they hope to reduce the down payment in the early stage and share in the sales of products after the success of products in the future.
Such a sales model does not exist in EDA and IP manufacturers today. We still charge the main fees with license authorization. However, from the characteristics of the Internet of things industry and chips, the above business ideas are a new direction that EDA / IP has to think about.
Third, the internal technical support and operation structure of the wafer factory also need to improve efficiency;
In the past, the customer structure of the wafer factory was 20%, accounting for more than 80% of the revenue and capacity. However, in the era of the Internet of things, wafer factories can not ignore the revenue contributed by 80% of the long tail customers and the innovation direction represented by the Internet of things. Because no matter the innovation departments of large companies or small and medium-sized companies, they need a variety of process technologies to try more subdivision directions. Although there is a demand for less than 1000 wafers a year after the quantity of many products is increased, the overall contribution of wafers is also objective. This is also why we believe that wafer foundry must improve the efficiency of current technical support and operation structure, and support the flexible needs of these long tail customers.
Fourth, the sealing and testing plant also needs to form efficient solutions for SIP and chiplet;
On the one hand, packaging factories have similar problems to wafer foundry, that is, the flexible demand for long tail customers, how to improve operation efficiency to support a large number of product R & D and verification needs and long tail mass production needs. On the other hand, the new demand for SIP and chiplet chips also makes them need to constantly improve the efficiency of packaging design, and even integrate products from different customers, To create a competitive SIP solution.
Test factories also need to continuously improve efficiency, because with the popularity of SIP, chip testing will require more and more accuracy and projects, and the proportion of testing in the cost structure will be larger and larger. At this time, inefficient testing process will become an unbearable part of product cost.
For the chip supply chain, this new battle is about to enter a white hot stage. As for who will be the final winner? In my opinion, it depends on who can take the lead in completing the new “revolution” under the new demand.