Interview: Sima Linwei
Written by: Chen Xiaorui
On April 18, Aethir launched the Aethir Edge product, which is powered by Qualcomm technology and equipped with Qualcomm Snapdragon 865 chip, during Token 2049 in Dubai.
Qualcomm is a semiconductor and communications technology company founded in 1985 and headquartered in San Diego, California, USA. Its main businesses include the research and development and sales of mobile communication chips, as well as the licensing of related patent technologies. Its products and technologies are widely used in a variety of wireless communication devices such as smartphones, tablets, automobiles, and IoT devices. Aethir is a decentralized GPU cloud infrastructure provider founded in 2021 and headquartered in Singapore. Public information shows that in July 2023, Aethir completed a $9 million Pre-A round of financing, with a valuation of $150 million at the time.
Recently, DeThings interviewed Andy Li, business development director of Qualcomm. The following is the full text, which has been edited.
DeThings: Can you briefly popularize and introduce Qualcomm's technology and business?
Andy Li: Qualcomm has accumulated a lot of technologies in the process of more than 30 years of development. What everyone knows best is our technology in the field of mobile communications, including the earliest CDMA, 3G UMTS and CDMA2000, 4G LTE, and now 5G. Qualcomm recently released the 5.5G evolution plan, which we call 5.5G, which is actually the evolution of 5G. Our laboratory has successfully realized 6G technology. Of course, the specific commercialization level depends on the market.
In addition to mobile communications, we also have technologies in computing fields such as CPU and GPU, such as Snapdragon CPU and GPU. Our GPU has achieved a good balance between high performance and low power consumption. Qualcomm has had its own CDSP architecture for a long time, formerly known as CDSP (C stand for computing), which is a DSP for computing. With the continuous development of AI algorithms such as computer vision, CNN, TensorFlow, etc. in recent years, Qualcomm has been adjusting the CDSP architecture to adapt to the latest technologies. Recently, we have also launched Qualcomm NPU, including matrix accelerators, tensor accelerators and vector accelerators, and have also made many optimizations in memory sharing bandwidth and access.
So Qualcomm's technology is not only about mobile communications, but also about technologies such as computing. The first step in applying these technologies to the Internet of Things is to connect objects, which requires the use of Qualcomm's connection technology, which includes not only cellular network connections, but also near-field communication technologies such as Bluetooth and WiFi. After connecting thousands of objects, the problem of data islands is solved. Because each object may be just a small sensor, but it is also constantly generating data. In the past, when it was not connected, the data of each device existed like an island, and it was necessary to manually pull and analyze the data. After the connection, these data form a network. At the network level, the AI big model can automatically learn and analyze these data to extract value from them.
DeThings: As we move from the PC era to the mobile Internet era, the new term "Internet of Things" (IoT) has emerged. What is your opinion on the term "Internet of Things"?
Andy Li: For Qualcomm, the Internet of Things is not a new term. The Internet of Things has actually existed for a long time. As the name suggests, the Internet of Things is to connect objects into a network, which is the first step. We have a very long history in this field. You may know Qualcomm as doing mobile communication-related business, such as the Snapdragon platform on the mobile phone side, but in fact, Qualcomm's business is very broad. Qualcomm positions itself as a partner that empowers thousands of industries, using our product and technology combination. The Qualcomm product that everyone is more familiar with is the Snapdragon brand, and recently it has also released PC and XR (VR/AR) brand products.
Whether it is Depins or the so-called Web3, they are essentially the same as the traditional Internet of Things, that is, connecting all devices. The difference is that in the past, due to the weak computing power of each device, a powerful centralized computing platform was needed for control. But with the increase in the computing power of each node, decentralization or Depins has become a possibility, because only when each node is powerful enough can decentralization be achieved, otherwise a powerful central brain is required for processing. When every detail has a certain computing power, a flat network can be built. This is where Qualcomm's opportunity lies. We have a combination of technologies on the edge and terminal device side that can give each node powerful capabilities, coupled with connection technology to connect them. Under this architecture, the so-called decentralized network can be truly realized.
DeThings: After the emergence of Chatgpt, everyone's attention has been attracted to AI. How does Qualcomm view computing in the AI era?
Andy Li: I think AI is the intelligent productivity that can change the production structure of the entire human society again since the second industrial revolution, and then change the organizational structure of our entire society. However, ordinary people may have a generalized understanding of AI, that is, artificial intelligence. So how can we implement artificial intelligence into our daily lives and make it concrete?
First, we may have seen general generative AI like ChatGPT. But ChatGPT represents a general AI, which is actually a kind of general knowledge. To achieve general knowledge, it means you need to understand all knowledge. Therefore, the parameter scale of general AI like ChatGPT will become larger and larger, and models with more than 200 billion parameters have already appeared.
But back to the IoT or our industry terminal field, my point is that whether you really need a general generative AI like ChatGPT needs further discussion. Because in the field of IoT or industry terminal devices, they will not be used to do everything. When designing a device, we already think that it needs to do a limited number of things in the field it faces. Just like we won't ask the radio "What gift should I give my wife for her birthday tomorrow", we will only ask operational questions related to the device. In this case, do we still need a general AI? I think it may not be necessary. It needs a customized, vertical AI model for its field.
In this way, we further focus on the scope of AI. From the perspective of hardware requirements, it will also be reduced accordingly. Because if you want to run a model with tens of billions of parameters, you can only run it in a data center. On a mobile phone, you may only be able to run a model with more than a billion parameters. For IoT devices, a small model with 7 billion or even 4 billion parameters is enough. What we need to do is not to stack hardware to run large models, but to focus on how to make AI models run more efficiently.
Qualcomm also has relevant arrangements in this regard. In addition to continuously improving chip capabilities so that devices that could not run large models before can now run, we have also launched Qualcomm's AI framework. It should be noted that Qualcomm does not make large models, but provides a complete set of tools. We launched AI Hub at the Embedded World in Nuremberg, Germany last week. In fact, hundreds of AI models in the open source community were ported to the Qualcomm Snapdragon platform, optimized, and open sourced for all developers to use. We hope that partner developers can develop AI solutions on the Qualcomm platform that can truly be used by thousands of people and in thousands of industries.
DeThings: In other words, do you or Qualcomm think that AI is a so-called "winner takes all" field? For example, we know that OpenAI is leading the development of a large model with trillions of parameters. In this powerful AGI (artificial general intelligence) situation, why is there a need for other AI?
Andy Li: That's a very good question you raise, and I'll actually talk about it in my talk. As you said, if AI is a "winner takes all" field, then it must be a giant AI that can only run in the cloud because neither edge devices nor terminal devices can support it.
First, every specialized AI model is derived from a general large model, we just customize and personalize it according to industry needs. Therefore, we need general AI and then segment it.
Back to your question, why don't all terminal devices and industry terminal devices be connected to the cloud? This involves the issue of economic benefits. As you know, a huge data center in the cloud has a very high cost. Even if it is just a simple inference, it needs to be fully started and consumes a huge amount of energy, including the energy itself and the energy consumption of the computer room air conditioner. Now, the data center is built in colder areas and closer to the power supply area to reduce these expenses. This is one aspect.
Second, we need to consider that there will be more and more AI application scenarios and more and more people using them. Assuming that there are 1 billion people using AI in the world in the future, the energy consumed by each AI startup and reasoning multiplied by so many application scenarios and people will have a huge impact on economic benefits. In this case, is it the best solution to put all calculations in the cloud? Or can we put some calculations on the edge and terminal devices? I think the latter is necessary.
More importantly, many industry terminal applications care more about latency and latency reliability. If it is through the cloud, even if you tell me that it does not exceed 20 milliseconds each time, the latency may vary from 10 milliseconds to 100 milliseconds. This uncertain latency is unacceptable for many industry applications. On the edge and terminal devices, you can get a certain and very low latency. That is, the reliability of the connection.
The third point is the security of data, which is also an important reason for us to promote decentralization, to ensure the security and reliability of data. Of course, we pay more attention to the security of data. For reliability, we hope that the data is stored locally or within a limited range that is controllable and accessible, as a node in a larger decentralized network.
Therefore, from the perspective of the AI era, general-purpose large models may not be able to meet all specific needs in terms of economic efficiency. In addition, general-purpose large models also have some defects in terms of reliability and security, and still require industry-customized solutions. General-purpose large models are not a universal answer.
DeThings: We know that global mobile phone shipments are shrinking. Everyone has this stereotype about Qualcomm, that is, Qualcomm's Snapdragon platform is very strong in making mobile phone chips. We want to know where Qualcomm thinks its next growth point is?
Andy Li: First of all, although global mobile phone shipments have declined compared to before the epidemic, they are slowly recovering.
Secondly, we still regard mobile phones as our main business, and there is no change. Because the research and development of many of our technologies started with mobile phones, the mobile phone market is actually the largest single demand market in history. Although there are only a few mobile phone brands, their needs are still very similar, so the mobile phone market is a very good soil for us to cultivate new technologies.
In addition, Qualcomm is indeed pursuing business diversification. You can see that we have entered the fields of IoT, XR (extended reality), PC and automobile. For example, in the automotive field, whether it is advanced driver assistance systems or in-vehicle infotainment systems, Qualcomm is in a leading position. We will continue to promote this diversification strategy.
The scale of these emerging markets is also very impressive. Take the field of Internet of Things as an example. We believe that its entire market capacity is as high as 720 billion US dollars, which is even larger than the mobile phone market. Although the Internet of Things is a highly fragmented market, the volume of each industry may not be large, but each industry has its own independent needs. But for us, we can penetrate the technical demands behind the applications of different industries, and these demands are actually the same. Our technology and patent portfolio is very flexible and can provide different combinations of technology, products and solutions for different industries. I believe that this combination of technology and patents is also very beneficial to customers, and can bring them the most advanced technology in the most effective way.
DeThings: In addition to the core business of mobile phones, what do you think is the most promising growth point for Qualcomm in the future?
Andy Li: In addition to the mobile phone business, we believe that one of the biggest growth points lies in the Internet of Things and the automotive field.
The Internet of Things is actually one of the biggest growth points. In addition, whether it is new energy vehicles or fuel vehicles, we believe that the automotive field has great potential. Even fuel vehicles have the need for intelligence. Although new energy vehicles do not have to worry about battery problems and are easier to design as new products, fuel vehicles also need intelligent upgrades. For us, the driving form of a car is just a form. Our goal is to make cars smarter and realize the interconnection between cars, cars and transportation facilities, and more importantly, cars and drivers. This is exactly what Qualcomm excels in.
In fact, cars can be considered part of the Internet of Things because we view them as smart connected devices. Therefore, cars still belong to Qualcomm's overall layout of the Internet of Things ecosystem.
We believe that Qualcomm has unique advantages in these areas because we have a wide range of partners, a large customer base and partnerships. Qualcomm has always been sincere in sharing our latest technologies with partners without hesitation. We believe that our customers will benefit from this, and the entire society will also benefit from it.
Dethings: What do you think of the concept of Web3? Because in addition to AI, the term Web3 has also been very popular in the past few years.
Andy Li: For Web3, we believe it is a very emerging economic activity with potential and prospects. For Qualcomm, we play the role of an enabler because Web3 is decentralized, there is no absolute boss, and everyone is an equal contributor.
As I said before, Qualcomm has been and will become more open. We will open our hardware platform and even open source our software, embrace more open source software, and we ourselves have a certain degree of open source code. We will empower everyone to enable our partners and customers to play a role in Web3 and even Web4 in the future. We provide them with weapons to enable them to display in this vast field. Because we believe that AI is a revolution in productivity, then Web3 is about the transformation of production relations. So Qualcomm treats this emerging industry in an open role as an enabler.
Dethings: Why did you choose Aethir, a Depin platform, for cooperation and experimentation? For companies like Qualcomm, it seems that they rarely get involved in the Web3 field. Web3 has strong financial attributes and involves many aspects of token economy. Does Qualcomm also have concerns in this regard?
Andy Li: Aethir is actually a partner we are very optimistic about. They have accumulated a lot of experience in both cloud computing and edge computing. If you visit Aethir's website, you can see real-time data display of their resources, such as the GPUs that are running, etc., which I think is very good. Because they are not a conceptual thing, but a real running business, which is truly implemented in business practice. For a pragmatic company like Qualcomm, we also hope to cooperate with such a pragmatic business partner, whether it is to build a role model or to build a real production tool.
In addition, we are not involved in the economic level you mentioned, which is beyond the scope of Qualcomm. We focus more on the technical level. We believe that technology is borderless and there is no good or bad technology itself. We empower everyone with technology. But I personally do not think that Web3 is a "disaster". We need to understand and embrace all new things. Only after understanding can we use them correctly.
Qualcomm's role in Web3 is an enabler. We hope that our customers can use Qualcomm's technology and products to create products that are conducive to social progress and development within the legal and ethical framework.
Dethings: We have noticed that companies like Google, Microsoft, and Nvidia already have some Web3 customers in terms of infrastructure construction. Does Qualcomm have any further observations or attempts in the Web3 field?
Andy Li: We are open to this and welcome all customers to contact us. We believe that Web3 represents a new type of production relationship. As a technology enabler, Qualcomm is willing to share its capabilities and technologies.