ChatGPT Will Drive Strong Long-term Growth in the Chip Market in the Future

Almost overnight, ChatGPT exploded out of the loop and swept through many countries and regions around the world. At the end of November last year, ChatGPT quickly became the fastest-growing consumer app in history after its launch, with more than 1 million registered users in just 5 days, and two months later the figure had already exceeded 100 million. In comparison, TikTok took 9 months to reach the 100 million user threshold, Instagram took 30 months, WhatsApp and Meta took 42 months and 54 months respectively, and Twitter took 60 months. So, what is the background of ChatGPT's explosion, and why? Does it mark the maturity of AI commercialization? What is the international landscape of this industry? And how will it stir up the underlying semiconductor industry, HPC, and cloud computing?

ChatGPT Sets Off a New AI Revolution

As a new artificial intelligence chatbot, ChatGPT's unstoppable power is believed to be setting off a new round of AI revolution.

It is understood that ChatGPT is a model developed by OpenAI using artificial intelligence technology trained to provide human-like content output through the analysis and training of large amounts of information and data from the Internet. Its appeal lies in its ability to answer open-ended questions in addition to specific knowledge-based questions and to interact with users in a conversational manner.

Today, people from all walks of life are sharing their experiences with ChatGPT on social media platforms. For example, students use it to write assignments and papers, self-publishing bloggers use it to write articles and advertisements, programmers use it to program, and others use it to write industry analysis reports, auto-translate, create novels, etc.

As to the reason for ChatGP's explosion, the general manager of Chundao Investment Consulting, Rixin Bu, said in an interview with Jiwei.com that, firstly, the ChatGPT model upgrade really gives everyone a bright feeling and is one step closer to the imagined artificial intelligence. Secondly, from the application field, taking search as an example, ChatGPT does have great practicality and helps to filter valuable information from complicated information.

“In addition, there is a more realistic reason that the AI field and the capital market also need a new windfall to stimulate after experiencing a -time low.” Multiple factors have contributed to the high level of industry interest in ChatGPT, noted Step Rixin.

However, some industry players also pointed out that although ChatGPT has a strong sense of fluency and logic, there are still problems such as low-quality content and logic stronger than content. Also, some of the responses to questions may not be accurate, truthful, and timely.

Benefit From the Iterative and Organic Integration of the Industry Chain

Undoubtedly, ChatGPT is a hot circle due to the rapid development, continuous iteration, and organic combination of high-performance computing, artificial intelligence, and semiconductor technologies.

It is reported that the development logic of ChatGPT is: by using high-quality artificial labeled data + reinforcement learning as the underlying logic, as well as continuous learning and iteration after large-scale corpus input, and finally relying on powerful computing power to support the product learning and input and output.

Further, ChatGPT is based on the enhancement of the GPT-3.5 model. The training dataset was initially trained on Microsoft's Azure AI cloud server, and then a substantial amount of code and human feedback mechanisms training were added to make it more accurate and user-friendly.

According to data, ChatGPT's total computing power consumption is about 3,640 PF-days, and industry sources say that ChatGPT has imported at least 10,000 high-end NVIDIA GPUs, while Citigroup estimates that ChatGPT will likely drive sales of Nvidia-related products to $3 billion to $11 billion within 12 months.

Driven by the rapid growth of ChatGPT, a GPU industry executive told Tibco that this will drive a big increase in arithmetic demand, with 350-watt inference chips becoming the norm, as well as a new surge in technology that will drive large model compression, while there are many services for enterprises and individuals that can be built on similar technology.

On the flip side, what does HPC mean for ChatGPT?

As Nvidia CEO Jen-Hsun Huang said, "You can't get to the moon even if the ladder is high, you need a rocket to get there." The question facing the industry is no longer what a computer can do, but who can be the first to use it to drive industry innovation. Today, AI, combined with accelerated computing and high-performance computing, forms a digital flywheel that drives super-exponential progress.

According to Jen-Hsun Huang, this industrial revolution in high-performance computing will sweep through data centers, public and hybrid clouds, and edge networks, and is about to reach a turning point in its development.

In this regard, Rixin Bu further said that high-performance computing chips are the underlying foundation of all AI, and currently there are two main types of Al chips: general-purpose GPU arithmetic chips and ASIC special chips. Al applications in the natural language processing class are currently still algorithm-based, and the underlying arithmetic chip is still mainly GPU, but the market pattern is more obvious, Nvidia occupies a major market position.

“At present, domestic companies are also trying, and gradually began to expand NLP in the intelligent terminal side of the speech semantic processing, which requires the chip function and performance upgrades to meet the terminal side of the arithmetic power and power balance, while this also belongs to the category of dedicated Al chips.” Bu Rixin said.

ChatGPT Will Drive Strong Long-term Chip Market Growth in the Future

As digital technology continues to penetrate into all areas of the economy and society, the global demand for computing power is growing rapidly. It is foreseeable that in the new round of technological and economic changes dominated by data centers, HPC will be intertwined and symbiotic with the cloud and AI. It is obvious that CPU+GPU will be an important cornerstone for the development of HPC, while the heterogeneous system of HPC is getting richer, such as high-performance hardware is shifting to take into account AI, big data processing, deep learning, and other applications. In addition, DPUs will also drive the overall performance of data centers because they can dynamically offload network, storage, computing, and other network infrastructure.

How will the rapid growth of HPC affect the future development of chips and other related industries? Rixin Bu said, “HPC belongs to the category of infrastructure construction, which is the core support of information technology. In the future, the popularity of AIGC applications represented by ChatGPT will greatly increase the new demand for computing power, which belongs to the incremental market.”

Although giants such as Microsoft, Google, and Meta are currently actively competing for the layout of ChatGPT, Rixin Bu believes that it is difficult to form an absolute monopoly in the AI field, and the explosion of ChatGPT can only be considered a beginning. As the industry's attention rises, related competing products will also be revealed one by one.

“Currently, there are already a number of domestic enterprises in the field of pre-training large models which have a long-term layout, such as Ali, Tencent, Baidu, Wave, and other related products, and the future of all these companies will gradually embed related technologies into their own platforms.” Bu Rixin said.

It is understood that China has an advantage in individual segments of the Al field, but overall it still plays a follower role. However, with ChatGPT becoming a global technological innovation breakthrough, China may set off another round of AI development boom. Meanwhile, the strong investments made by a number of international technology giants will also increase China's confidence in AI basic R&D, capital investments, and other strategic investments, as well as in the promotion and popularization of AI.

JP Morgan Asia Pacific technology, media, and telecommunications research co-head Gokul Hariharan (Gokul Hariharan) said: "We expect that the overall chip market in the first half of 2023 will be a significant decline. However, at the same time, strong long-term growth will be seen in the coming years, driven by high-performance computing."

Overall, ChatGPT in the global blowing, artificial intelligence applications whirlwind at the same time, will also stir the underlying semiconductor industry, HPC, cloud computing, and other industries synergistic development, which in turn will lead to the relevant hardware and software and ecological complement each other, and even promote the recovery of the semiconductor industry and the sustainable long-term development of related industries.

Related News

1、Chip Packaging Lead Time Has Grown to 50 Weeks

2、Eight Internet of Things (IoT) Trends for 2022

3、Demand for Automotive Chips Will Surge 300%

4、Volkswagen CFO: Chip Supply Shortage Will Continue Until 2024

5、BMW CEO: The Car Chip Problem Will Not Be Solved Until 2023

6、Shenzhen: This Year Will Focus on Promoting SMIC and CR Micro 12-inch Project