Where Are the Entrepreneurship and Investment Opportunities in the AI Era?
-
In 2023, ChatGPT and GPT-4 attracted widespread attention, igniting a new wave of enthusiasm for artificial intelligence. So, where exactly are the entrepreneurial and investment opportunities in the AI field? What constitutes a good industry, company, or team? What can we do in the AI era, and what should we avoid? In this article, Professor Zhang Yu from CEIBS will provide a detailed analysis. (Note: The discussion is based on information available as of early November 2023.)
Before analyzing the investment and entrepreneurial opportunities in the AI industry, let's first review the development of GPT.
Although ChatGPT is currently a global sensation, profoundly impacting various industries, the breakthrough in this technological approach only occurred in recent years, particularly in 2022. In 2019, AI technology was still quite primitive, with some even derisively calling it "artificial stupidity." At that time, large language models had very few parameters and were far from perfect.
The GPT-3 released in 2020 was a general-purpose large language model containing over 100 billion parameters, capable of text generation, translation, and question answering. The GPT-3.5 version released in March 2022 improved upon GPT-3, enhancing conversation quality and logical rigor.
The November 2022 release of ChatGPT truly demonstrated the power and appeal of GPT models. This specialized conversational tool could answer questions based on its knowledge base up to March 2021. Initially just a test version, ChatGPT quickly gained global attention. Within less than three months, it reached 100 million active users, becoming the fastest internet application in history to achieve this milestone. After that, GPT began to accelerate its iteration. GPT-4, released in March 2023, is a multimodal large language model capable of not only generating text content but also understanding image inputs, allowing users to interact directly with images. For developers, the most exciting aspect of GPT-4 was the simultaneous release of its API interface. Users could develop their own applications using the API to create various connections.
The real "game-changer" is GPT-4 Turbo, a major upgrade to GPT-4. GPT-4 Turbo supports a context length of 128K tokens, equivalent to roughly 300 pages of text material. It offers better controllability, a more extensive knowledge base (updated to April 2023), and supports more modalities, including text, audio, video, and images, along with customization and higher concurrency.
How should we view the new wave of AI investment opportunities brought about by GPT's rapid development? Typically, a good investment target should meet the "three good principles": a good industry, a good company, and a good price. What determines a good industry? Ultimately, it boils down to two dimensions: first, whether the potential scale is sufficiently large; second, whether the structure is sufficiently sound. If an industry has a large enough scale and a solid structure, investing in its top companies is likely to yield favorable returns.
Industry scale is determined by demand and cost. Beyond high demand, the ability to meet that demand at a sufficiently low cost is crucial for scaling the industry. Economically, the intersection of demand and cost curves determines an industry's sales volume. In practical terms, it depends on how much value the products and services create for users and partners. The size of an industry is influenced by the value delivered to customers, the resources invested, and the opportunity costs of those resources.
What value does AI actually create? Large language models like ChatGPT, which generate not only text but also images, audio, and video, significantly enhance the efficiency of content and application production while reducing costs. Take text interaction as an example—it generates standardized text far faster than humans. At least for now, AI can replace humans in repetitive tasks. Though imperfect and occasionally inaccurate, it has undeniably transformed our work and daily lives. In the future, AI's intelligence, reasoning, and cognitive abilities will continue to strengthen and improve. What constitutes a good industry structure? Porter's Five Forces model suggests there are five forces determining the scale and intensity of competition in any industry: bargaining power of suppliers, bargaining power of buyers, threat of new entrants, threat of substitutes, and rivalry among existing competitors.
How is created value distributed? Customers, companies, and suppliers all claim portions of it. Whether investing or starting a business, one must consider how much value can be captured from collaborations with customers and suppliers, and what factors determine the creation and appropriation of value.
What makes a good company? Beyond operating in a favorable industry, successful companies and teams require high growth and returns, preferably protected by moats such as brands, patents, proprietary technologies, or switching costs. Additionally, scale effects or network effects are also crucial. When investing or starting a business, we must consider: as the business improves and the user base grows, will it attract even more users? If the number of users continues to increase, can the costs be distributed? If the business we are engaged in or the team we are investing in possesses both network effects and scale effects, then it is undoubtedly a great project or idea. Otherwise, it will only result in linear growth.
In the foreseeable future, AI cannot replace the experience and creativity of professionals because the underlying algorithm is a model of probability and statistics, focusing only on correlation rather than causality. This inevitably leads to issues, but it also presents opportunities for entrepreneurial investment. Regardless of the type of company, it is essential to consider whether network effects and experience curve effects can be generated in specific scenarios, products, or applications. Amazon provides an excellent example in this regard. Its growth flywheel operates on two positive feedback loops: more users attract more sellers, and a greater selection improves customer experience. The resulting growth reduces costs and prices, which in turn attracts more customers and increases traffic—creating a compounding effect. Amazon's market capitalization grew 2,500 times over 20 years, a phenomenal growth rate in the mobile internet and internet era, all thanks to this positive feedback effect.
This is precisely the strategy ChatGPT has adopted. Even when users were queuing up to purchase access, the company did not raise server prices—ensuring widespread adoption of its services. The larger the user base, the higher the efficiency, leading to lower costs and continuous improvements in performance and product quality, ultimately capturing the market.
The AI industry's leading companies fall into several categories. First are well-funded, technologically advanced corporations with dedicated teams capable of purchasing computing power, fine-tuning or enhancing large models, and developing chips—such as OpenAI, NVIDIA, Google, Facebook (Meta), and Amazon. Another category comprises companies with unique expertise in specific scenarios and data applications. How to compete with giants? First, do what they cannot and do not intend to do; second, do what they are unwilling to do. Even the most powerful giants and corporations have things they cannot or will not do, and these are your opportunities for investment and entrepreneurship.
The final point is a good price. Price is intrinsic value multiplied by market sentiment. In the short term, the market is a 'voting machine,' but in the long term, it is a 'weighing machine.' If market sentiment is overly optimistic, with everyone chasing trends and hype, entering the market at such a time will come at a high cost. The AI industry is just beginning to develop, with its industrial chain consisting of upstream data, midstream models, algorithms, and computing power, and downstream applications and distribution. Although the scale and potential of the upstream, midstream, and downstream sectors are all substantial, we need to carefully consider what is feasible and what is not.
Currently, China is experiencing a 'hundred-model battle.' In this context, unless one has substantial financial and technical resources, it's advisable not to invest in chips or large models when considering related investments or entrepreneurship. Large models and computing power are critical national strategies, leaving little room for ordinary investors and entrepreneurs. Conversely, opportunities for ordinary individuals may lie in application creation. Since ChatGPT remains primarily software-oriented, current investment and entrepreneurial opportunities in China are likely concentrated in downstream applications/distribution or upstream data supply. It must be emphasized that if AI is only used as a tool for cost reduction and efficiency improvement, the end result will be a competition in computing power and costs. What we truly need to do is cultivate unique experiences, creativity, and inspiration, which require proactive thinking and exploration. In fact, truly outstanding companies, while meeting customers' basic needs, often also consider providing value in emotions, lifestyle, self-actualization, and transcendence. Whether developing B2C or B2B products, we must think about how to deliver personalized value to customers.
Personalized value stems from pain points in customers' lives or work. In the past, solving such pain points might have required a team of 8–10 people, but now, empowered by AI, the team may only need to focus on fine-tuning. In the future, we may see many solo-entrepreneur teams or companies—a scenario never seen in the past two decades—and this presents an opportunity for each individual to create value. Among the first thousands of ChatGPT assistants and apps to emerge, those that truly achieve commercial success or strike gold in this field will likely not be the fastest or most accurate, but those capable of solving specific needs for particular user groups. The same logic applies to investment—the team you invest in doesn't necessarily need the most impressive model, but must have profound insight into user needs and value.
Does the value created differ when you're involved versus when you're not? If the company you founded or the team you're part ceased to exist, would the world be different? Does the business you're engaged in, the company, or the team provide additional value to the upstream and downstream of the industrial chain? In the complex landscape of entrepreneurship, the high-tech industry's supply chain is particularly intricate. Whether investing or starting a business, we must carefully consider these questions and determine whether our business is a relatively important or unique link in the supply chain. If starting a business, you must contribute to the upstream and downstream of the industrial chain; if investing, you must invest in such teams, companies, and enterprises. Many people view AI as a replacement for humans, but its more crucial role is in equalizing opportunities. It rapidly empowers everyone with skills that might have taken eight, ten, or even more years to acquire traditionally. As long as you have an irreplaceable advantage in a certain area, you hold your own value. Perhaps, when everyone is pursuing AI, providing excellent coffee services to AI professionals becomes a great business; when everyone is making coffee, perfecting the art of making jianbing (Chinese crepes) becomes a lucrative venture. There is no hierarchy in entrepreneurship and investment; the most important thing is not to overbid and avoid investing in overcrowded areas.
Professor's Profile Dr. Zhang Yu is a Professor of Strategy and Chair of the Department of Strategy and Entrepreneurship at CEIBS. He earned his Ph.D. in Management from INSEAD. Prior to joining CEIBS, he served as an Assistant Professor of Strategy at the Paul Merage School of Business, University of California, Irvine. His teaching areas primarily include strategic management, industry and competitive analysis, innovation and competition, as well as corporate transformation and upgrading.
Professor Zhang's research interests mainly focus on the interaction between strategy and capital markets. His research has been published in top-tier international journals such as the Academy of Management Journal, Organization Science, and Strategic Management Journal. He has been invited to deliver academic lectures at renowned business schools in Europe, America, and Asia, and has given keynote speeches at prestigious industry conferences such as the Annual Conference of the National Association of Corporate Directors.