Author: Yiwen Hao
In 2017, artificial intelligence (AI) set the trend in the venture capital industry. Source Code Capital believes there are some hidden gaps and dangers in the AI investment sector that are often overlooked or not yet recognized. Thoroughly analyzed and written, Source Code Capital exclusively presents Source Code Research Report Issue No. 2.
- From our view in China, the majority of existing AI projects are still in the inception stage of development and still have a long way to go before getting to the stage of commercialization.
- Currently, we see two gaps in the AI talent space. There are not enough outstanding qualified tech talents, and next, there is a gap in terms of qualified business talents.
- The main exit strategy for AI-orientated projects tends to be acquisition by industry giants (in 2015, we counted 69 M&A exits, for 2016, 84). At present, we do not believe IPOs will be the main exit path in this area. We therefore note that it is still unclear how Venture Capital (VC) firms backing AI projects are able to achieve liquidity.
AI: The hottest trend in the VC space in 2017
According to Rhino Data’s statistics, between late 2016 and early 2017, AI was mentioned a total of 48 times in 98 separate discussions pertaining to last year’s investment trend by large domestic institutions. This was 1.8 times higher than “culture and entertainment,” the second most discussed topic.
According to Venture Scanner’s data from 2016, on a global scale, the AI sector and AI start-ups received USD $5 billion to invest in 658 AI companies. The AI market is undoubtedly red-hot. Nevertheless, behind this excitement and hype, we believe there are some “hidden dangers” that are not often identified.
Commercialization and valuation bubble
We agree that the popularization and explosion of AI technology is indeed an inevitable trend. Nonetheless, the majority of AI projects we have seen are mostly in the inception stage of the development, and likely still have a long way to go before culminating in commercialization. In comparison to the SaaS industry, which is B2B-based and makes solving practical corporate problems its main focus, there is definite potential for a valuation bubble within the AI sector.
Within the industry, there is a consensus with regards to the valuation of a SaaS company, that is based on a Forward P/S ratio of 5-10. However, valuation-setting for AI-type projects are still being debated, and right now, some valuation methods include reviewing of resumes, looking at the number of papers published, or looking at P/P.
Furthermore, there is a plethora of “to-be AI” projects that have experienced an overhaul and are now flooding into the market. There are even some consumer-based projects where the key points of a business model do not involve the application of AI technology, but are part of this “to-be-AI” hype and managed to secure financing amidst the craze. On one hand, technical jargon is often tossed at inexperienced investors, misguiding them about the actual value of the start-up. On the other hand, this results in a lot of additional technical due diligence for others to uncover an AI company’s actual unique characteristics.
There has been an increase in demand for AI tech talent and a two-fold increase in pay, as start-up companies find it difficult to recruit. Since 2016, there has been a 3:10 ratio in supply and demand for AI development jobs based in North America. The supply and demand disparity of qualified professionals in China is even more exacerbated with a ratio of 1:10. The industrial sector is seeing an influx of talent from academic circles, while many large companies have gone on to create their own start-ups.
The largest gap in AI talent is the lack of qualified, tech-savvy individuals, and we think the next gap will be the lack of qualified business talents in the AI business. If you understand technological boundaries and the difficulty of implementation mechanisms, then you must understand the necessity for “interpreter-type” abilities.
Right now, the majority of Chinese AI entrepreneurs and founders are returnees who have doctorates from colleges abroad, and have worked in the industry for 3-5 years. They swiftly climbed up the corporate ladder and do possess business acumen.
Second, R&D departmental executives in large companies can be compared to an assistant professor of 30-40 years old with good academic influence. These individuals are often determined to abandon all other pursuits in order to join the AI business profession. Scientific entrepreneurs need to hasten their search for business partners who understand the industry they wish to engage in.
Liquidity and Exit Pathways
The main exit strategy for AI-orientated projects tends to be acquisition by industry giants (in 2015, we counted 69 M&A cases, and for 2016, 84 cases). At present, we do not believe IPOs will be the main exit path in this area, unlike the big data segment. We therefore note that it is still unclear how VC firms backing AI projects are able to achieve liquidity.
The time it takes for technology to go from the invention process, to the production line, and then to be widely recognized by the industry is steadily improving. However, the window period for technology to retain an acquisition value is becoming shorter. A prime example of this is Prisma, a popular software which utilizes neural style-transfer technology. This concept was first introduced in August 2015 by a paper, “A Neural Algorithm of Artistic Style.” Within 3-4 months, the Open Source Software Community was able to enhance the algorithm speed by several orders of magnitude.
In the first half of 2016, only 2-3 start-ups had possession of this technology. Since Prisma launched their product in July 2016, they influenced more than 20 teams to start their own version of this same business, and this became the standard used in photo-editing and photo-beautifying. The problem is that the acquisition value of this technology will experience a dramatic decrease due to the proliferation of the technology, and so over-valued start-ups may find themselves in an awkward position if they were over-valued before they managed to find an exit.
Technology Maturation Cycle
Start-ups in the cutting-edge science and technology sector need to respect commercial guidelines, search for value creation sources, and understand the maturation cycle of technology. Every year, Gartner announces the Hype Cycle for Emerging Technologies. Using this, we encourage entrepreneurs to circumspectly evaluate their respective strengths and search for emerging technologies that is closer to commercialization. All great enterprises start by satisfying customer demands instead of relying on attaining a world-class academic status.
Giants in the AI sector are actively expanding, and platform-type technologies will soon become commercialized. For example, the framework of Google TensorFlow is now being applied by a number of AI companies, just like the popularization of Hadoop three years ago. However, it is impossible to use this framework to create AlphaGo, even if it is enough for analyzing offline business data.
AI Company Characteristics which Source Code Views Positively
Here at Source Code Capital, we view some characteristics of AI positively. We believe that data is production data, and calculations are productivity. Through surveying the entire industry chain, we are more optimistic about the application-layer of the business solutions. We also find value creation points in verticals, where companies can take advantage of AI technologies and use it to improve business efficiency, or meet new demands that can only be realized by reliance on AI. At the same time, we are quite optimistic about companies that are able to accumulate vast amounts of self-owned data. These companies may be able to harness their original data and create new business lines or more value through AI tech down the line.
Source: NetEase and Wuzhen Institute, summarized by Source Code Capital
At Source Code Capital, we believe the following AI+ industries should see many interesting investment opportunities emerge soon: information distribution, finance, healthcare, education, security, and logistics, to name a few verticals.
We believe that AI opportunities are forming their own “investment track,” and will spread across various applications and domains.
Source Code Capital believes that AI technology is one of the most exciting and transformative opportunities of our time, and within 2-3 years, AI will become a type of independent investment field. We think AI can become like a “basic capability,” as in how electricity was in the second industrial revolution, and that AI can be a base from which countless new business opportunities and products can be created across industries.
Profile of a “Promising” AI Start-up
- It starts with a technology-based venture, with technology as its entry point. After identifying its advantageous edge, it can be applied to a specific industry.
- It has a solid entry point, with an executive management team who possess industry know-how.
- It has awareness of the need to secure a favourable position through data computing.
- It has a reliable algorithm computing platform and a good price-performance ratio. This means that basic algorithms are up to date, thus allowing for the founder to solve engineering problems.
- 2B start-ups make up the majority, while 2C’s are the minority. Revenue generation is predominately project-based, and initially, they operate as subcontractors. Enabling traditional intermediaries, integrators and agents are some good business models in this industry track.