SINGAPORE - Over the past 100 years, the companies that have commercialized the most new technologies have tended to be startups, e.g., personal computers, integrated circuits, enterprise software, and e-commerce services were all predominantly commercialized by startups. Even when incumbents have commercialized products, it has often been startups who have ended up becoming the dominant companies. 

On the surface, artificial intelligence (AI) is following the same trend. A large number of AI startups were founded in the 2010s and venture capital (VC) funding of AI startups set records in 2020 and 2021. These data suggest AI startups are doing quite well, and that AI is also doing very well. 

Below the surface, however, a more nuanced picture of the situation emerges. Only two publicly traded startups, SoundHound and c3.AI, can be truly defined as AI firms, and neither are profitable or have large market capitalizations (See Table 1). 

Table 1 lists many other startups that use algorithms, and many of these claim to use AI, thus implying that their algorithms use AI. Many readers will assert that these are not true AI startups, but with so few publicly traded AI startups, one must define this liberally to talk about it. This small number is interesting in and of itself too. Given the record amounts of funding and such hype surrounding AI, one would have expected many more startups to have gone public. 

Table 1. America’s Publicly Traded AI/Algorithmic Unicorn Startups