My husband’s small fund invested in a computing power IAAS company 7 years ago (oh my god, there was no way to exit and all my family’s money was locked up), so I have been consciously paying attention to Singapore’s AI industry since last year.

IAAS (infra as a service), to put it simply, is a CPU+GPU server center or IDC-Internet Data Center.

After several months of research, we concluded that Singapore’s AI industry is far from mature:
1. There are only a few companies training large models
2. Local AI application companies and their corporate clients for toB services are not in Singapore
3. The toC business of Singapore’s AI industry is still dominated by big companies such as OpenAI
4. Many Chinese AI companies have expanded to Singapore, but they only use Singapore as a springboard, and this springboard can only be used for a while.

The causes are:
1. Talent bottleneck: A professor at Nanyang Technological University said that one of his undergraduate students majoring in cloud computing, who will graduate in 2024, was recruited by an American AI company with a monthly salary of 16,000 SGD. It is difficult to find talent in AI large model training or AI application companies.
2. With few local corporate clients, it is extremely difficult for AI companies to expand sales locally. They must go overseas to countries and regions outside of Singapore, and when they go overseas, they face business sales and customer service operating costs in other countries.

In addition, AI computing centers also face these problems when developing their business in Singapore:

1. Although AI computing power is scarce, it is probably in excess if it is used by local AI companies. AI computing companies must also expand to customers outside of Singapore (mainly Europe, the United States, and Japan)
2. The cost of operating a computing center in Singapore is extremely high (building server buildings, loans and electricity supply are strictly controlled by the government with licenses). There is no need to place the hardware of the computing center in Singapore. If you open a company in Singapore, you still have to go to Malaysia, Indonesia, Thailand and other places to build infrastructure (so if you go to work, you have to run around the mountains and ditches all the time, haha)
3. The infrastructure of AI computing power should be close to toB customers. If the customers are not in Singapore, then in the long run, the computing power center will not be in Singapore, but in the United States, Europe, Japan, and maybe Hong Kong (Hong Kong also depends on Sino-US relations. Currently, as in China, Nvidia cannot sell H100 chips to Hong Kong companies)
4. In the long run, AI computing companies will still have to go to the United States, Japan, and Europe to find customers and build infrastructure. The Singapore office will become a discarded child or useless (the boss and his family moved to Singapore mainly to take care of the children's education, and doing business in Singapore is secondary)

As for Web3 AI, it is still in the narrative stage, and its practical application remains to be seen. However, due to the influx of a large amount of funds, the idle computing power of Web2 has made a lot of money from the mining projects of these Web3AI companies.

https://mp.weixin.qq.com/s?__biz=MTQzMjE1NjQwMQ==&mid=2656018583&idx=1&sn=4df7edb5c5d7b9383a53cc9e2313d3fe&chksm=66d80f0951af861fe3e897bdf036dbc169f669a8cfaf236673f7c4d683e21e61b235bc943e7f&mpshare=1&scene=2&srcid=0630DQOeyLc8epenexE9GhTW&sharer_shareinfo=8e439af13f7ab157f9870ef59d528219&sharer_shareinfo_first=8e439af13f7ab157f9870ef59d528219#rd…