China AI data centers need profits as borrowing costs climb
China's AI data center industry is growing fast, but actually making money is proving tough as borrowing costs climb.
According to a new Goldman Sachs report, investors are now more interested in seeing real profits than just bigger facilities.
Data center operators need to show they can generate returns that exceed the cost of capital.
AI token factories offer highest returns
Goldman Sachs compared three types of AI data centers: traditional "powered-shell" buildings (steady but low returns), integrated centers offering GPU computing (potentially better at around 15% returns), and the standout: AI token factories.
These token factories, which make money by selling computing power for AI tasks, could bring in around 21% returns for an asset-light token factory and around 43% for one owning its GPU infrastructure.
The report sums it up: smart monetization and using capital wisely are key if this industry wants to keep growing.