In a report published July 9, S&P said the number of AI companies founded in the U.S. between 2013 and 2023 totaled 5,509. That’s more than all the other countries combined and nearly four times more than second-placed China, which saw the creation of 1,446 in that span. Rounding out the top five are the U.K. with 727, Israel with 442, and Canada with 397.

AI startup activity was also reflected in private-sector investment. The U.S. saw cumulative investment of $335.2 billion during that decade, according to S&P Global.

That’s more than triple the $103.7 billion from China’s private sector, and it dwarfs the U.K.’s $22.3 billion, Israel’s $12.8 billion, and Canada’s $10.6 billion.

But S&P also cautioned that governments play a big role in AI investments, noting that much of China’s spending on AI will come from the public sector and won’t be as transparent.

In fact, China has a three-year AI development action plan, which includes programs for creating a domestic AI-skilled labor force and attracting foreign workers for AI projects, the report added.

Worldwide, S&P estimated that private investments in AI startups could reach $800 billion-$900 billion by 2027, representing a compound annual growth rate of at least 70%-74%.

The data on the robust U.S. private-sector AI industry comes as the Chinese government is censoring AI developers.

Sources told the Financial Times this past week that the country’s top internet regulator, Cyberspace Administration of China (CAC), is testing companies’ large language models to ensure they “embody core socialist values.”

The CAC has a “special team doing this, they came to our office and sat in our conference room to do the audit,” an employee at a Hangzhou-based AI company told the FT. “We didn’t pass the first time; the reason wasn’t very clear so we had to go and talk to our peers. It takes a bit of guessing and adjusting. We passed the second time but the whole process took months.”

Ensuring “core socialist values” also means companies must monitor sensitive questions or keywords that could contribute to “inciting the subversion of state power” or “undermining national unity,” the FT added.

The result is that queries related to the Tiananmen Square massacre or even most questions about President Xi Jinping are rejected by Chinese chatbots.

The Financial Times previously said that a research center reporting to the CAC is developing a large language model trained on “Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era.”