Speaking at the 36kr WAVES 2026 conference in Guangzhou, FreeS Capital founding partner Li Feng framed AI valuations as downstream of central bank balance sheets, with global money now rotating between assets rather than growing.
The most honest way to read the AI rally in mid-2026 is not through model release notes or benchmark leaderboards. It is through the central bank balance sheets that funded it. FreeS Capital founding partner Li Feng, opening the 36kr WAVES 2026 conference in Guangzhou with a keynote on capital and liquidity, argued that the AI boom was always a liquidity trade layered on a real technology cycle, and that the liquidity half is now the binding constraint on what AI market caps can do next.
His mechanism is unfashionable in tech press coverage, because it treats AI valuations as downstream of money rather than ideas. From 2020 onward, pandemic-era central bank expansion added roughly US$12 trillion in base money globally. Bank-dominated systems like China's multiply that base money at roughly 6x to 8x into broad money; the US runs a roughly 3x to 4x multiplier. Using a conservative global multiplier near 4, Li Feng pegs broad money added at roughly US$40 trillion to US$50 trillion since 2020, a figure he described as an unprecedented expansion in human financial history.
The translation into AI assets happened in two steps. First, when US, European, and Chinese assets each became unattractive around 2022, the global liquidity pool concentrated into US dollar assets. Second, the arrival of ChatGPT in late 2022 provided the narrative lift that turned that parked liquidity into an AI rally. The technology was real; the funding was mechanical.
What changes now, in Li Feng's read, is the trajectory of the underlying pool. In April-May 2026, he sees liquidity rotating back toward dollar assets in a zero-sum seesaw pattern across global assets. Total global liquidity is no longer expanding because no major central bank is reflating. That distinction, growth versus rotation, is the entire bet. The same capital is chasing the same AI names; new capital is not arriving to widen the bid.
The valuation context sharpens the stakes. Li Feng places US equity market capitalization at roughly US$80 trillion against US GDP near US$30 trillion, a Buffett Indicator reading of 2.3 to 2.4, meaning US stocks are worth more than twice US annual output, a level historically associated with late-cycle excess. US markets now account for over 60% of global equity market cap, according to his figures. China's Buffett Indicator sits below 1 by his reading, which leaves room for Chinese AI assets to reprice if global liquidity ever tilts back toward them.
The implication for AI funding is direct. The AI boom in Li Feng's framing is a liquidity phenomenon riding a genuine new technology cycle, and the same logic dictates whether AI market caps and burn rates stay funded. If total liquidity is flat and rotating, then incremental AI funding has to come from rotation, from one asset into another, rather than from a rising tide. That is the difference between an expanding market and a zero-sum reallocation, and it changes how an investor should read an AI earnings beat or a Chinese model release in the second half of 2026.
FreeS's own deployment pattern illustrates the framing. Reporting from letschuhai.com describes FreeS and a Zhipu ecosystem fund co-investing in a humanoid robotics company, a deal that reads as capital chasing the AI-and-robotics intersection rather than pure technology conviction. FreeS's 36kr PitchHub profile confirms the firm as active across AI and hard tech, and a separate 36kr piece on Anthropic, noting that 80% of code merged at the lab was produced by Claude, is a useful reminder that the technology side keeps moving even as the capital conversation dominates the agenda. The Anthropic reporting is adjacent context, not a load-bearing data point on valuations.
The honest caveats matter. Li Feng is one prominent China VC speaking at an industry-organized conference, and the speech was edited and organized by 36kr rather than delivered verbatim. The monetary figures, the roughly US$12 trillion base money, the roughly 4 global multiplier, the roughly US$40 trillion to US$50 trillion broad money total, are his round numbers, not independent audit results. The Buffett Indicator readings are his own. The "we are in the back half" call and the April-May 2026 seesaw reading are predictive VC opinion, not a market call. None of that invalidates the framework, but it tells the reader how much weight to put on each line item.
The watch items for the rest of 2026 fall out of the framework cleanly. First, whether any major central bank (Federal Reserve, European Central Bank, People's Bank of China, Bank of Japan) restarts balance sheet expansion. That is the leading indicator for incremental AI funding, not earnings season. Second, whether the seesaw rotation tilts back toward China. A Buffett Indicator below 1 is mechanically room for repricing if the liquidity ever points that way. Third, whether AI capex commitments continue to rise against a flat liquidity backdrop, which would force a rotation trade inside US tech rather than a new inflow. The technology leg keeps moving on its own clock; the money leg is now what determines which AI names stay funded.