The 2025 enterprise-software IPO wave: do first, talk later, and PS back below 10x
Over the past two or three years the entire enterprise services market was overvalued by 50% to 100%, with PS falling from 30-40x to below 10x. Beisen's IPO was a "bargain-basement IPO," but it was the first shot in the systematic listing of Chinese SaaS, and a wave of enterprise-software IPOs will come in 2025.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Beisen's IPO is the first shot in the systematic listing of enterprise software
Xiong Fei reads the signal from Beisen's listing as this: Chinese enterprise services companies now include excellent companies listing systematically in Hong Kong or on domestic capital markets. He retraces the past two years as a rollercoaster — valuations slid but company quality did not change; the distortion came from the external environment. In fiscal 2020 Beisen could have been cash-flow profitable, to the tune of several tens of millions, but after the pandemic the US dollar flood inflated the whole sector's valuations three- to fivefold, and at least six to eight companies in the same sector each raised tens of millions of dollars to spend more heavily, changing prices, costs and per-person efficiency enormously — a prisoner's dilemma for every company. Going public is combat: through listing you rebuild per-person efficiency and get gravity pointing upward all the way.
— Xiong FeiBring growth down, bring quality up
Xiong Fei observed that over the past two or three years the vast majority of companies chased 100%, 150% or even 200% growth, and what that bought was doubled headcount and doubled losses — radishes happy, mud unkneaded. Now most of the companies he is involved with are working toward single-month break-even by the end of this year and full-year break-even next year, by cutting growth to 60%, adding only 10% to 20% more people, and lifting per-person efficiency 40% to 50% a year. He cites Stripe: profitable before the pandemic, then valued at nearly $100 billion after it, quickly becoming a company losing several hundred million dollars a year with per-person efficiency down 60% — about 2,000 people before the pandemic, close to 10,000 after, and now it has optimized away a great many people.
— Xiong FeiThe 2025 IPO wave: 300-500 million in revenue plus profitability
Xiong Fei says enterprise services companies are now much closer to listing than before. More than six to eight of the companies he has invested in have revenue of 100-200 million or even several hundred million, and there are still plenty of companies with tens of millions in revenue. He and top-tier investment banks share a consensus: there will be a wave of enterprise-software IPOs in 2025. The reason is that these companies are at a scale like Beisen's, perhaps 700 million in revenue, but more are at 100-200 million to 200-300 million, and after two more years of growth to 300-500 million or close to 500 million, both Hong Kong and A-shares are very OK; second, they will have built profitability through 2023 and 2024. On domestic listing standards, application software mostly goes to ChiNext, and hitting 30-50 million or 40-50 million in net profit for the year with 30-40% or around 40% growth is the ChiNext consensus.
— Xiong FeiThe enterprise services market was overvalued by 50% to 100%
Xiong Fei says very candidly that over the past two or three years the entire enterprise services market was priced at a premium — even though his own seat is in this sector, he believes it was overvalued by 50%, 100%. At the hottest point in 2021, many companies had only 20-30 million or 30-50 million in revenue but valuations of nearly a billion, a 30-40x PS, or 20-30x PS. He judges that in the second half of this year or the first half of next year, industry-wide PS will fall back below 10x, and excellent companies in the primary market can return to 6-8x. At the same time, most companies have low visibility, but excellent companies will hit single-month or full-year profitability this year and next, and by the first half of next year at the latest we will see a batch of companies that are fully profitable for the year, still growing in the high tens of percent around Series B, and around 40% from Series B/C onward — and then PS becomes reasonable.
— Xiong FeiAirbnb and Uber improved markedly after going public
Xiong Fei believes going public is a very correct, very wise decision. He cites Airbnb choosing to IPO in its worst moment, in 2020, holding itself to public-company standards, and its financial position is in a fairly OK position among the new generation of unicorns; conversely, other companies that did not IPO simply did not set that bar for themselves. He says that when a company has 300-500 million in revenue, whether it goes to A-shares or Hong Kong, it is already a qualified company, and holding itself to qualified-public-company standards increases its influence, is healthy and effective for the company, and lets it put down its baggage and go at its own pace. He observes that after Airbnb and Uber listed, the whole company's economics, health and operational refinement improved markedly versus before listing, because there is an invisible hand — or a real hand — making adjustments.
— Xiong FeiVersus 2015-17: real money, closer to capital, consolidation
Xiong Fei compares 2015-17 with now and sees three differences. First, in 2015-17 much of it was concept; back then the first wave of foreign enterprise services, Jetforce and Workday, were valued at only a few billion or one or two hundred billion dollars, and many Chinese companies had revenue of maybe tens of millions of RMB, whereas now there are plenty of companies with 100-200 million to several hundred million in revenue. Second, these companies are much closer to the capital markets and can be tested by them. Third, the market is entering a trend of consolidation: 2015-17 was a great explosion of Han weapons, while 2020-22, especially from 2022 onward, consolidation in sectors like e-commerce, HR, CRM and electronic contracts has been very severe and clear, essentially shrinking to competition among the top two or three, with very few opportunities left for new companies.
— Xiong FeiMidjourney may be built on top of Stable Diffusion
Xiong Fei discusses Midjourney and Stable Diffusion: Stable Diffusion can build its own model and launched DreamStudio, with an interface like Adobe Photoshop; Midjourney is like Meitu Xiuxiu, extremely easy to pick up. He thinks there is also a professional market, where professional users are willing to spend a few days setting up Stable Diffusion themselves and training the model they want, turning it into their own digital asset — controllable, so that next time they draw the same character it will come out exactly as expected. He mentions that outside material speculates Midjourney is built on top of Stable Diffusion, because Stable Diffusion has not disclosed at all how its underlying layer is trained, Midjourney has very few people, only ten or twenty, and outside observers guess it simply cannot afford to build its own underlying model, and Midjourney's new versions and Stable Diffusion's new versions update similarly.
— Xiong FeiThe iPhone moment has arrived; the Android moment has not
Xiong Fei believes ChatGPT will exist long-term, like the Mac or the iPhone — closed-loop, independent, with a significant advantage — but that cheap, open-source AI will certainly emerge, just like Android. He believes the early versions will certainly be very wrong; right now Dolly 2.0 and Stable LM are out and everyone complains about the gap, but one is 175 billion parameters and the others are 7 billion and 3 billion, just as the first generation of Android was also very wrong. He gives an example: if OpenAI does $10 billion a year in revenue, then leaving OpenAI to start an open-source company is like Android — very profitable, because the downside is nothing to lose; open-sourcing a pre-chain might be 0.3x as good as OpenAI, or even 0.2x as good, but 0.2x or 0.3x is already usable in a great many enterprises, because enterprises have sensitive data they do not want sent to OpenAI and want their own model.
— Xiong FeiIn their own words · checked verbatim
Among the companies I'm involved with, most are working toward single-month break-even by the end of this year and full-year break-even next year.
我参与的公司里 大部分大家都在做 今年年底单月盈亏平衡 明年全年盈亏平衡
Xiong Fei5:04
To be very candid, over the past two to three years the entire enterprise services market was priced at a premium.
非常坦诚的说 在过去的两到三年 整个企业服务市场是被溢价的
Xiong Fei13:56
I think basically, in the second half of this year or the first half of next year, industry-wide PS will fall back below 10x.
我觉得基本上 今年下半年或明年上半年 全行业的PS就会回到十倍以下
Xiong Fei14:48
I think the whole company's economics, health and operational refinement improved markedly versus before listing, because there is an invisible hand — or a real hand — making adjustments.
我觉得整个公司的经济性 健康性经营的经济化程度 比上市之前要显著优化 因为这是有一个隐形的手 或者是一个真实的手 在去做调整
Xiong Fei17:30
So people are animals of their environment. When the environment changes, now when we ask most top-tier companies, all of them feel they can do it.
所以人是环境的动物 当环境变化了 现在我们去问 大部分一头公司 所有公司都觉得可以做到
Xiong Fei26:14
The iPhone moment has already arrived; the Android moment has not yet appeared.
现在已经有了iPhone时刻 还没有安卓时刻出现
Xiong Fei1:01:49
This venture capital industry now has a gravity effect. No matter which sector is rising now, that gravity means in the end you still look at the business, at revenue, at monetization.
现在这个创投行业 它就是有重力效应 不管是哪个行业现在起来 这个重力就是最后还是要看商业 看收入 看变现
Xiong Fei1:10:55
Figures
| Beisen fiscal 2020 cash-flow profit | several tens of millions | 2:02 |
| Number of companies in the same sector that raised tens of millions of dollars | at least six to eight | 2:02 |
| Stripe headcount before the pandemic | 2,000 | 4:03 |
| Stripe headcount after the pandemic | close to 10,000 | 4:03 |
| Stripe per-person efficiency decline | 60% | 4:03 |
| Number of enterprise services companies Xiong Fei invested in | more than 6 to 8 | 7:04 |
| Revenue scale of enterprise services companies | 100-200 million to several hundred million | 7:04 |
| ChiNext net profit standard | 30-50 million or 40-50 million | 9:15 |
| ChiNext growth requirement | 30-40% or around 40% | 9:15 |
Glossary
- PS / price-to-sales ratio
- Market value divided by revenue, a measure of how high or low a valuation is.
- NDR / net dollar retention
- The share of revenue retained from existing customers through renewals and upsells.
- PMF / product-market fit
- A product finding a market that genuinely needs it.
- fine tune
- Continuing to train an existing large model on your own data.
- pre-chain / pretrained model
- The product of a large model's base training stage.
How to listen
Founders, investors and engineers watching enterprise services, SaaS and AI infrastructure — especially anyone weighing the enterprise-software IPO window and the AI open-source inflection point.
31:45 to 38:26 covers the investor's career transition and personal choices; low information density.