Li Bei, founder of Banxia Investment, said in a recent dialogue that the opportunity in real estate has been upgraded from a once-in-a-decade event to a once-in-twenty-years event. She noted that the new policy on completed-home sales has caused a significant short-term negative impact on real estate stock prices, but the level of the opportunity has actually increased, and the duration and height of the subsequent real estate bull market will be extended. In her view, under the influence of the new real estate rules, the industry will undergo a more thorough shakeout over the next one to two years, with more companies continuing to exit; in the long run, entry barriers will rise, the gap between companies will widen, and a small number of leading developers with both product strength and financing advantages may actually become beneficiaries of the new rules, with some companies' share prices potentially rising five to ten times over the next few years.

The new policy on completed-home sales is the key variable in her judgment of changes in the industry landscape. For homebuyers, purchasing a completed home allows them to defer down payments and monthly mortgage payments while avoiding risks such as unfinished projects and uncertain delivery quality. Based on this, she estimates that completed homes have room to raise prices by about 3% to 5% compared with pre-sale homes, developers can share part of the premium, and buyers can also benefit from lower capital costs and reduced risks. On the developer side, the market generally worries that completed-home sales will lengthen turnover cycles and squeeze profits. But her calculations show that if land payments can be made in installments, project capital turnover time may only increase by 20% to 30%, and combined with a slight increase in home prices and a decline in land prices, the net profit margin may actually rise by 20% to 30%. From a longer-term perspective, she expects industry concentration to increase substantially. The proportion of companies that ultimately exit may rise from 90% to 95%, or even 98%, and the number of companies capable of continuously acquiring land and expanding nationwide in the future may be no more than a double-digit figure. At the same time, land payment requirements, sales fund supervision, and the main bank system have raised barriers for new entrants, making financing capacity and major shareholders' guarantee capacity core competitiveness. Li Bei believes that given the gap in financing costs, the advantage of central state-owned enterprises over private enterprises will further expand. Even after the industry reaches a high level of prosperity in a few years, new players will not easily consider entering, and a highly oligopolistic landscape may persist for a long time.

Beyond real estate, Li Bei also flagged the risk of an AI bubble. She judges that although overseas AI capital expenditure is still growing, its quarter-on-quarter growth rate may peak in the middle of next year. Previous increases in investment by cloud vendors were based on the assumption of rapid linear growth in AI model revenue; however, revenue growth at model companies has slowed significantly, and the correction in AI upstream-related companies after July mainly reflects valuation contraction rather than an immediate downward revision in earnings expectations. In her view, the "second wave of decline" in the AI sector may only appear when capital expenditure truly peaks and profits begin to fall. She judges that when the AI "boom" fades, the U.S. economy declines, U.S. Treasury yields fall, and the dollar begins to depreciate, China's real estate and consumption may instead become a "desert oasis" for global assets. She believes consumption has already gone through five to six years of adjustment, and the steepest downward phase in households' marginal propensity to consume may be over. Next year, even if the economy takes another step down due to declining real estate investment and exports, consumption will remain relatively stable; many leading companies in essential consumption and some discretionary consumption are valued at historical lows, and combined with dividends and buybacks, some companies' static returns are already attractive, making Chinese consumer stocks "gold everywhere." When AI capital expenditure cools and domestic demand policies intensify, these companies are expected to enter a main upward trend.

Real Estate Opportunity: From Once in a Decade to Once in Twenty Years

Tencent Finance: How do you view the impact of the new completed-home sales policy on real estate? Li Bei: Previously, it might have been 90% of companies being eliminated and 10% remaining; now it may be 95% eliminated with only 5% left. Completed-home sales will further intensify the reshuffling, and this must be viewed from both long-term and short-to-medium-term perspectives. First, looking at the stage of the industry, real estate is now at the end of a down cycle and in a phase of bottoming and differentiation. In fact, the new home market has been improving since the second half of 2024, and quality companies' new projects can almost all achieve a net profit margin of 10%. Compared with second-hand homes, fourth-generation new homes have higher usable area ratios, better design and quality, and even if they are more expensive, buyers are willing to purchase them, with many projects selling out on launch day. Second-hand homes are convenient, and the national index is still drifting lower, but the decline has narrowed significantly compared with last year. Shanghai has already begun to rise, and more than a dozen second-tier cities are also stabilizing and recovering. At the company level, 90% of companies have already been eliminated, and new home supply has fallen sharply and is still falling. New construction starts are only about one-quarter of the peak, and new project launches are about one-fifth of the peak, still down 20% to 30% year on year. However, some surviving companies with strong product strength and financing capacity have actually seen significant improvement in operations over the past two years, though this is not yet visible in their financial statements due to lag. This is because financial statements lag sales by two years, and some legacy burdens are still being provisioned for impairment. What happens after the new policy? Many people believe completed-home sales will greatly lengthen developers' turnover cycles and raise financial expense ratios, thereby sharply lowering ROE. I do not think so. On the one hand, home prices can rise; on the other hand, land prices will fall, and land payments can most likely be made in installments. First, home prices. If you buy a pre-sale home, you need to pay the down payment now and make mortgage payments for two years, with delivery only after two years; if you buy a completed home, you only need to pay a 5% deposit now, and pay the down payment, arrange a loan, and start monthly payments two years later. For the same house, most people will choose the latter. Therefore, even if a completed home is 4% to 5% more expensive, based on a mortgage cost of 3% per year, or 6% over two years, it is still more cost-effective than a pre-sale home, and there is no risk of unfinished projects or delivery uncertainty. For the same property, completed homes have room to raise prices by 3% to 5% relative to pre-sale homes, and consumers are willing to accept this. In addition, land prices will also be pushed down. Several land auctions after the new policy showed that, under the same conditions, if installment payments are not allowed, land prices are commonly about 30% lower than comparable plots before the new policy; if installments are allowed, it is equivalent to local governments bearing part of the capital turnover pressure. When the new policy was first introduced, I judged there would be two changes: first, home prices might rise; second, installment land payments would most likely appear. Both have now been confirmed: two cities have issued detailed rules for land installment payments, and dozens of projects nationwide have canceled discount offers, equivalent to implicit price increases. The root reason the pressure of the new policy can be passed on to consumers and local governments is that developers' bargaining power has strengthened, whether toward homebuyers or land sellers. For consumers, developers that can build good homes, avoid unfinished projects, and gain recognition for quality are becoming increasingly scarce; for local governments, the number of companies able to acquire land is already limited and will be even smaller in the future. Looking at the steady-state new financial model one year from now, for a leading advantage developer: assuming 50% of land payment is made upfront and the remainder is paid after completed-home sales, capital turnover is extended by only 20% to 30%. With moderate support from the group and a modest increase in leverage, sales scale can be maintained. If home prices rise about 3%, land prices fall 10% to 20%, and the net profit margin may instead increase by 20% to 30%, then even looking only at the short to medium term, corporate profits will not decline but rise. In the long run, there are three outcomes. First, the shakeout will be more thorough. The proportion of eliminated companies may rise from 90% to 95%, or even 98%. I judge that no more than a double-digit number of companies will be able to continuously acquire land and expand nationwide in the future. Second, barriers for new entrants will rise substantially. Any industry with high prosperity and high profitability will attract new players, but the new policy requires developers to pay land payments in a lump sum, or at least half first; at the same time, fund supervision and the main bank system are implemented, and before project completion, sales funds must remain in supervised accounts. Even if land payments can be made in installments, funds will still be occupied for a long time. The old model of acquiring land with a small amount of own funds and then rolling through mortgage loans is basically no longer viable. Although some local rules leave openings, they mainly benefit state-owned enterprises backed by large groups. For example, Wuhan's detailed rules allow a minimum 10% down payment for land, and when supervised funds exceed project needs, they can be withdrawn with a group guarantee. But basically only large state-owned enterprises, especially central SOEs, can provide such guarantees; private enterprises find it very difficult. Third, financing capacity becomes core competitiveness. As capital occupation increases, financing capacity is crucial, and the gap between private enterprises and central SOEs is enormous. Even for mixed-ownership enterprises, financing costs may be nearly 100 basis points higher than those of central SOEs, leaving almost no competitive advantage. In the future, even if the industry becomes profitable again, when leading central SOEs achieve a net profit margin of 15%, private enterprises may only have 5% to 6%. Therefore, the threshold for new companies to enter the industry will be greatly raised, and a highly oligopolistic landscape may persist for a long time.

AI Is Currently Falling on Valuation; the Second Wave Is Still Ahead

Tencent Finance: Since the beginning of this year, the capital market has shown a "K-shaped divergence," with technology and new productive forces sectors and traditional sectors experiencing completely different conditions. However, the technology sector has recently pulled back. There is considerable market debate over whether AI is a bubble. How do you view this situation? Li Bei: AI capital expenditure has indeed grown rapidly this year and will continue to grow next year, but from a quarter-on-quarter perspective, it may peak around the second quarter, and even if not in the second quarter, it could be in the third or fourth quarter. This round of increased capital expenditure by cloud vendors mainly stems from AI models succeeding in the coding field and Anthropic's ARR rising rapidly. After this happened in the first quarter, these cloud vendors sharply raised capital expenditure in the second quarter. The assumption at the time was linear extrapolation: based on the first-quarter trend, total model revenue by the end of this year could be around 500 billion (with Anthropic alone accounting for 300 billion), and could exceed 1 trillion next year or the year after. If revenue really reaches more than 1 trillion, then it matches the current annual capital expenditure of more than 1 trillion, and it is not a bubble but reasonable. But in fact, ARR growth slowed significantly in the second quarter. I repeatedly warned of this risk in May and June, predicting that ARR would slow substantially, and this has indeed been confirmed. This is precisely the reason for this round of market adjustment. From April to June, the market sharply raised expectations for 2027 earnings because capital expenditure was increased; at the same time, valuations based on 2027 were also higher because everyone believed this was a sustainable level and could continue to grow. So this decline is not a "Davis double kill"; what is falling is valuation: the market still recognizes substantial profit growth in 2027, but if ARR cannot rise, the investment level is unsustainable, and capital expenditure is very likely to peak in 2027, or if not in 2027, then in 2028. If end-demand revenue cannot rise, industrial chain investment cannot be maintained, and high valuations cannot be justified. My view is that the second decline in AI will have to wait until capital expenditure truly peaks, possibly in the middle of next year. By then, when everyone sees profit expectations beginning to decline, AI will fall in a second wave.

By This Time Next Year, Real Estate and Consumption May Instead Become a "Desert Oasis"

Tencent Finance: What is the logic behind judging that there are investment opportunities in consumption and real estate? Li Bei: Some people will ask, if the economy is very bad, why would consumption and real estate instead have opportunities? Because their profits will no longer decline. Consumption has actually experienced five to six years of decline, and now consumers' mindset has stabilized, and the steepest downward phase in the marginal propensity to consume has passed. Previously, due to the pandemic and the sharp drop in home prices, people engaged in some panic saving, and their propensity to consume fell very low. After experiencing phased shocks, everyone's mindset began to stabilize. After the rapid adjustment period passed, the entire industry also restabilized. Essential consumer goods are inherently less affected by economic fluctuations, and the key is that these essential consumer goods and some discretionary consumer goods are now valued at historical lows. Let me give two examples: as an essential consumer goods company, the leading dairy company's performance fluctuations are inherently relatively small. In 2021, when consumption was sought after, it traded at 40 to 50 times PE, and now it is only 10 times. Even more exaggerated are the leading frozen food company and the leading condiment company. In 2021, both had triple-digit PE ratios, similar to today's tech bubble, and now both have returned to just over 10 times, both at particularly low historical valuation levels. Moreover, these companies also pay dividends. For example, the company that makes konjac snacks has a dividend yield of more than 7%. On average, the dividend yield plus buybacks of such companies can provide a static return of about 4% to 7%. So they are in fact extremely undervalued, except that the current atmosphere is all focused on AI. Once the AI narrative collapses, or once domestic demand policies are introduced, everyone will discover that gold is everywhere here, and the market trend will turn. So I think one can absolutely stay in this sector now and wait for the wind to come; it is already too cheap, and it is not sensitive to the economy. Real estate follows another logic. Previously, developers sold pre-sale homes, which could only enter the financial statements after completion and settlement, so this year's profits reflect sales from two years ago. As mentioned earlier, new home sales began to improve in the second half of 2024, and the homes these companies sold already had a net profit margin of 10 points, which corresponds exactly to the financial statements in the second half of this year; the financial statements in the first half of this year correspond to the first half of 2024, which was precisely the worst period for new home sales. In other words, by the second half of this year, the profits of the leading companies will improve significantly, and in the first half of next year they will rise by another step. Now some leading companies may show a development business net profit margin of only one or two points, or two or three points, in their statements. By this time next year, you may find it at five or six points, or six or seven points, and by this time the year after next, it may be seven or eight points or even around ten points. This comes partly from the time lag between settlement and sales, and partly from the clearing of impairment burdens. Over the past two years, they have continuously provisioned for asset impairments to deal with the historical burden of previous price declines. Last year was the peak of this process, some companies have already begun to see less this year, and next year basically everyone will significantly reduce it. So by this time next year, the reported performance of real estate companies will reflect 2025 sales, and sales in the first half of 2025 were actually quite good. You will find that real estate and consumption have instead become a "desert oasis."