For more than a decade, one of the easiest investment decisions was also one of the most profitable: own the United States, particularly its largest technology companies, and allow American exceptionalism to do the rest. The combination of deep capital markets, extraordinary companies, a strong entrepreneurial ecosystem and ultimately the artificial-intelligence boom made the US market difficult to compete with.
The problem is not that this story has suddenly become false. The problem is that a very good story can become an expensive story, while an uncomfortable one can become sufficiently cheap to deserve another look.
As of 31 August 2026, the MSCI USA Index traded at about 25.8 times trailing earnings, 20.1 times expected earnings and 5.7 times book value. The MSCI China Index stood at approximately 14.0 times trailing earnings, 10.8 times forward earnings and 1.45 times book value. In other words, Chinese equities were trading at roughly half the US earnings multiple and about one quarter of the US price-to-book multiple.
That is an enormous difference. It is not, however, enough to say “America is expensive, China is cheap, therefore buy China.” The Chinese discount includes real economic, political, regulatory and geopolitical risks. Conversely, paying a high multiple for a company capable of compounding earnings and capital at exceptional rates can still be rational.
The more useful question for a long-term investor is therefore broader:
Has the combination of extreme enthusiasm around American AI, very different valuations and China’s growing economic and technological capabilities created a case for diversifying some of the portfolio away from the most popular US assets and toward the assets that investors have spent years avoiding?
Our answer is increasingly yes—but as a diversification argument, not as a prediction of American decline.
1. The US problem: the businesses may be excellent while the expectations become demanding
It is easy to caricature concerns over American valuations as another attempt to call the top of a market that has repeatedly confounded pessimists. That would be a mistake.
The strongest version of the bearish argument is not that artificial intelligence is fake. It is that an enormous amount of capital is now being invested on the assumption that AI revenues, margins and competitive advantages will eventually justify today’s expenditure and valuations.
The MSCI USA Index’s largest ten positions accounted for almost 37% of the index at the end of August, led by Nvidia, Apple and Microsoft. Technology alone represented 37.9% of the S&P 500.
Concentration by itself is not evidence of a bubble. Successful companies naturally become large. But it does mean that what happens within a relatively small number of businesses increasingly matters for investors who believe that they own a broadly diversified American index.
Eisman’s Achilles’ heel
Steve Eisman, remembered for his successful bet against subprime mortgages before the Global Financial Crisis, has recently focused on one potential vulnerability in the AI ecosystem: its dependence on a few frontier laboratories.
In August 2026, Eisman estimated that OpenAI and Anthropic together represented roughly 70% of AI-related revenues at Microsoft, Amazon, Google and Oracle and as much as 25–35% of those companies’ cloud revenues. These are Eisman’s estimates rather than audited segment disclosures, but the reasoning is worth considering: a considerable amount of hyperscaler infrastructure investment is ultimately justified by expectations for continued rapid growth in demand from relatively young AI businesses.
His concern is particularly interesting because he has not concluded that investors should immediately short AI. His argument is almost the opposite: before the subprime trade he could observe deteriorating mortgage data; with AI, the comparable evidence remains incomplete because crucial companies remain private and industry economics continue changing rapidly.
That is a sensible distinction between identifying vulnerability and predicting collapse.
Michael Burry has gone further, questioning whether all the demand appearing within the AI ecosystem ultimately originates from independent end customers or whether some of it is supported by interconnected financing, investment and purchasing relationships between chipmakers, infrastructure providers and AI companies. Again, circularity is not in itself proof of a bubble. But it makes final customer demand and return on invested capital increasingly important metrics to watch.
2. The $1.65 trillion “hidden debt” story needs some unpacking
One of the more alarming numbers appearing in the debate is $1.65 trillion.
A Nikkei analysis estimated that Alphabet, Microsoft, Amazon, Meta and Oracle had around $1.65 trillion of obligations that did not appear as conventional balance-sheet debt, including future data-centre leases, purchase commitments and infrastructure arrangements. Moody’s has separately examined very large debt-equivalent obligations associated with data-centre development.
Calling all $1.65 trillion “hidden debt” risks exaggerating what the accounts actually say. Many of these obligations are disclosed in financial-statement notes and consist of future contractual commitments rather than money already borrowed. A filing-based critique of the headline calculation argues precisely this point: leases and purchase commitments are economically relevant obligations, but they should not automatically be treated as equivalent to bank loans or bonds.
This is actually a more interesting story than a sensational debt headline.
For decades, the great technology companies were prized partly because they were extraordinarily asset-light. Software could be reproduced at negligible marginal cost. AI is forcing some of these same companies into an unprecedented infrastructure race involving chips, electricity, data centres, networking and long-duration commitments.
The business model has not necessarily become bad. It has become more capital intensive.
And that changes the valuation question.
3. Buffett’s Alphabet purchase is more interesting than either bulls or bears admit
Warren Buffett provides a useful counterweight to the simple AI-bubble argument.
Berkshire Hathaway built a substantial Alphabet position, and Buffett confirmed in July 2026 that he had personally initiated it. Yet his explanation was hardly an unconditional endorsement of the entire AI boom.
He emphasised the extraordinary amounts of capital that Google and its competitors now have to deploy, while simultaneously arguing that Alphabet’s record makes it more likely to emerge as a winner than most investments marketed on Wall Street. He also remarked that there were several Berkshire businesses he still preferred to Alphabet.
It is possible to believe that Alphabet is an exceptional business while questioning the economics of the wider AI investment boom. It is possible to own US technology while believing that valuation and capital intensity matter. And it is possible to expect several winners while also expecting a large amount of capital elsewhere in the ecosystem to earn disappointing returns.
That is very different from “AI is a bubble.”
4. SpaceX and the danger of extrapolating enormous addressable markets
Discussing the AI component of SpaceX, including xAI, Aswath Damodaran highlighted banker estimates of a roughly $22 trillion total addressable market, describing the number as closer to “hallucination than estimate.” His objection was not that AI lacks a gigantic potential market; it was that TAM calculations become dangerous when virtually every conceivable future revenue stream is added together and then used to justify today’s valuation.
SpaceX provides an unusually vivid example. Elon Musk has moved the company’s internal target of $1 trillion in annual revenue forward to 2030. The ambition may ultimately prove visionary, but even from a hypothetical $100 billion annualised revenue base, reaching $1 trillion within four years would require extraordinary compounded growth.
Investors have already shown that there are limits to their patience. SpaceX shares fell sharply after its first public quarterly results as spending on AI infrastructure accelerated, despite strong revenue growth.
None of this means Musk cannot succeed, but spectacular company and a spectacular investment are not necessarily the same thing at every price..
5. One important correction: the American rally is broadening
There is also evidence against the most extreme bubble narrative.
If a handful of giant technology companies were doing all the work while the remaining market quietly deteriorated, concern would be considerably greater. Yet by 8 September 2026 the S&P 500 Equal Weight Index was up about 12.84% year to date, compared with roughly 12.1% for the conventional S&P 500.
Small- and mid-sized US companies have also participated strongly. The implication is that investors should not confuse high valuations and AI concentration with evidence that the entire American market has become a narrow speculative façade.
This remains an extraordinarily productive economy containing extraordinarily productive companies.
The Guidefinances position is therefore not “bet against America.”
It is stop assuming that the price of American superiority is irrelevant.
6. The “anti-bubble”: why China deserves another look
Marc Faber has used a provocative description for the opposite side of the trade: China as an “anti-bubble.” At the peak of the technology boom in March 2000, investors were willing to pay extraordinary prices for fashionable New Economy companies while many traditional businesses had become deeply unfashionable and unusually cheap. Some of the best subsequent opportunities were therefore present at the same moment as one of history’s largest speculative peaks. A market does not need to crash before neglected assets become attractive.
A bubble and an anti-bubble can coexist. China is a plausible contemporary candidate because the contrast is unusually large. MSCI China currently trades at 13.96 times trailing earnings versus 25.8 for MSCI USA, while forward multiples are approximately 10.76 and 20.06 respectively.
But valuation is only the starting point. To understand whether China is merely cheap or genuinely interesting, investors need to understand something of the system producing those companies.
7. China is not simply a cheaper United States
One of the recurrent errors in Western investment analysis is to examine China as if it were an American economy with different accounting ratios.
It is not.
Political institutions, relationships between government and business, historical memory, capital allocation and the role of the state are different enough that an investor who focuses only on price-to-earnings ratios misses part of what is being purchased.
Ray Dalio’s June 2026 essay The Tribute System: The New World Order offers one provocative framework for thinking about those differences. Dalio draws on more than four decades of visiting China and argues that Chinese leaders’ behaviour is easier to understand when viewed through China’s own history rather than exclusively through Western strategic categories.
His framework should not be mistaken for academic consensus. But it is useful.
Confucianism, hierarchy and the “state family”
Dalio places considerable weight on the Confucian tradition, in which social stability depends partly upon people understanding obligations associated with their positions within families and wider society. He links this to a more hierarchical understanding of political order than is customary in liberal Western systems.
The Chinese word for country, guójiā — 国家 — combines characters commonly associated with state/nation and family/home, an etymology Dalio uses to illustrate this family-like conception of political order.
We should be careful not to extract an investment thesis from a Chinese word.
Modern China is also shaped by Legalist political traditions, Taoism, Buddhism, Marxism, nationalism, market economics and enormous regional differences. It contains major ethnic and cultural minorities despite the numerical dominance of the Han population.
Nevertheless, the deeper point remains useful: Chinese political legitimacy is historically associated not only with procedures but with order, performance, continuity and the capacity of the state to deliver results.
Chinese leaders are also unusually explicit about studying long historical cycles. The traditional idea of a dynasty gaining or eventually losing the “Mandate of Heaven” remains part of China’s historical vocabulary for understanding how disorder, poor government and declining legitimacy can lead to systemic change.
For investors, this helps explain why Beijing often assigns exceptionally high value to political stability, social order and national unity—even where the economic cost appears considerable.
8. The tribute system: a useful model, but not a law of history
Dalio’s most controversial claim is that China’s growing power could produce a modern analogue of the historical tribute system.
For long periods of imperial history, Chinese governments maintained relations with neighbouring states through arrangements in which differences in status were formally acknowledged, diplomatic rituals reinforced hierarchy and economic relationships provided mutual benefits. In Dalio’s interpretation, the stronger state offered benefits to countries that accepted the relationship, while pressure could be applied when they challenged what Beijing considered the legitimate hierarchy.
This fits his broader argument that China may prefer influence and dependency to territorial conquest, especially when objectives can be obtained without direct war.
There is, however, an important academic qualification.
Historians and international-relations scholars have repeatedly challenged the idea that there was one coherent “tribute system” governing East Asia for two thousand years. Zhang Feng’s research argues that the traditional framework is too simple, while Yuan-kang Wang shows that relations were sometimes hierarchical and at other moments much closer to diplomatic equality depending upon the actual balance of power.
That distinction matters enormously.
The tribute system is best understood as a lens through which to examine Chinese statecraft, not a deterministic forecast of Chinese behaviour.
9. Sun Tzu, Go and winning without the decisive battle
A second part of Dalio’s framework draws on The Art of War.
Sun Tzu’s famous ideal is to “subdue the enemy without fighting.”
The corresponding investment insight is not that modern Chinese officials mechanically follow a 2,500-year-old military manual. It is that economic leverage, technology, trade, regulation, diplomacy, access to markets and control over strategically important resources can all produce geopolitical effects before anybody fires a weapon.
Dalio contrasts chess with the Chinese game of Go. Chess culminates in the destruction of the opposing king; Go is fundamentally about surrounding territory and gradually restricting the opponent’s room to manoeuvre.
Western investors often focus on dramatic events—wars, sanctions, elections, invasions. Chinese strategy can sometimes be easier to understand by watching the slower accumulation of economic dependencies, industrial capacity, infrastructure, trade relationships and political influence.
10. From American exceptionalism to a bipolar world?
Dalio takes the argument considerably further.
Following the latest confrontation around the Strait of Hormuz, he has compared the changing perception of US power with Britain’s experience during the Suez crisis, which became a symbolic marker of the decline of British imperial power. He argues that some Asian governments are becoming less confident that the United States can or will indefinitely guarantee the existing international order, while simultaneously accommodating China’s increasing economic strength.
Historical analogies are especially dangerous when used to forecast markets. Britain’s relative decline unfolded over decades; the dollar, American capital markets, the US military and the country’s technology ecosystem remain extraordinarily powerful. But the possibility of a less US-dominated world does not require believing in imminent American collapse.
A world in which China is simply more powerful relative to the United States than it was twenty years ago is already enough to alter portfolio logic. Diversification becomes more valuable when the world itself becomes less unipolar.
11. AI may be one of the clearest examples of China’s progress
The strongest argument for taking China seriously may no longer be cheap labour or manufacturing scale.
The 2026 Stanford AI Index concludes that the performance gap between leading US and Chinese AI models has narrowed dramatically. As of March 2026, its measures showed the leading American model ahead by only 2.7%, with US and Chinese models having exchanged leadership positions during the preceding year. China also leads in AI publication volume, citations and industrial-robot installations, while the United States retains significant advantages in private investment and the production of top-tier frontier models.
The investment gap remains huge: Stanford estimates private US AI investment at $285.9 billion in 2025 versus $12.4 billion in China, although private-market figures inevitably understate the importance of Chinese government-supported capital.
This combination creates an unusual competitive environment.
America has far more capital.
China has shown that it can sometimes compensate with engineering efficiency, enormous domestic scale, lower prices and open-weight models.
Chinese models from companies including DeepSeek, Alibaba and Moonshot have progressively narrowed capability gaps while strengthening the competitive case for open-weight AI.
For American AI companies, the threat is not necessarily that Chinese models become unequivocally superior.
They only need to become good enough at a sufficiently lower price.
That is where Eisman’s argument becomes particularly important. If enterprises decide that the most expensive frontier intelligence is unnecessary for a large proportion of routine workloads, cheap open models could place pressure on prices and therefore on the revenue assumptions being used to justify gigantic infrastructure spending.
This is not yet evidence that OpenAI or Anthropic will lose. Recent business-spending data still show strong demand for both. But the competitive threat has become credible.
12. Dual circulation: self-sufficiency does not mean isolation
The economic component of China’s strategy is often described through dual circulation.
The easiest way to understand it is as two interconnected economies.
The first is China’s enormous domestic system, in which Chinese households, companies, banks and governments transact with one another.
The second is the external economy through which China trades, invests and competes with the rest of the world.
The policy objective is not complete autarky. Chinese government documents describe domestic circulation as the mainstay while domestic and international circulation reinforce one another. More recent policy documents for the 2026–2030 Five-Year Plan simultaneously emphasise greater technological self-reliance, stronger domestic demand and continued high-standard opening to the outside world.
China wants the benefits of global trade while becoming less vulnerable to external pressure in semiconductors, energy, software, industrial equipment and other strategically important sectors.
13. China’s external economy is strong; its internal economy is much less comfortable
China’s aggregate economic success can obscure an important contradiction.
Its external industrial economy remains formidable. China continues to run very large trade surpluses and possesses extraordinary production capacity across sectors ranging from machinery and electronics to batteries, electric vehicles and renewable-energy technologies.
Internally, conditions are much less straightforward.
The IMF expects Chinese real GDP growth of around 4.6% in 2026, following 5% in 2025, but continues to highlight weak domestic demand, the prolonged property-sector adjustment, elevated debt, deflationary pressures and an aging population as significant medium-term challenges.
The Chinese government itself clearly recognises the problem. Its new consumption plan for 2026–2030 aims to increase household consumption’s role in economic growth, while senior officials continue to identify insufficient domestic demand, real-estate risks and local-government debt as policy priorities.
This is one reason simplistic comparisons between China and the United States fail.
China can simultaneously be an extraordinarily competitive industrial power and have a troubled property market.
Its exporters can dominate world markets while Chinese households remain cautious.
Its technology companies can innovate while the overall equity market remains cheap.
Two apparently contradictory stories can both be true.
14. Why Chinese equities are cheap
Valuation sceptics sometimes treat China’s low multiples as though Western investors simply failed to notice them.
The discount reflects several genuine risks.
The property adjustment is unresolved. Local-government and corporate debt remain significant. The population is aging. Domestic consumption has not replaced investment and exports as rapidly as policymakers would like. Government intervention in industries can change the economics of private businesses rapidly. Strategic competition with the United States creates sanctions, tariff and technology-access risks. Taiwan remains a potentially enormous geopolitical tail risk.
And there is market evidence that Chinese equities deserve a higher risk premium.
MSCI reports three-year annualised volatility of around 22.6% for MSCI China versus 13.1% for MSCI USA. Its recorded maximum drawdown for MSCI China is 73.3%, compared with 55.4% for MSCI USA.
Cheap does not mean safe.
It means the investor is being paid more—in valuation terms—to accept the uncertainty.
| Factor | United States | China | Guidefinances interpretation |
|---|---|---|---|
| Trailing P/E, 31 Aug. 2026 | 25.8× | 14.0× | Large valuation advantage to China |
| Forward P/E | 20.1× | 10.8× | Expectations remain far higher in the US |
| Price/book | 5.69× | 1.45× | US profitability/quality premium is enormous |
| 3-year volatility | 13.1% | 22.6% | China is materially riskier in market terms |
| AI position | Frontier leadership, huge capital base | Rapidly narrowing capability gap, strong open-weight ecosystem | Competition may pressure future AI economics |
| Domestic economy | Stronger consumption and capital markets | Property, demand and debt problems | China’s low valuation has fundamental causes |
| Political/regulatory risk | Relatively strong shareholder protections | Higher state and regulatory intervention | A persistent China discount is rational |
| Geopolitical risk | Global power but heavily exposed to AI/Taiwan supply chains | US rivalry and Taiwan are major tail risks | Diversification helps, but neither market is geopolitically isolated |
Source: MSCI data to 31 August 2026 and cited research above.
15. Taiwan: the risk that sits above both markets
Dalio expects China to use progressively stronger economic, diplomatic and strategic pressure in pursuit of its objective of reunification with Taiwan, rather than assuming that a full-scale military confrontation is necessarily Beijing’s preferred route. That is consistent with his broader Art of War interpretation.
Investors do not need to decide whether this forecast is correct.
They merely need to acknowledge the asymmetry of the risk.
A serious Taiwan crisis could be disastrous for Chinese equities. But it would hardly be an isolated Chinese event. Stanford notes that a single Taiwanese foundry remains responsible for an extraordinary proportion of leading-edge AI-chip fabrication.
A geopolitical crisis involving Taiwan could therefore simultaneously damage China, global semiconductor supply, American AI companies and technology valuations worldwide.
This is precisely why geopolitical diversification is harder than buying securities listed in different countries.
Supply chains connect the risks again.
16. What should investors actually do?
The Guidefinances response is not to replace the American consensus with a Chinese one.
That would merely exchange one concentration for another.
The first principle remains a diversified global portfolio. An investor already using a broad global equity index already owns both American and Chinese companies, although the enormous growth in US market capitalisation means that such portfolios are increasingly dominated by the United States.
The more interesting decision concerns incremental capital.
For an investor whose portfolio has become heavily concentrated in US mega-cap technology, directing some new money toward other regions, smaller companies, value-oriented strategies or emerging markets can reduce dependence on one valuation regime without requiring a heroic prediction that the AI boom is about to end.
China can reasonably form part of that diversification.
Broad exposure is preferable to attempting to identify the single Chinese technology company that will defeat its American counterpart.
For European and Swiss investors, UCITS vehicles tracking broad Chinese indices are readily available; MSCI itself lists products from iShares, Xtrackers and HSBC among the funds linked to its China index. Investors should compare not only the TER but also fund domicile, replication, liquidity, securities lending, trading currency, withholding taxes and the precise mix of A-shares, Hong Kong shares and overseas listings before choosing a vehicle.
The same logic applies on the US side.
Reducing concentration does not require selling America. An investor can remain exposed to American economic success while shifting part of the allocation away from the most expensive mega-cap companies through broader, equal-weight, smaller-company or value exposure.
That is perhaps the most useful part of Faber’s “anti-bubble” idea.
At a market top, the answer need not be cash.
Sometimes the alternative is simply the part of the market nobody currently wants.
17. What would make the China thesis stronger—or weaker?
A valuation gap can persist for decades, so price alone is not a catalyst.
For the Chinese investment case to become substantially stronger, we would want to see evidence that domestic consumption is becoming structurally more important, the property adjustment is stabilising, private-sector confidence is improving and Chinese companies are increasingly able to translate technological progress into durable shareholder returns.
Improving relations with major trading partners would also reduce the political discount embedded in Chinese assets.
Conversely, a renewed property deterioration, entrenched deflation, greater intervention in private enterprise or a serious escalation around Taiwan would make even today’s low multiples less compelling.
The US side has equivalent tests.
The expensive-market thesis weakens if AI infrastructure begins producing convincingly high returns on capital across a broad set of customers, productivity improvements accelerate and earnings continue catching up with valuations.
It strengthens if capital expenditure rises faster than monetisation, if frontier-model pricing collapses, if corporate customers increasingly substitute much cheaper models for premium ones, or if the AI financing chain begins depending more heavily on itself than on external customer demand.
These are observable conditions.
They are more useful than predicting the date of the next crash.
Guidefinances conclusion: hedge your bets
There is an appealing simplicity to grand narratives.
America is in decline. China will dominate.
Or the opposite:
China is uninvestable. American technology will dominate indefinitely.
Both may make entertaining market commentary. Neither is a sensible foundation for a long-term portfolio.
The United States remains the world’s deepest capital market and hosts many of its strongest businesses. AI may create enormous economic value, and betting aggressively against American innovation has historically been an expensive habit.
But price still matters.
At roughly 26 times trailing earnings for MSCI USA against 14 times for MSCI China, investors are being asked to pay vastly different prices for the two futures. Those prices imply vastly different expectations.
China, meanwhile, is not merely a collection of cheap stocks. It represents a different economic and political system with different risks, strategic objectives and relationships between companies and the state. Understanding Confucian traditions, the history of hierarchy, the debate around the tribute system, The Art of War, dual circulation and technological self-reliance does not provide a formula for forecasting Chinese equities.
But it helps explain what we are buying.
Faber may ultimately be wrong to call China the “anti-bubble.” Dalio may be wrong that a modern tribute system will characterise the emerging Asian order. Eisman and Burry may underestimate the ability of AI companies to monetise today’s immense investment. Musk may even deliver something close to the extraordinary growth he predicts.
The portfolio does not need any one of them to be right.
The better response to radical uncertainty is not radical conviction. It is diversification.
Keep exposure to the American companies that continue to compound capital successfully. Be careful about allowing enthusiasm for AI to turn a diversified portfolio into an accidental technology bet. Look again at markets and sectors that have been ignored precisely because their narratives are uncomfortable. And recognise that China is now too large economically, industrially and technologically to be understood only as a geopolitical risk.
The goal is not to identify the next empire.
It is to build a portfolio that does not require one empire to win.
Selected sources and further reading
MSCI USA Index and MSCI China Index, valuation, composition and risk data as of 31 August 2026.
Stanford Institute for Human-Centered AI, The 2026 AI Index Report, particularly the sections on US-China model performance, open models, investment and infrastructure.
International Monetary Fund, China 2025 Article IV Consultation and July 2026 projections, for growth, property, debt, domestic-demand and demographic risks.
Ray Dalio, The Tribute System: The New World Order, 18 June 2026, for the cultural, historical and geopolitical framework discussed in this article.
Zhang Feng, “Rethinking the ‘Tribute System’,” and Yuan-kang Wang, “Explaining the Tribute System: Power, Confucianism, and War in Medieval East Asia,” for academic qualifications to the tribute-system interpretation.
Aswath Damodaran, Musings on Markets, 2026, for the discussion of AI economics, total addressable markets and SpaceX/xAI valuation.
Warren Buffett interview with CNBC, July 2026, for his comments on Alphabet, AI capital expenditure and long-term return on capital.
Steve Eisman, August 2026 commentary and interviews, for the concentration of AI demand around OpenAI and Anthropic and the potential challenge from lower-cost models.
Chinese State Council and 15th Five-Year Plan materials, for domestic demand, technological self-reliance and the evolving dual-circulation strategy.
Editorial note: market valuations, index data and current economic information were reviewed on 10 September 2026. They will change and should be checked again before making investment decisions.
Guide Finances 