The All Weather portfolio: Ray Dalio's answer to the AI bubble

12 min read12 min

03/08/26

He still regards All Weather as an essential core after 30 years, but did not reaffirm the popular 30/55/15 mix as today's target allocation.

Planned total 2026 capital expenditure, not pure AI spending. Alphabet and Meta are ranges; the combined figure of roughly $728bn uses their midpoints. Amazon: plan updated after the Q2 report. Sources: company statements and earnings calls, as of 3 August 2026.

The total is striking, but not automatically irrational. Amazon reported 37 percent revenue growth at AWS in the second quarter and an annualised revenue run rate above $25 billion for its AI business. Microsoft had already named a $37 billion annualised AI revenue run rate in April, while Alphabet reported a $240 billion Cloud backlog. Unlike many dot-com companies, these groups earn substantial profits, own established businesses and see real demand.

That is precisely what makes the position difficult. A technology can be real, its winners can be profitable, and their shares can still price in too much of the future. Railways transformed the economy and still ruined many providers of capital. The internet prevailed even though the Nasdaq fell deeply after March 2000. Technological success and investment return are two different questions.

Three warning signs of a bubble

1. Spending is growing faster than free cash flow. Amazon reported a free cash outflow of $7.6 billion for the twelve months to June 2026, against an $18.2 billion inflow in the previous period. It attributes the deterioration mainly to a $66.1 billion increase in property and equipment investment, primarily for AI. Meta's second-quarter free cash flow fell from $8.55 billion to $784 million as capital expenditure reached $31.08 billion. That does not prove misallocation. It shows how large future returns must become to justify the capital committed.

2. Index investors already carry the trade. In May 2026 the ECB described technology and AI valuations as particularly high and euro-area investors' exposure to US equities as having grown sharply. If productivity, adoption or margins disappoint, the repricing will therefore not stop with thematic funds. The same large names carry it into world ETFs, retirement savings and insurance portfolios.

3. An equity risk is turning into a credit risk. The AI build-out was initially financed largely from hyperscaler profits. According to the ECB, AI companies and infrastructure are now relying increasingly on credit finance. Historically, 15 percent of years with especially rapid growth in both equity prices and business debt were followed by a financial crisis within two years. That is expressly not the probability of an AI crisis. It is a conditional historical warning sign.

The International Monetary Fund has put the transmission mechanism into a scenario. A moderate correction in AI stock valuations, accompanied by tighter financial conditions, would reduce global growth by 0.4 percentage points relative to the baseline. For a private portfolio, the relevant message is not that this scenario must happen. It is that a concentrated expectation has become large enough to turn a sector correction into a macroeconomic shock.

What Ray Dalio actually recommends in 2026

On 23 March 2026, Dalio did more than recount the history of All Weather: he explicitly reaffirmed it. For most investors, he argued, two things matter most: a well-engineered, broadly diversified mix and little or no market timing. He describes All Weather as a passively held combination of investments, not a single product. Thirty years on, he still regards the strategic core mix as essential and continuously holds his own evolving version for his family and foundation.

He had been even more direct in a public AMA in September 2025: a risk-balanced All Weather portfolio was the best starting point for most investors most of the time. In the simultaneously AI-enabled, debt-burdened and turbulent world, he also highlighted companies capable of productive AI advances, hard-money assets rather than debt assets and excellent global diversification. Those are current emphases around a strategic core, not a new fixed formula.

That is not a new endorsement of the popular five fixed weights. Dalio says his approach has evolved and improved. He intended to publish an updated “recipe”, but none was publicly available by 3 August 2026. The familiar 30/55/15 mix should therefore not be presented as his current personal portfolio or as a newly confirmed target allocation.

On 15 June, Dalio made the connection to AI concentration himself. A small number of predominantly AI-driven companies now dominated markets and the economy. Even the eventual winners of previous technology cycles suffered enormous interim losses; revolutionary technologies make overinvestment or underinvestment difficult to avoid. His answer is to seek roughly 15 good, independent investments whose risks are balanced rather than make one concentrated bet.

Dalio's own, explicitly uncertain assessment put real equity returns over the next five to ten years between minus 5 and minus 10 percent. That is not a dependable forecast and not a reason for hurried selling. It shows how seriously he currently regards the combination of valuation, concentration and prospective return.

His recent comments also change how the defensive building blocks should be read. In February 2026, amid high government debt, Dalio suggested underweighting debt assets and overweighting gold, with a smaller allocation to Bitcoin. He treats gold first as a strategic portfolio holding; his analysis suggests a range of 5 to 15 percent depending on the other assets and the investor's risk preferences. That is consistent with All Weather's logic, but not with blindly copying a US-heavy bond allocation.

What All Weather really means

Bridgewater developed All Weather in 1996 for Ray Dalio's family trust. It began with a simple observation: a conventional 60/40 portfolio may hold 60 percent equities and 40 percent bonds, but almost all of its risk comes from equities. Capital is diversified; risk is not.

All Weather instead organises assets around their response to two surprises: growth and inflation. Either can come in above or below what the market expects. That creates four economic seasons. No forecast says which one comes next, so each should find a part of the portfolio built for it.

The important term is risk balancing. The institutional strategy does not simply put equal amounts of money into four buckets. It recognises that equities, long-duration bonds, gold and commodities move by very different amounts, and it can use derivatives or leverage to balance their risk contributions. That is not the same thing as the familiar retail mix.

The familiar retail version

The implementation popularised by Tony Robbins and commonly described as the “Ray Dalio All Weather” allocation divides capital as follows:

These weights are a public simplification, not the disclosed composition of Bridgewater's institutional fund and not a universal recommendation. Dalio's 2026 comments reaffirm the approach, but not these exact weights. They are also built heavily around US Treasuries. For an investor in the euro area, currency, tax rules, available UCITS products and personal horizon all change the practical implementation.

How the four economic seasons fit together

This mapping is not a law of nature. Correlations change, and several building blocks may fall in the same year. The benefit comes from the structural difference in their sources of return, not from a guarantee that one always rises.

What carries the portfolio through an AI correction

An AI crisis can arrive by at least three routes, and that uncertainty is precisely the case for a weather-resistant core.

• Demand disappoints. Data centres are less fully used than planned, depreciation rises and earnings expectations fall. That is mainly a negative growth shock. High-quality government bonds may then stabilise the portfolio, provided inflation eases as well.
• The build-out creates an inflation shock. Power, copper, cooling and grid connections become scarce. Equities and nominal long bonds may then suffer together; commodities, gold and inflation-linked bonds become the more relevant counterweight.
• AI meets the high expectations. Productivity and profits rise. The equity sleeve still participates. An All Weather core does not require missing the technological winner; it only refuses to make the entire portfolio depend on it.

The portfolio therefore does not short the AI trade. It dilutes its power. A 30 percent global equity allocation still contains the large platforms, but a price collapse can no longer determine the entire outcome by itself. Government bonds, inflation protection, gold and commodities receive a role in advance, before the crisis supplies its own persuasive narrative.

Where All Weather itself can fail

All Weather does not mean “no losses”. Four limits belong in any honest mandate.

Long bonds can fall hard. When real yields and inflation rise together, long-duration bonds are especially sensitive. The year 2022 demonstrated that equities and nominal bonds can lose at the same time. Duration is insurance against deflation, but a risk during inflation.

Gold pays no running income. It can lag equities and bonds for years. Its role is a different response to certain crises, not a dependable return premium.

Commodity products are not spot prices. Futures-based funds carry roll costs and may behave very differently over long periods from the commodity index an investor has in mind. Gold ETCs and commodity ETPs also bring product, issuer and tax questions.

The simple copy is not risk parity. Mechanically adopting the five familiar weights produces a conservative multi-asset allocation, but not automatically the same risk balance as Bridgewater. Without leverage, derivatives and continuous risk measurement, the result remains capital-weighted.

Turning it into a resilient core

For self-directed investors, the most valuable element is not the exact percentage but the architecture. One possible and expressly illustrative rulebook looks like this:

1. Core before theme: 80 to 90 percent of the long-term portfolio follows a broadly diversified allocation built across several economic environments. Themes such as AI remain a clearly limited satellite.
2. Measure risk, not the number of funds: A world ETF and a technology ETF are not two independent bets when the same companies dominate both. Look through to the company level.
3. Choose the defensive currency deliberately: The bond sleeve is meant to stabilise spending and liabilities. Unhedged foreign-exchange risk can overwhelm that job.
4. Cover inflation explicitly: Gold, broad commodities and inflation-linked bonds perform different roles. “Bonds” alone are not complete inflation protection.
5. Rebalance by rule: Review annually and act only at a threshold fixed in advance. Direct new contributions first to underweight sleeves to reduce sales and tax friction.
6. Do not turn the thesis into the rule: Neither an AI headline nor a price target may change the core allocation. A change requires a different horizon, goal or capacity for risk.

Who the strategy suits

An All Weather core is more likely to suit investors who want less dependence on equities, have a multi-year horizon and accept that they will deliberately lag an all-equity portfolio during equity bull markets. That lag is not a defect. It is the cost of diversification.

The familiar 30/55/15 version is less suitable for very young investors with high risk capacity and a maximum-growth objective, for money needed soon, or for anyone who does not understand long-duration bonds and would sell at the first rise in yields. Even a robust strategy fails when its weakest period is emotionally impossible to hold.