Figure on a Swiss alpine balcony facing a financial skyline dominated by one unusually tall tower

Last December, I rebalanced around five theses for 2026. Five months later, moving money from US large companies into cheaper markets had mostly worked. Europe was the exception. As always, this is my 2026 midyear portfolio review, not a recommendation.

Portfolio performance review through May

Through May 29, the portfolio gained 5.2% in Swiss francs on a time-weighted basis. A time-weighted return removes the effect of deposits and withdrawals.

My global 60/40 benchmark gained 5.4%. It holds 60% in the MSCI All Country World Index and 40% in global bonds. The bond portion is hedged against movements in the Swiss franc. The S&P 500 total return gained 9.0% in francs.

The portfolio therefore matched the diversified benchmark and trailed the US index by almost four percentage points.

From January 2 to May 29, the portfolio gained 5.2% in francs, global 60/40 gained 5.4%, and S&P 500 gained 9.0%. The March trough was near -4%.

I was up modestly through January, lost those gains by late March, and bottomed near -4%. The portfolio then climbed steadily.

A Middle East conflict began in late February. A brief threat to the Strait of Hormuz sent oil sharply higher. Markets had expected inflation to keep slowing. For several weeks, the oil shock replaced that story with persistent inflation. I think that single event reset the economic backdrop for everything that followed.

In 2025, this approach beat the S&P 500 by a wide margin in Swiss francs. The dollar’s fall against the franc drove that result. The USD/CHF exchange rate fell about 11.5% that year. Through May 29, 2026, it had fallen only about 1.2%.

The currency tailwind that made the strategy look brilliant is gone. Without it, the portfolio did what I expected: matched a balanced benchmark and trailed the US index.

What worked, what didn’t

Returns through May 29: emerging markets 20.7%, small caps 13.2%, Japan 13.1%, US large caps 9.9%, and Europe 4.8%. Gold gained 1.0%; corporate bonds 0.2%; global bonds were flat; euro bonds -0.1%; Treasuries -1.3%; listed private equity -13.5%; Bitcoin -19.3%.

Emerging Markets gained about 20.7% in Swiss francs and contributed the most relative to its portfolio weight. US Small Cap gained 13.2%, and Japan gained 13.1%. My valuation thesis worked in three of the four market segments where I had added money.

I only half-anticipated what drove the emerging-market gain. In December, I had expected Chinese stimulus. Instead, Korean and Taiwanese chipmakers at the centre of the AI supply chain led the move. My broad emerging-market holding captured that gain; my dedicated Chinese-technology holding did not. I was right about Asia for the opposite reason I had written down. Fair enough.

The fourth rotation was Europe, and it did not deliver. It gained 4.8% in francs, compared with 9.9% for US large companies. Europe trailed the other three rotations by eight to sixteen percentage points.

Germany announced a €500 billion infrastructure fund, €400 billion for defence, and €600 billion in private commitments. Citi’s European equity team maintained its overweight view. An overweight position exceeds the market’s share in a benchmark. Citi projected 13% compound annual growth in German earnings per share through 2029.

None of that had appeared in prices by May 29. A cheap market can remain cheap until an event forces investors to value it differently. The event I expected had not done its job, at least not yet.

Bitcoin and listed private equity were the worst calls, down about 19% and 13% in francs. Bitcoin’s behaviour in March changed how I size it.

I am watching listed private equity more carefully. Its share-price discount to net asset value (NAV) widened and caused the loss. NAV is the value of assets minus liabilities. I see a crowded trade in a year with wider differences between investment returns, rather than a broken thesis.

Valuations got more extreme, not less

The Shiller cyclically adjusted price-to-earnings ratio (CAPE) was the spine of my December argument. CAPE compares share prices with ten years of inflation-adjusted earnings. When I wrote that article, the S&P 500 traded at 40.5 times earnings. The long-run mean was 17.3.

I expected the gap to close through lower valuations or higher earnings. Investors often call the first path multiple contraction.

S&P 500 CAPE, 1990 to May 2026: 42.7 in May, 39.8 in December 2025, 17.3 long-run mean, and 44.2 dot-com peak

Neither path closed the gap. The CAPE was about 42.7 in late May, up from 39.8 at the end of December. No reading since December 1999 had been closer to the dot-com peak of 44.2. The valuation rose while earnings kept pace, and the rest of the world remained cheaper.

Concentration is messier. Nvidia’s index weight rose from 7.2% in December to 8.17%. Apple represented 6.70%, while Microsoft fell from 5.9% to 4.96%. The ten largest companies represented 39.1% of the index.

The concentration problem had become a risk tied to one company, not the whole group known as the Magnificent Seven.

S&P 500 weights: Nvidia 8.17%, up from 7.2%; Apple 6.70%; Microsoft 4.96%, down from 5.9%; ten largest companies 39.1%

The Europe trade is alive but not paying

UBS’s Year Ahead 2026 described Europe’s discount using expected earnings as 22%. Its figures placed the US at 23 times forward earnings and Europe at 14 times. A forward ratio uses expected earnings.

By May 29, the US traded around 28 times past earnings, while Europe traded around 18 times. This trailing ratio gave Europe a discount near 35%. The cheap market had become cheaper relative to the expensive one.

Being early on a valuation gap that will not close is indistinguishable from being wrong for as long as it lasts.

May 2026 trailing price-to-earnings ratios: US 28, Japan 19, emerging markets 19, and Europe 18, a 35% discount to the US

Two quarters of strong US earnings supported the high valuation. FactSet reported a record 13.4% net profit margin for the first quarter of 2026. The index therefore looked expensive without looking fragile.

JPM’s Lakos-Bujas raised his year-end S&P 500 target to 7,600 in April. He raised expected 2026 earnings per share (EPS) to $330, a 22% annual increase. Morgan Stanley’s Mike Wilson set a 7,800 target and called the market “early cycle.” Bank of America’s Subramanian was more cautious at 7,100.

Forecasts had gathered near the high end of the range. For my Europe position, the lesson is that a wide discount can persist for years. A low valuation was never enough on its own. Europe needed a catalyst, and Germany’s fiscal shift had not yet changed the valuation.

Dollar, AI, bonds

Dollar. In December, I expected the dollar to fall 4% to 10% during 2026. By May 29, it had fallen about 1.2% against the franc. That was a stall rather than a fail.

Goldman Sachs, UBS, Pictet, ABN AMRO, and MUFG kept their weaker-dollar forecasts. They moved most of the expected decline into the second half of 2026. The underlying factors remained: US fiscal and current-account deficits, a smaller interest-rate advantage, and concerns about government debt. The International Monetary Fund highlighted the fiscal concerns in April.

The war revived demand for the dollar as a safe asset. A Federal Reserve focused on inflation also preserved the dollar’s interest advantage. I assume the thesis is delayed, not dead.

AI capital expenditure. The largest cloud companies were forecast in December to spend $571 billion on long-lived assets during 2026. First-quarter results and company forecasts raised the conservative range to roughly $660 billion to $725 billion. One financial-analyst estimate reached $805 billion. No major research firm reduced its forecast. Cloud revenue accelerated with the spending.

I still think AI may become a competitive commodity. In that case, much of the value flows to users rather than one model provider. But the investment cycle had clearly not peaked. Betting against it on valuation alone would have cost me.

Bonds. This is the clean miss. I built the holdings around two or three Federal Reserve cuts in 2026, and none occurred by May 29. The energy shock pushed inflation back towards 4%. The Federal Reserve held rates at every meeting.

By spring, markets expected the European Central Bank to raise rates before the Federal Reserve changed them. Swiss-franc corporate bonds returned about 0.2% in francs. Hedged euro government bonds returned about -0.1%, while US Treasuries returned about -1.3% in francs. The currency loss erased a roughly flat dollar return.

Interest income has been fine. My expected capital gain has not arrived. I had assumed that lower policy rates would raise bond prices.

The two diversifiers split

The clearest lesson from the first five months is that my two “diversifiers” are not the same animal. The March shock separated them.

Gold behaved like a diversifier when it mattered. It held its value during the energy-shock drawdown while equities fell. Central banks also continued buying gold. They bought 244 tonnes in the first quarter, according to the World Gold Council.

Over the full five months, my Swiss-franc-hedged gold holding gained only about 1%. The currency hedge removed gains that unhedged gold received from the dollar. I accepted that cost deliberately. Gold did its job during the only period that tested it. It remains my portfolio’s one genuine diversifier.

Bitcoin did not. During the March drawdown, it fell with equities and fell further than they did. It was down about 19% in francs for the year through May 29. Bitcoin behaved like an amplified bet on risk appetite and market liquidity, not like ballast.

Three portfolio allocation changes

I made three small changes. None reversed the strategy. A midyear review mostly earns the right to leave things alone, and that is what this one did.

Changes across 14 holdings: gold rose from 5% to 6%, crypto fell from 4.5% to 3.5%, and Japan stayed at 3% with a franc hedge

Gold from 5% to 6%. It is the portfolio’s one genuine diversifier and held up during the year’s only stress period. I funded the increase with new cash rather than selling anything. The currency hedge reduced this year’s reported gain, and I am keeping it anyway. I own gold to reduce equity losses in francs, not to make a separate dollar bet.

Crypto from 4.5% to 3.5%. This is not a loss of faith. I keep a meaningful position because the potential gain exceeds the amount at risk. But crypto’s job was diversification, and it failed that job in March.

I am sizing it for what it actually is. I let its weight fall and redirected new contributions instead of selling after the decline. The change therefore had no transaction cost.

Japan, switched into a Swiss-franc-hedged share class. Japanese equities gained well into the teens in local currency. A falling yen erased most of that gain for me as a franc investor.

I kept the equity exposure and hedged the currency. The Swiss National Bank’s policy rate was 0%, and Japanese rates were also low. The small rate difference made the hedge inexpensive compared with the currency move it removed. Same 3% weight, different currency.

I did not apply the same logic to the dollar. Hedging dollars cost roughly 4% because of the interest-rate difference. That cost was too high to offset a one-percentage-point currency move on a 2% Treasury position. I preferred to keep the extra interest income.

I also measured the AI exposure hidden inside my supposedly less concentrated portfolio. Market-capitalisation-weighted funds hold more of the largest companies, including Nvidia and TSMC. This indirect AI exposure was roughly 13% of the portfolio. No single company exceeded about 2%.

The reduction in concentration worked. I do not have a concealed AI bet masquerading as diversification. That was the outcome I most wanted to rule out.

What would change my mind

The thesis rests on a few load-bearing assumptions. A useful review names what would break each one.

For AI and US concentration, price alone is not the trigger because expensive assets can remain expensive. I would react if the largest cloud companies cut investment while their cloud revenue stopped growing. A major efficiency gain that made current infrastructure look excessive would have the same effect.

Through May 29, capital expenditure was still rising and generating revenue. If that relationship reverses, I will reduce the portfolio’s direct and indirect AI exposure.

For interest rates, falling inflation and Federal Reserve cuts would restore the bond case that I got wrong this spring. I would then buy bonds with longer maturities. Longer-maturity bonds respond more strongly to changes in interest rates.

If inflation remains high and energy costs feed into wages, gold must provide protection. I would keep the bonds at shorter maturities.

For Europe, I will keep the overweight position while the public-spending plan remains intact. I view the problem as a missing catalyst, not a broken thesis. If German spending stalls or analysts cut earnings forecasts, I will reduce the position to its benchmark weight.