From transmission risk to portfolio resilience: did diversification hold up?

In January, I argued that the best response to uncertainty around the AI capex boom was not prediction, but diversification* across asset classes, sectors, regions, liquidity profiles and risk premia. Part one revisited the transmission channels; this second part asks whether that portfolio response earned its keep.

Six months is too short to judge a strategic allocation, and the sharpest stress test reversed quickly. Using broad benchmarks and US-listed ETF flows as proxies, the answer is qualified: diversification helped only where it introduced genuinely different return drivers.

A broader opportunity set

The strongest headline return in 2026 so far has come from commodities. The Bloomberg Commodity Total Return Index is up approximately 22% year to date, ahead of global equities at around 10% (MSCI ACWI) and the S&P 500 at 10%, while the Bloomberg US Aggregate Bond Index is essentially flat at 0.2%.^1,2,3,4

Energy drove most of the commodity gain, rising 58%, while the Bloomberg Gold sub-index fell 8%.⁵ After reaching an intraday record of $5,595 an ounce on 29 January, spot gold declined by more than 28% to its 24 June closing trough of $3,999.⁵

Regional exposure also helped. Developed markets outside the US gained around 9%, while emerging-market equities rose approximately 17%, led by South Korea and Taiwan at 55% and 49% respectively.⁶ Yet both markets are closely connected to the semiconductor supply chain. Geographic diversification did not always mean economic diversification.

Concentration met broader leadership

The clearest evidence came from within equities. US value outperformed growth by 17.6 percentage points; the S&P 500 Equal Weight Index gained around 12%; and the Russell 2000 returned approximately 20%, all against 10% for the cap-weighted S&P 500.⁷ This does not prove value, equal weight or small caps will always outperform, but reducing dependence on a handful of dominant companies helped during this period.

Listed infrastructure offered a more nuanced result. The S&P Global Infrastructure Index returned approximately 11%, broadly matching the S&P 500 despite a 38-basis-point rise in the US 10-year Treasury yield.⁸ Its maximum drawdown was 5.9%, versus 8.9% for the S&P 500.⁸

Where did the money move?

US-listed ETFs attracted just over $1 trillion in the first half. Non-US equity ETFs drew approximately $140 billion and emerging-market equity ETFs a further $39 billion, together representing around 26% of equity ETF inflows.⁹

Investors also addressed concentration directly. The largest S&P 500 equal-weight ETF attracted approximately $10 billion, broadly matching the $11 billion drawn by the small-cap ETF category.⁹ Equal weight appeared the clearer concentration-management tool.

Performance and flows did not always align. Although value outperformed growth by 17.6 percentage points, pure value ETFs attracted only $2.5 billion, less than the $6.7 billion directed to growth. The three largest dividend ETFs drew $15.9 billion, while representative quality-factor ETFs recorded approximately $7 billion of outflows.^9,10 Investors preferred an income-based expression of value and quality.

Fixed income played a different role. Bond ETFs attracted approximately $300 billion, or 29% of total ETF flows despite representing around 16% of assets. Four short-term government-bond ETFs drew $32.5 billion, more than the $18.6 billion directed to representative core aggregate and investment-grade vehicles, while long-duration Treasuries saw outflows.⁹ Investors captured yield while limiting exposure to an uncertain rate path.

Options-based income strategies attracted $45.3 billion. Covered-call vehicles dominated, while buffer and defined-outcome flows were more modest.¹⁰ These strategies may exchange upside for income or specified protection, but express 'prepare, don't predict' by choosing a payoff profile rather than one market forecast.

Theory versus reality

So, did diversification hold up? Partially and unevenly. It helped most where it changed the underlying return driver: regional breadth beyond US mega caps, equal-weight exposure, value and income, short-duration fixed income and explicit downside management.

That distinction matters. Much of 2026's apparent market broadening still appears tied to the same AI investment cycle driving mega-cap returns. Russell 1000 Value returned 18.8% year to date to 17 July 2026 versus 1.3% for Russell 1000 Growth, while the S&P 500 Equal Weight Index returned 12.4%, outperforming its cap-weighted counterpart at 9.6% by roughly 280 basis points.⁷ Yet beneath those headlines, concentration remained striking. Among the largest active contributors to Russell 1000 Value portfolio return, Micron Technology** alone delivered 3.18 percentage points of active contribution to return, equivalent to roughly 19% of the portfolio's total active contribution.¹¹ AI-linked holdings accounted for the dominant share of active outperformance despite representing a relatively small portion of average portfolio weight.¹¹

Leadership broadened, but the underlying earnings and return engine remained far more concentrated than headline index performance suggested. A value tilt exposed to semiconductor equipment, AI infrastructure and data-centre supply chains was often participating in the same cycle as a growth allocation through a different wrapper.

Six months cannot settle a long-term argument, but January's central conclusion has survived contact with the data. We cannot know the enduring AI winners, whether today's investment will generate adequate returns or how financing conditions will evolve; investors can, however, reduce their portfolios’ dependence on any single outcome by diversifying across genuinely different return drivers.

Diversification has not been a universal shield. It has, however, helped limit the cost of being wrong where exposures genuinely introduced different return drivers. In a concentrated, capital-intensive and increasingly interconnected AI cycle, 'prepare, don't predict' remains the most practical conclusion.

*It should be noted that diversification is no guarantee against a loss in a declining market. **For illustrative purposes only. Reference to a particular security or index is on a historic basis and does not mean that the security is currently held or will be held within an L&G portfolio. The above information does not constitute a recommendation to buy or sell any security. Figures shown are approximate, point-in-time figures for illustrative purposes and are not a recommendation. Past performance is not a guide to future performance.

Sources

• [1] Bloomberg, BCOMTR Index (Bloomberg Commodity Total Return Index, USD); total return 31 December 2025 – 17 July 2026.
• [2] Bloomberg, MXWD Index (MSCI ACWI Total Return, USD); total return 31 December 2025 – 17 July 2026.
• [3] Bloomberg, SPX Index (S&P 500 Total Return, USD); total return 31 December 2025 – 17 July 2026.
• [4] Bloomberg, LBUSTRUU Index (Bloomberg US Aggregate Total Return); total return 31 December 2025 – 17 July 2026.
• [5] Bloomberg, BCOMENTR, BCOMAGTR, BCOMINTR and BCOMGCTR sub-indices (USD total return); XAU Curncy (Gold Spot, USD per troy ounce); 31 December 2025 – 17 July 2026. Gold intraday high of $5,595 on 29 January 2026; closing price of $3,999 on 24 June 2026.
• [6] Bloomberg, NDDUEAFE, MXEF, MXKORT1N and TAMSCI indices (USD total return); 31 December 2025 – 17 July 2026.
• [7] Bloomberg, RLV, RLG, SPW, RTY and SPX indices (USD total return); 31 December 2025 – 17 July 2026.
• [8] Bloomberg, SPGTIND Index (S&P Global Infrastructure Index, USD total return); SPX Index; GT10 Govt (US 10-year Treasury yield); maximum drawdown calculated on daily closing prices of IGF US Equity as proxy; 31 December 2025 – 17 July 2026.
• [9] Bloomberg fund-flow data, US-listed ETFs; AUM as at 30 June 2026 and flows from 31 December 2025 to 30 June 2026. Categories and vehicles as specified in the underlying analysis.
• [10] Bloomberg fund-flow data for representative quality-factor, options-based income, covered-call and buffer/defined-outcome ETFs; AUM as at 31 December 2025 and flows to 30 June 2026. Quality and buffer figures are directional proxies.
• [11] Portfolio attribution report, Russell 1000 Value holdings; active contribution to return by holding, average active weight. 31 December 2025 – 17 July 2026.

All flow and AUM data are in US dollars. Factor and defined-outcome ETF flows are based on representative vehicles and should be treated as directional.