Key points:

• AI meets expensive money: AI spending can remain strong, but higher funding costs, rapid obsolescence and rising reinvestment needs will increasingly separate companies that can comfortably finance the AI race from those spending simply to stay competitive.
• Inflation needs more than one hedge: Oil is only part of the risk. El Niño, fiscal spending and the AI infrastructure buildout could also keep pressure on food, power, commodities and bond yields.
• Bonds are competitive again: With the US 10-year Treasury yield around 5%, investors no longer need to rely as heavily on equities for returns. They should assess whether recent strong stock-market gains have pushed portfolios away from their intended allocations.
• Diversification should extend beyond sectors: Europe and Asia offer different earnings, policy and currency drivers at a time when continued US and dollar outperformance should not be taken for granted.

The bar has moved higher

Growth is still proving resilient. The IMF expects the global economy to expand by around 3% in 2026, corporate earnings remain supportive and AI investment continues to add momentum even as higher borrowing costs work their way through the economy. But the longer yields remain elevated, the greater the risk that financing pressure eventually weighs on households, companies and investment.

Inflation is becoming harder to dismiss at the same time. US headline inflation rose to 3.4% in August, oil climbed back above $100 a barrel and geopolitical disruptions have added another supply-side risk just as central banks were hoping inflation would become less complicated.

That leaves monetary policy unusually open-ended. One or two additional Federal Reserve hikes would matter, but they are not the main risk. Markets can adjust to individual 25-basis-point moves. The bigger question is whether persistent inflation means the hiking cycle itself is not yet finished. That possibility has moved rapidly from a tail risk to the centre of the debate.

If a broader tightening cycle is required, the investment consequences go well beyond the next Fed meeting. Long-term yields could remain high, refinancing costs continue to rise and the return investors should demand from equities could move higher with them. Earnings remain supportive, but with the risk-free rate above 5%, valuation expansion has less room to do the work. Q4 equity returns may increasingly need to be earned through profit growth and cash flow.

The November US midterms will add uncertainty around fiscal and trade policy, but for longer-term investors, the more important issue is whether policy changes the trajectory of deficits, inflation and yields rather than the election result itself.

Our Q4 view is therefore not to retreat from markets. It is to raise the bar for the risks investors choose to own.

AI meets the cost-of-capital test

In Q3, we divided the AI opportunity into builders, users and efficiency winners. That framework helped identify where spending was flowing.

The AI investment cycle is entering a more demanding phase. So Q4 question is no longer who can spend the most, but who can earn an adequate return on that spending when the cost of capital is so much higher.

With bond yields above 5%, investors have a credible alternative to equity risk. That raises the hurdle rate for long-duration investments and makes the economics of AI capital expenditure harder to ignore.

Four issues stand out:

• Who can self-fund the race? Cash-rich companies funding AI from operating cash flow are better positioned than those relying heavily on debt, leases or external capital. With yields high, financing structure matters again.
• How quickly does AI investment become obsolete? Rapid advances in chips and models can shorten the useful life of infrastructure. That supports recurring upgrade demand for suppliers, but raises depreciation and reinvestment risk for the companies funding it.
• Is capex still creating growth? Some spending should open new revenue streams, but part of it may increasingly become maintenance capex — money that has to be spent simply to remain competitive.
• Who can benefit with less capital? Software companies and AI adopters may capture productivity gains without bearing the same infrastructure costs as hyperscalers.

AI's obsolescence problem

The speed of technological change creates an unusual challenge for AI infrastructure.

Traditional infrastructure can often be financed against decades of useful economic life. AI assets, particularly compute, can face much faster technological depreciation. Chips improve, model architectures evolve and more efficient ways of training and running models emerge. That can shorten the useful economic life of investments and force companies to reinvest before their previous spending has generated the expected return.

This makes the distinction between growth capex and maintenance capex increasingly important.

If every major platform has to buy the newest chips, train larger models and expand compute simply to maintain its competitive position, part of today's extraordinary AI investment may eventually become the cost of staying in the game.

For investors, the critical question is not how much a company spends on AI, but whether it can generate durable cash flows faster than technological changes force it to reinvest.

What could slow the AI investment cycle?

The bigger risk to the AI investment cycle is not necessarily a collapse in demand, but a longer path from investment to monetisation.

That could come from slower model progress, additional testing and safety requirements, export restrictions, legal challenges, infrastructure constraints or simply slower adoption by businesses and consumers.

Any delay matters more when the cost of capital is high. Longer payback periods make it harder for companies to justify very large upfront investments and could increase scrutiny of whether AI spending is genuinely creating growth or simply becoming necessary to remain competitive.

At the same time, higher fixed costs and greater complexity could favour the largest incumbents, which have the balance sheets, computing capacity and distribution networks to absorb them. A slower frontier-model race could also shift attention towards extracting value from the AI capabilities that already exist.

What this could mean for portfolios

The case for maintaining core AI exposure remains, but Q4 calls for greater selectivity around capital intensity.

That means looking beyond companies that benefit simply because AI capex keeps rising and focusing instead on:

• companies capable of self-funding investment;
• selective software businesses that can monetise AI without enormous physical capex;
• financials, industrials and healthcare companies that can turn AI into lower costs or higher productivity;
• businesses where AI strengthens an existing competitive advantage rather than merely defending one.
• equal-weight US indices which continue to make sense alongside market-cap-weighted exposure

This is not a call for mega-cap tech to collapse. The infrastructure buildout still has further to run. But if AI gradually shifts from a capex story to a productivity story, earnings leadership should have more room to broaden.

Equal weight strategies offer a simple way to participate in that broadening without abandoning the US market.

Build an inflation sleeve

Q4 inflation risk is no longer simply about whether oil rises or falls. Energy has become a geopolitical variable again, with Middle East disruption adding a risk premium to crude just as the Northern Hemisphere heads into winter and energy inventories may need to be rebuilt. That combination could keep oil and gas markets tighter than usual and feed back into inflation expectations and bond yields.

Where the Middle East goes from here still matters. Further escalation or prolonged disruption around key shipping routes and export infrastructure could keep energy prices elevated, while a credible de-escalation could remove part of that premium relatively quickly. That makes energy an important, but volatile, part of the inflation picture rather than a one-way inflation hedge.

Several independent forces could keep inflation and long-term yields elevated.

The main ones are:

• Oil and geopolitics: Brent has moved back above $100 as Middle East disruption threatens supply, feeding quickly into inflation and Fed expectations.
• El Niño: NOAA says there is now a greater than 90% chance of a very strong El Niño through the Northern Hemisphere autumn and winter, raising the potential for disruption to agricultural production, food prices and power markets.
• Fiscal spending: Defence, infrastructure and energy-security programmes support activity but also increase government financing requirements and competition for capital.
• AI infrastructure: Data centres are adding demand for electricity, copper, cooling, construction, grids and skilled labour.

AI creates an important contradiction. It may ultimately reduce inflation through higher productivity, but building the infrastructure required to deliver that productivity is resource-intensive today.

Our view is not that inflation will necessarily accelerate from here. It is that portfolios should not rely on a smooth return to 2%.

What this could mean for portfolios

Instead of relying on a single inflation hedge, consider a diversified inflation sleeve:

• Gold for fiscal, monetary and geopolitical uncertainty;
• energy for direct supply shocks;
• power and infrastructure for structural electricity demand;
• selected commodity producers exposed to physical bottlenecks;
• pricing-power equities that can protect margins;
• inflation-linked bonds where appropriate.

These positions will not all perform well at the same time. That is precisely why the sleeve should be diversified.

Bonds pay again, and that changes the maths

For much of the previous decade, bonds provided little income and investors were pushed towards equities to generate returns.

That is no longer the market we are in.

The US 10-year Treasury yield touched 5% on 14 September 2026 for the first time since 2023, reflecting inflation concerns, still-resilient growth, fiscal pressures and heavy government and corporate issuance.

This does not make equities unattractive. Over long horizons, equities remain important for participating in economic and earnings growth.

But equities now have competition.

When high-quality bonds can again provide meaningful income, investors should reassess whether every additional unit of equity risk is still warranted.

That is especially relevant after years of strong stock-market performance. A portfolio that began with a balanced allocation may now have a significantly higher equity exposure simply because equities have outperformed.

What this could mean for portfolios

• Check portfolio drift. Strong equity gains may have pushed the portfolio materially above its intended equity allocation.
• Rebalance rather than trying to call the market top. Rebalancing is about returning to an intended risk level, not forecasting a crash.
• Use bonds for income and stability. Short- and intermediate-duration high-quality bonds can provide useful carry without requiring a large bet on long yields falling.
• Consider when the money will be needed. Investors expecting to draw more regularly from their portfolio in the coming years may place greater value on predictable income and lower volatility than investors with much longer horizons.

Long-duration government bonds can still perform strongly if growth weakens sharply and inflation falls. But with inflation uncertainty still elevated, investors do not need to make a large duration call to make bonds useful again.

Diversify the earnings and currency engine

US equities and the US dollar have rewarded investors for years, supported by strong earnings, deep capital markets and global technology leadership.

But concentration remains concentration even when it has worked.

The dollar can still strengthen if the Fed keeps hiking, US yields remain high or risk sentiment deteriorates. At the same time, the trade-weighted dollar is around 8% below its previous peak, and fiscal concerns plus improving prospects elsewhere leave open the possibility of further diversification away from the currency.

Investors do not need to make a dramatic dollar call to reduce that concentration.

Different regions already provide different earnings and policy drivers:

• Europe: financials, industrials, infrastructure and defence offer exposure to a different fiscal cycle.
• Japan: higher domestic rates and reflation can support financials and selected domestic businesses.
• Korea and Taiwan: remain important beneficiaries of AI hardware and memory demand, but are increasingly cyclical and exposed to any moderation in AI capex.
• China: technology, AI, advanced manufacturing and biotech remain more compelling than relying on a broad property recovery. Tighter US AI regulation could also improve China’s relative position.
• Singapore and broader Asia: financials, income and domestic growth provide still different earnings drivers.

Five portfolio checks for Q4

Q3 was about keeping AI while building a stronger portfolio around it.

Q4 is about recognising that the bar has moved higher.

AI has to prove that extraordinary levels of investment can still generate attractive returns when financing is expensive and technology depreciates quickly. Equities now have to compete with bonds that offer meaningful yields. Inflation protection has to account for more than just oil. And diversification needs to extend beyond sectors to include currencies and economic drivers.