The Only Easy Day Was Yesterday

The only easy day was yesterday. That may seem like an unusual way to describe a year shaped by ongoing energy disruptions, rising trade uncertainty and a rapidly expanding artificial intelligence (AI) investment cycle. But the forces supporting markets are increasingly colliding with the risks they were expected to overcome. Renewed threats to global shipping are reviving inflation concerns, tariffs have returned to the centre of the market narrative and advances in Chinese technology are challenging some of the assumptions underpinning the AI investment cycle.

Nowhere has the pattern been clearer than in energy markets. Investors have repeatedly treated de-escalation in the Middle East as evidence that the energy shock was nearing an end. Oil prices fell, inflation concerns receded and attention shifted back towards resilient growth and AI. Yet, reopening the Strait of Hormuz is not the same as securing it. Repeated shifts between disruption and de-escalation have widened the range of possible outcomes, with markets placing a higher probability that normalization fails again. That uncertainty might persist for years until alternative infrastructure can reduce the world’s dependence on shipping through Hormuz. The world will eventually become more resilient to such shocks but also less efficient.

That transition is not necessarily more bearish, but it is less forgiving. Long-term bond yields are increasingly being shaped by fiscal supply and term premia as much as monetary policy (graph 1). Equity markets continue to reward firms positioned to benefit from AI and other critical network buildouts, but concentrated leadership and elevated expectations leave less room for disappointment. Meanwhile, recurring geopolitical and trade disruptions make it harder for markets to treat periods of relief as permanent. The global economy may have avoided recession, but the measures required to absorb yesterday’s disruptions have made the next stage of the cycle more expensive.

This environment remains difficult for fixed income. The AI investment cycle is becoming a key source of growth in the United States and other countries closely connected to that network (graph 2). Spending tied to resilience and AI infrastructure has offset some of the drag from higher energy prices, limiting the slowdown in economic activity. But that has also raised the threshold for a conventional bond rally. Without a more decisive deterioration in growth, inflation uncertainty and elevated financing requirements could continue to outweigh the appeal of duration.

More recently, bond yields have been shaped by rising inflationary pressures emerging from several directions. Higher energy prices are lifting transportation and production costs, tariffs are increasing the price of traded goods and the AI buildout is generating substantial demand for equipment, power and infrastructure. These common pressures have increased the co-movement of bond yields across major jurisdictions, particularly at the long end of the curve (graph 3). But common pressures need not produce common policy responses. For monetary policymakers, the key questions are whether these pressures persist, how broadly they spread and the starting point for underlying inflation.

Both exposure to these inflation pressures and the starting point for policy vary considerably across regions. Europe remains the most exposed to energy disruptions, although its more limited participation in the AI investment cycle reduces that source of demand pressure. North America is relatively insulated from the energy shock, but the United States is more exposed to price pressures from tariffs and the AI buildout. US inflation is also starting from a higher level (graph 4), meaning it would take a sustained run of soft readings for the Federal Reserve’s preferred measure to return to its 2% target. Canada has a more favourable starting point. Underlying inflation is considerably softer, and unusually strong readings would be required to push these measures above 3%.

These common inflationary forces are therefore unlikely to produce common policy responses across regions. We expect the ECB to deliver another rate increase later this year, while the Bank of England is also likely to resume tightening. Those forecasts are broadly consistent with market pricing. Our main disagreement with market pricing is in North America, where investors continue to see rate increases from both the Federal Reserve and the Bank of Canada. While our Fed call is in jeopardy, we still expect both central banks to remain on hold for the rest of the year.

Canadian bonds continue to outperform on a relative basis in this environment. Across the curve, yields on Canadian bonds have risen less, likely reflecting less pronounced inflationary pressures and softer economic growth (graph 5). Investors also continue to require relatively less compensation to own Canadian long-term debt (graph 6). That advantage is particularly evident compared to the United States, where stickier inflation, larger fiscal deficits and uncertainty surrounding the Federal Reserve’s strategy have added to volatility at the long end of the Treasury curve.

The US dollar remains on stable footing following a bout of weakness in 2025. Renewed disruptions in the Middle East have supported demand for liquidity, reinforced expectations that US interest rates could remain higher for longer and highlighted America’s relative energy security. Together, these forces have pushed the greenback higher against most major currencies, with the dollar index recently trading near its 2026 high.

But the source of US dollar demand is changing. Foreign investors have become more selective about holding US government debt as large fiscal deficits, heavy issuance and the use of financial sanctions encourage reserve diversification. At the same time, capital continues to flow towards US equities at the centre of the AI investment cycle. That’s supportive of US dollar strength, but potentially less durable. Unlike official reserve demand, equity flows are more sensitive to earnings, valuations and the ability of US firms to capture the returns from AI investment. Recent advances by Chinese technology companies could challenge the margins and scarcity premiums embedded in US technology valuations, weakening an important source of dollar demand if equity leadership begins to fade. For now, however, higher yields, relative energy security and continued demand for US equities make a sustained decline in the greenback difficult to envision.

The Canadian dollar remains constrained by trade uncertainty, and we expect limited appreciation this year. Avoiding a breakdown in CUSMA negotiations is not the same as restoring long-term certainty. The latest tariffs on Canadian goods were broadly consistent with our expectations, and negotiations are likely to remain volatile before a more durable agreement can be reached. Higher energy prices can provide support, but they are unlikely to fully offset the trade-related discount embedded in CAD. We expect the Canadian dollar to strengthen more meaningfully next year as trade uncertainty recedes and interest rate differentials between Canada and the United States narrow.

Equities can still perform well in this environment. Economic activity remains strong, particularly in the US, and that has been supporting earnings growth (graph 7). Firms connected to AI, energy, infrastructure and strategic supply chains stand to benefit directly from the resilience of the investment cycle. But the tradeoff is upward pressure on discount rates and greater dispersion (graph 8). The result is a market that can weather incoming shocks at the index level while becoming less forgiving beneath the surface.

The AI trade is also entering a more selective phase. Following an extraordinary rally, semiconductor stocks have experienced several bouts of heavy profit-taking. Investors are differentiating between companies supplying scarce AI infrastructure and those committing capital to consume it. Hyperscaler earnings illustrate the dilemma. Aggressive spending guidance supports demand for semiconductors, servers and data centres, but can weigh on hyperscaler valuations if investors become concerned about free cash flow and the eventual return on that investment. Rising Chinese competition adds to the wall of worries surrounding AI expenditures.

Leverage has made this transition disorderly in some jurisdictions. South Korea provides the clearest example. AI-related enthusiasm left the KOSPI increasingly concentrated in Samsung Electronics and SK Hynix, which at one point accounted for 50% of the Korean benchmark index. The use of single-stock leveraged products and retail margin borrowing further amplified the rally and the subsequent correction (graph 9). Regulators are now working on ways to improve financial stability. While Korea represents an extreme case, the same concerns apply to US markets with a rising share of leveraged ETF products concentrated in technology names.

The scale of the buildout is also making the AI trade more sensitive to long-term interest rates. Hyperscaler capital spending is expected to reach roughly US$700–800 billion in 2026, approximately double the amount spent in 2025, with a growing share being financed through debt markets. Hyperscalers have been tapping bond markets globally, typically favouring longer maturity issuance. While these investments should support earnings, they also increase demand for physical resources, which can contribute to inflation pressures. The additional bond supply can also lift term premia. Both can increase the discount rate applied to AI-related earnings, creating a circularity in which the investment supporting the cycle also makes valuations harder to sustain. So while valuations on some of these hyperscalers have been stabilizing or improving, their credit default swaps have been widening, reflecting this risk.

Financing requirements are increasingly visible in credit markets. In Canada, Amazon issued a record C$14 billion of Canadian-dollar debt in June, surpassing Alphabet’s C$8.5 billion transaction only one month earlier. Meta has also announced a more than C$13 billion data centre in Alberta, adding to expectations that technology companies will continue tapping Canadian capital markets to finance the buildout.

The challenge remains finding viable diversification away from the technology theme. Emerging markets have become more dependent on the same AI capex cycle supporting US equities, as evidenced by the recent price action. Canadian equities remain a key pillar of that diversification, and we expect continued strength going into the end of the year. However, headwinds from trade should continue to weigh on returns in the near term.