Global outlook: Volatile Repricing or Tipping Point?

Volatile Repricing or Tipping Point?

Markets are wrestling with a profound tension: the same AI boom that is extending the growth runway for parts of the global economy is also raising long-term borrowing costs, straining infrastructure and reviving questions around inflation, neutral rates and valuation discipline. The result is a market regime that looks less like a conventional risk-off episode and more like a repricing of capital in an economy being rebuilt around power, data centres, chips and software automation.

The key question is whether the recent turbulence marks the start of a materially deeper equity-market drawdown or a blip on the road to a more productive, AI-enabled economy. The better answer is that it is neither simple panic nor simple opportunity. It is a transition from a world in which falling inflation and easier policy supported both earnings and valuation multiples to one in which growth may remain solid, but the cost of capital is structurally higher.

The immediate trigger for the latest volatility has been the jump in long-dated yields, compounded by renewed Middle East tensions and rising oil prices. The 30-year US Treasury yield recently reached levels last seen in 2007 as investors digested larger fiscal deficits, heavier Treasury issuance, AI-related corporate borrowing and oil-led inflation risks linked to renewed Middle East tension. Long-end yields are not moving higher simply because of short-term Fed expectations; they are rising because investors are demanding a larger term premium to lend into an environment of persistent deficits, duration supply and policy uncertainty.

Treasury issuance remains the dominant source of duration supply, but AI financing has become a material marginal amplifier. Wall Street estimates of AI-related investment-grade issuance are centred on around $300 billion in 2026, potentially creating as much as $360 billion of 10-year-equivalent duration supply (around one-eighth of that generated by US Treasury issuance). The effect is magnified by long-dated corporate bond issuance, pay-fixed swap activity and the displacement of financial-sector issuers that would otherwise absorb duration.

That additional duration supply matters because equities, especially richly valued growth equities, are discounted against long-duration real rates. The old market equilibrium rested on the assumption that AI would deliver a productivity boom while inflation continued to glide lower and central banks could keep easing. The newer equilibrium is more complicated. AI may still boost earnings and nominal GDP, but the financing required to build the infrastructure behind that story is also contributing to higher yields.

This explains why equity volatility can rise even while the fundamental AI thesis stays intact. Markets are recalibrating the price they are willing to pay for future earnings, particularly when those earnings are long-dated, capital-intensive or reliant on generous assumptions about falling rates.

Blip or tipping point?

The base case is still a volatile repricing rather than the start of a severe bear market. Volatility was unusually low heading into the recent moves and AI capital expenditure is expected to add materially to US growth over the next two years. That is not the backdrop typically associated with an imminent collapse in corporate earnings.

At the same time, this may be more than a routine bout of market turbulence. A true tipping point would come from a feedback loop in which higher long-term yields raise funding costs, widen credit spreads, reduce the returns on data-centre and infrastructure projects, and then force companies to trim capex or lower earnings expectations. Should this dynamic become self-reinforcing, the AI theme could stop justifying premium multiples, and attention would turn instead to the burden on cash flow and balance-sheet resilience.

The distinction therefore lies in whether the economy remains in a world of strong nominal growth and contained spreads, or shifts into one where the supply of duration overwhelms demand and investment plans begin to buckle. For now, the evidence points more towards rotation and selectivity than systemic collapse. Markets may well correct further, but the more likely path is a narrower and more discriminating market rather than a uniformly lower one.

What the equity market is likely to reward

In this regime, the winners are less likely to be “AI” in the abstract and more likely to be companies that can convert AI spending into durable cash flow without overstraining their balance sheets. That includes cash-rich hyperscalers, selected semiconductor companies, grid and power-infrastructure beneficiaries, and software firms with visible pricing power and strong free-cash-flow conversion.

By contrast, the more vulnerable part of the market consists of long-duration equities whose valuations depend heavily on lower discount rates or distant monetisation. Companies tied to the AI supply chain but reliant on continual refinancing, speculative capacity expansion or uncertain returns on invested capital face a tougher environment if yields remain elevated. The market is moving away from a phase where thematic exposure alone was enough and towards one where capital discipline matters again.

That is why this period should be viewed as a test of market leadership rather than a blanket verdict on growth. If AI is as transformative as many expect, the underlying economic opportunity remains substantial. But equity markets will increasingly separate the owners of the economics from the owners of the narrative.

The regional divide

The AI build-out is also widening the gap between regional equity markets. The US retains the deepest exposure to hyperscaler earnings and AI-driven productivity, but it is also most exposed to the higher-yield, higher-valuation regime because the bulk of infrastructure spending and corporate duration issuance sits there. US AI investment is estimated at $581 billion in 2026, or about 1.8% of GDP, underscoring both the scale of the growth impulse and its sensitivity to a rising cost of capital.

Europe is less exposed to the direct AI capex cycle, and therefore less vulnerable to a US-style de-rating of richly valued, long-duration technology, but it is not insulated from the current bond-market volatility. Higher global real yields, with German Bund yields also at their highest levels in more than a decade, raise financing costs and weigh on already subdued investment and growth.

Europe’s opportunity is more indirect, through industrial automation, power networks, defence, cybersecurity and selective software; its constraint remains weaker trend growth and a much smaller AI investment base.

Asia is the clearest external beneficiary of the US-led capex boom. Korea and Taiwan sit at the centre of the semiconductor, memory, hardware and data-infrastructure supply chain, enabling their companies to capture AI demand at generally more reasonable valuations than US hyperscalers, albeit with greater cyclicality and geopolitical risk.China requires a more selective lens. Its private AI investment remains below that of the US, although private data understate the role of state-directed funding and industrial policy. The more compelling exposure is likely to lie in semiconductors, automation, power equipment and advanced manufacturing rather than in a direct replica of the US hyperscaler trade. China’s AI opportunity is therefore a supply-chain and productivity story as much as a platform-company story.

Where neutral rates may be heading

One of the most important implications of the current environment is that neutral rates may be higher than investors were accustomed to in the pre-pandemic period. AI-related investment demand and productivity expectations may be lifting estimates of the neutral rate, even as structural fiscal deficits and energy constraints add to aggregate demand.

A sensible working assumption is that the nominal neutral policy rate in the United States now sits closer to 3.5%–4.0% than to the 2.5%–3.0% range that dominated much of the previous cycle. In real terms, that implies something like 1.5%–2.0%, plus the Fed’s 2% inflation target. This is not a precise point estimate; it is recognition that a world of stronger investment demand, reindustrialisation, fiscal borrowing and power-hungry technologies is unlikely to support ultra-low real rates on a sustained basis.

This higher-neutral-rate framework matters for valuation. Should policy settle at a higher plateau than the ultra-low rates that once supported extreme duration multiples, investors will need to exercise greater valuation discipline, particularly when growth is deferred or capital-hungry.

Is AI inflationary or deflationary?

The best answer is that AI is inflationary in the build phase and potentially disinflationary in the diffusion phase. That sequencing is essential. At this stage, the AI boom is bidding up scarce inputs such as electricity, grid capacity, data-centre space, advanced semiconductors, construction services and long-duration financing. These pressures do not necessarily dominate the whole CPI basket, but they are economically meaningful because they raise the marginal cost of the infrastructure required to deliver future productivity gains.

Over time, however, AI should become disinflationary across a broader range of tasks. As firms redesign workflows around coding, research, customer service, compliance and administrative functions, unit labour costs in those areas can fall and total-factor productivity can rise. The near-term picture looks different: demand for energy, equipment, skilled technical labour and capital is rising faster than productivity gains are being realised. That balance should shift only once adoption broadens and today’s infrastructure bottlenecks—including power, chips and data-centre capacity—begin to ease.

Final market take

The most persuasive conclusion is that current volatility reflects a structurally important repricing, but not yet a decisive break in the equity bull case. The market is adjusting to the reality that AI may be enormously growth-enhancing while also making capital more expensive, neutral rates higher and capital allocation more important. That is uncomfortable for equities, but it does not invalidate the transformation thesis itself.

The practical implication is to avoid both extremes. It is too early to conclude that the AI boom is ending, and too complacent to dismiss the bond market’s message as noise. This is a new phase of the cycle: less about indiscriminate multiple expansion, more about productivity, pricing power, financing resilience and evidence that massive investment will actually earn an adequate return.

If long yields stabilise and credit spreads remain orderly, this episode is likely to look like a difficult but necessary repricing on the road to a more productive economy. If long yields keep climbing, oil stays high and AI capex begins to undermine rather than reinforce free cash flow, the probability of a deeper and more durable equity-market decline rises materially. For now, the balance of evidence favours volatile repricing over a tipping point. That assessment rests on long yields stabilising before higher financing costs begin to impair capex returns, corporate cash generation and ultimately confidence in the AI investment cycle.