NextFin News - The S&P 500 is trading near record highs above 7,700, the US economy keeps confounding recession callers, and the person steering one of the world's largest private banks' investment strategy says the rally has further to run — not because of easy money or fiscal sugar, but because artificial intelligence is finally doing what economists have waited decades for: lifting productivity. Willem Sels, global chief investment officer at HSBC Private Bank and Premier Wealth, told Bloomberg Television on Tuesday that the US economy and corporate sector have proved "more resilient than people thought," and that further AI-driven productivity gains are a core reason he remains "more bullish on the US than on Europe."
Further AI-driven productivity gains are a core reason he remains "more bullish on the US than on Europe," Sels said, framing the 2026 bull market as a productivity story rather than a liquidity story.
The claim matters because it reframes the entire 2026 bull market. If AI is merely a capex boom, the cycle ends when spending slows. If it is a genuine productivity shock, earnings can grow faster than wages and margins can hold even as the labour market tightens — and the valuation multiples investors are paying today stop looking stretched. Sels is betting on the latter.
The Backdrop: A Rally Underpinned by Spending, Not Just Sentiment
The backdrop is unusually favourable for a productivity-led equity thesis. The S&P 500 closed at 7,718.60 on 4 September 2026, up 18.71% over the prior year, according to S&P Dow Jones Indices. The SPDR S&P 500 ETF closed at $770.19 that day, little changed from its August peaks. Volatility has been absorbed without breaking the trend.
HSBC's stance is not a lone voice. In its Q4 2026 Perspectives outlook dated 3 September, Sels wrote that "AI adoption continues to accelerate, with more powerful models and new applications fuelling investment and market optimism." Critically, he argues the benefits are no longer confined to the chip and cloud giants: "The benefits aren't just seen in the tech sector, as AI also increases demand for electricity, materials and financing. Robotics, automation, healthcare, consumer services and other industries are also benefitting from AI-powered innovation."
The scale of the spending is the engine. HSBC estimates total AI capital expenditure will top $1 trillion in 2027, up from $735 billion in the first quarter of 2026 — a roughly 36% increase in under two years. That is the demand side. The supply side is the productivity payoff that Sels says is only now beginning to flow through the data.
The bank's positioning backs the words. HSBC says it has "recently added exposure to global equities, favouring the US and Asia," while maintaining a broad sector allocation. Within technology, it prefers cloud, semiconductors and AI-enabled applications "where monetisation and revenues are clear and sustainable."
But there is a tension worth naming immediately. The same HSBC that is bullish on US stocks has, at points this year, trimmed the magnitude of its overweight on US equities to mitigate valuation risk. And while Sels favours the US over Europe, a growing chorus of strategists argues Europe is where the relative value now sits. The question this piece answers: is the AI-productivity trade a structural regime shift that justifies US outperformance, or a cyclical capex wave that will revert once the spending peaks?
Why Productivity Changes the Earnings Math
The first-order effect of AI is straightforward: hyperscalers and chipmakers book the capex. The second-order effect is what Sels is really pointing at. When firms deploy AI to automate routine work, accelerate research and development, or optimise supply chains, output per hour rises without a matching rise in headcount or wages. In national accounts terms, total factor productivity accelerates. For equity investors, that is the difference between a bubble and a boom.
A capex bubble ends badly because spending creates capacity that never earns its keep — the fibre-optic overbuild of the late 1990s is the textbook case. A productivity boom compounds because each dollar of investment lowers the cost of producing the next dollar of revenue. Margins expand, return on invested capital rises, and — this is the key — the rally becomes earnings-driven rather than multiple-driven.
Sels' own framing supports the second reading. He notes that "the US tech sector is trading at much lower price-to-earnings multiples than last year, as the market rally has been underpinned by earnings growth. So we don't think that the tech sector is expensive." If true, the most hated part of the rally is the part with the best fundamental cover.
Cyclical Capex Wave or Structural Regime Shift?
This is a structural shift riding on top of a cyclical upswing — and the distinction decides the conclusion.
The cyclical leg is real and mean-reverting. AI capex, like any capex cycle, will have a peak. When the $1 trillion annual spending figure is reached, growth in that spending must slow even if the level stays high. Order backlogs normalise. Chip lead times ease. The hyperscalers' capital intensity plateaus. Any investor positioned purely for perpetual capex acceleration will get hurt at that inflection. History offers three clear analogues: the railway mania of the 1840s, the electrification boom of the 1920s, and the internet fibre build-out of 1998–2000. In each, the infrastructure builders saw violent re-ratings once deployment peaked, even though the underlying technology transformed the economy.
The structural leg is different, and it is why Sels' US-over-Europe call has a durable foundation. General-purpose technologies — steam, electricity, the microprocessor — do not revert because they change the production function itself. The evidence that AI belongs in this category is accumulating: adoption is spreading beyond the tech sector into robotics, healthcare, consumer services and materials; the spending is flowing into enabling infrastructure — electricity generation, data centres, grid upgrades — that will be used for decades; and the productivity gains are beginning to show up in corporate guidance rather than just vendor roadmaps. HSBC's own Q1 outlook captured this. Americas CIO Jose Rasco said:
"Productivity gains from accelerating AI adoption are only beginning to flow through. As AI deepens its models and expands globally, innovation is extending its life and broadening the opportunity set beyond mega tech."
So the judgment: the capex cycle is cyclical and will peak; the productivity regime is structural and will not revert on its own. The investable implication is to own the enablers and the monetisers, not just the builders.
The Second-Order Trade the Market Hasn't Fully Priced
The consensus view is that AI capex benefits the Magnificent Seven and the semiconductor supply chain. That trade is crowded, priced, and well understood. The second-order trade is the one Sels is gesturing toward: the broadening of the AI payoff into the sectors that sit downstream of the build-out.
Electricity is the cleanest example. Data centres do not run on optimism; they run on gigawatts. AI-driven demand for power is pulling forward investment in generation, transmission, and grid equipment — a capital-intensive, regulated, backlog-rich industry that has nothing to do with software margins. Defence is another. HSBC notes global defence order backlogs reached $2.9 trillion in 2025 and are still rising; modern militaries are increasingly software- and autonomy-dependent, and the same AI supply chain serves both commercial and security demand. Financials, materials, and energy round out the list.
This is where the US-over-Europe call gets its teeth. The US has the hyperscalers, the chip designers, the venture capital, and — critically — the energy grid and capital markets capable of funding a multi-year build-out. Europe has excellent industrial champions in grid equipment, defence, and automation, and Goldman Sachs strategists have argued the STOXX Europe 600 has actually outperformed the S&P 500 since the start of 2025 — 54% versus 34% in dollar terms as of mid-August. But Europe lacks the scale of the AI platform companies that capture the highest-margin layer of the stack. Sels' call is not that Europe will fall; it is that the US captures more of the value added.
The Strongest Counter-Thesis — and What Would Break It
The strongest counter-thesis comes from valuation discipline, and it deserves a full hearing. HSBC itself has trimmed its US overweight at points this year "to mitigate valuation risks." The argument runs like this: the S&P 500 at roughly 7,700, up nearly 19% in a year, already embeds a large share of the productivity payoff. Forward earnings estimates are at record highs. If AI-driven productivity arrives more slowly than the capex cycle peaks — if the $1 trillion spending figure proves to be the high-water mark rather than a stepping stone — then multiples have nowhere to go but down, and the index could fall even if earnings keep growing.
There is also the interest-rate channel. HSBC's base case is that "a US rate hike is unlikely in the near term, as long as high energy prices aren't translating into higher core inflation." That is a conditional statement, and the condition is the risk. AI data centres are power-hungry; power is energy-intensive; if energy prices re-accelerate into core inflation, the Federal Reserve's hand is forced, discount rates rise, and the long-duration growth stocks at the centre of the AI trade compress. The equity rally and the productivity thesis are both levered to the same assumption: benign inflation.
Sels' answer to this is embedded in his sector preference: favour companies "where monetisation and revenues are clear and sustainable," and diversify into bonds, gold, infrastructure and alternatives. In other words, the bank is not betting the house on a single linear extrapolation. It is expressing a conviction while hedging the conditionals.
The falsifying signal is specific and observable. Watch US non-farm productivity — output per hour — and unit labour costs over the next four quarters. Productivity grew just 1.2% a year on average from 2010 to 2019, according to Congressional Research Service calculations of Bureau of Labor Statistics data, and the most recent print — 1.4% in the second quarter of 2026 with unit labour costs up 1.3% — is only modestly above that. If productivity fails to accelerate meaningfully above the 1.2% post-2010 norm while unit labour costs rise faster than 3% for two consecutive quarters, the structural-productivity thesis is wrong and the rally is a capex cycle heading for its peak. At that point, the defensive posture — quality bonds, gold, non-US diversification — becomes the right trade.
Who Benefits, Who Is Exposed, and What to Watch
The mechanism cashes out into a clear set of beneficiaries and exposures.
Who benefits. The enablers: cloud infrastructure, semiconductors, and AI-enabled applications with visible monetisation — HSBC's stated preferred layer of the stack. The infrastructure adjuncts: electricity generation and grid equipment, data-centre construction, copper and other materials. The capital providers: financials that underwrite the build-out. Across regions, the US remains the primary beneficiary, with Asia — particularly the chip supply chain and data-centre expansion — as the second.
Who is exposed. Pure-play capex beneficiaries with no path to recurring revenue face the peak-cycle risk. Companies that priced in perpetual double-digit AI spending growth are vulnerable to the first earnings miss that suggests deceleration. Europe is not "exposed" in a bearish sense — its industrial champions may well outperform on relative value — but it captures less of the highest-margin software and platform layer.
Time horizons. Short term, over the next quarter: sentiment and liquidity dominate; the index can move on rate expectations and energy prices regardless of the productivity story. Medium term, six to eighteen months: earnings delivery decides. If third- and fourth-quarter 2026 results show AI monetisation spreading beyond tech, the rally broadens and the US overweight is vindicated. Long term, three to five years: the structural question resolves. If AI lifts trend productivity to 2% or above, US equities earn their multiples through earnings growth rather than multiple expansion, and the current levels look cheap in hindsight.
Scenarios. Base case: productivity accelerates modestly, capex growth slows but stays elevated, and the S&P 500 grinds higher on earnings rather than multiples. Upside case: AI monetisation surprises to the upside across healthcare, robotics and enterprise software; productivity breaks above 2.5%; the rally broadens beyond mega-cap tech. Downside case: capex peaks in 2027 as forecast but productivity fails to show in the data; unit labour costs accelerate; multiples compress 15–20% even as earnings hold.
What to watch. Four signals: first, US productivity and unit labour costs, the falsifier named above; second, hyperscaler capex guidance for 2027 — any cut is the first warning of a peak; third, core inflation and energy prices, which gate the Federal Reserve's next move; fourth, the breadth of earnings growth outside the top 10 S&P 500 names.
Sels' central judgment can be compressed into one line: this is not a rally betting that AI will change the economy someday — it is a rally betting that AI has already started changing the earnings math today. If the productivity data confirms it, the US overweight survives the next capex peak. If it does not, the peak will look, in retrospect, like the moment the cycle turned.
Explore more exclusive insights at nextfin.ai.
