AI Spending Is Fuelling Inflation. Here's What Expats Should Do

AI Spending Is Fuelling Inflation. Here's What Expats Should Do

Most expats are watching AI as a technology story. They should be watching it as an inflation story.

The four largest technology companies deployed over $380 billion in infrastructure spending in 2025 alone. Goldman Sachs projects hyperscaler capital expenditure could exceed $600 billion in 2026 and $1.15 trillion cumulatively through 2027. That kind of spending, increasingly debt-funded rather than earnings-funded, is beginning to flow through into chip prices, energy costs, and credit markets in ways that have direct consequences for your portfolio. This post breaks down the mechanism, the risks most advisors aren't flagging, and the practical steps worth considering now.

How $600 Billion in AI Spending Becomes Your Inflation Problem

The spending numbers have moved well beyond what most financial commentary has caught up with. Amazon is leading at $125 billion in 2025 capital expenditure, followed by Microsoft at $96 billion, Alphabet at $91 billion, and Meta at $70 billion. Goldman Sachs now projects that 2026 hyperscaler capex could reach $600 billion or more, up from a cumulative $477 billion across the three years from 2022 to 2024 combined.

What matters for your portfolio is not the headline figure. It is how that spending is being financed and what it is doing to input costs across the economy.

The Shift From Earnings-Funded to Debt-Funded

In the early phase of AI build-out, hyperscalers were funding infrastructure largely from operating cash flows. That is no longer the case. Bank of America data shows these companies borrowed $75 billion in just the final two months of 2025 for data centre spending. Hyperscaler debt issuance hit $121 billion in 2025, four times the historical average.

J.P. Morgan estimates the full data centre build-out will require $1.5 trillion in investment-grade bonds over the next five years. When companies of this scale shift to debt financing at this pace, it competes for capital across credit markets and puts upward pressure on corporate borrowing costs broadly.

The Energy Bottleneck

Goldman Sachs estimates AI will account for 19% of total data centre power demand by 2028. New power generation capacity takes years to build. Chip deployment takes months. The structural mismatch between those two timelines creates a cost bottleneck that does not resolve quickly.

For an Australian expat in Singapore holding a diversified portfolio with significant long-duration bond exposure, this is the precise mechanism that erodes real returns quietly over time.

The Rate-Cut Reversal Risk Nobody Is Pricing In

U.S. inflation as of early 2026 sits at approximately 2.8%, still above the Federal Reserve's 2% target. The Fed resumed cuts of 50 basis points total in September and October 2025 after a nine-month pause. The assumption baked into most expat portfolios right now is that rate cuts continue smoothly. That assumption deserves scrutiny.

Aviva Investors has named the risk of central banks ending rate-cutting cycles, or even beginning to hike again, as one of the principal market risks for 2026, citing building price pressures from AI investment combined with government stimulus in Europe and Japan. Carmignac portfolio manager Kevin Thozet has increased holdings of inflation-protected Treasuries specifically because he views inflation risk as significantly underappreciated given the current growth cycle.

The BNP Paribas debate framing is useful here: if AI raises the neutral rate, premature rate cuts could ignite inflation expectations. If central banks wait too long to respond, growth stalls. Neither scenario is straightforward for a portfolio built around the assumption of declining rates.

What This Means for Long-Duration Exposure

Long-duration bonds are among the most rate-sensitive assets in a portfolio. If the rate-cut trajectory stalls or reverses, those positions take the sharpest hit. Mercer's Julius Bendikas, who oversees $683 billion directly and advises on $16.2 trillion, has publicly stated he is moving out of debt markets most exposed to an inflation shock. That is not a fringe view. That is the largest institutional money in the world repositioning.

As discussed in The Hard Truth About Market Returns You Can't Time It But You Can Prepare For It, the goal is not to predict where rates go. It is to structure a portfolio that does not rely on a single macro outcome to hold together.

The Revenue Gap: The Risk Hiding Behind the Capex Numbers

Here is the part of the AI spending story that most investors are not seeing. Sequoia Capital has tracked a growing shortfall between infrastructure investment and actual AI revenue generation. That gap now sits at an estimated $500 to $600 billion, up from $125 billion in 2024.

Using a standard 10-year asset life assumption for data centres, depreciation costs alone are running at roughly double current AI-related revenues, estimated at $15 to $20 billion annually. To break even at a 25% gross margin would require scaling AI revenues to $160 billion per year. That revenue does not yet exist.

Deutsche Bank estimates AI data centre capex could reach $4 trillion by 2030, warning that rapid expansion risks creating bottlenecks in chips and electricity that could send costs spiralling. HP has specifically flagged that memory chip cost inflation will pressure prices and profits later in 2026 as data centre demand continues to surge.

Why Concentration Risk Is the Practical Problem for Expat Portfolios

AI-related stocks have accounted for 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth since ChatGPT launched in 2022. A globally mobile professional who has been accumulating a simple S&P 500 index position as their primary investment vehicle has, without necessarily intending to, built a portfolio with substantial concentration in this single theme.

If the revenue gap does not close as expected, or if the rate environment shifts, the repricing risk in that concentrated position is material. Think You're Diversified? Think Again: A Guide for High-Income Expats covers this structural problem in detail and is worth reading alongside this post.

What Portfolio Managers Are Actually Doing Right Now

This section is deliberately practical. Here is what institutional money is moving toward, based on publicly stated positions from named managers.

Inflation-protected bonds. Carmignac is actively increasing holdings of TIPS as inflation risk re-accelerates. For a British expat earning in GBP but invested in USD-denominated assets, TIPS provide a layer of real-return protection that nominal bonds do not.

Rotating away from long-duration debt. The Mercer repositioning noted above is the clearest institutional signal on this point. Duration risk is being reduced, not extended.

AI infrastructure over pure software. Goldman Sachs notes that the average stock in its AI infrastructure basket returned 44% year-to-date in 2025, while software and services stocks have underperformed due to unproven AI-enabled revenue. If you want AI exposure, the evidence currently favours physical infrastructure over software-layer bets.

Commodities and real assets. Power demand from AI is driving investment in nuclear and renewables. Energy infrastructure exposure gives a portfolio a direct link to a theme that benefits regardless of which AI software applications ultimately win.

Broad diversification away from index concentration. This is not a call to exit equities. It is a call to review whether your equity exposure is genuinely diversified or effectively a leveraged bet on a single cycle.

Frequently Asked Questions

Q: Does AI investment actually cause inflation, or is this just media noise? A: The mechanism is real. AI capex creates demand for chips, electricity, and construction at a pace that supply chains cannot immediately match. When demand exceeds supply in inputs like High Bandwidth Memory and grid power, costs rise. Deutsche Bank has formally modelled this risk through to $4 trillion in cumulative spending by 2030. It is not media noise. It is a structural supply-demand imbalance with a multi-year timeline.

Q: Should expats be reducing equity exposure because of AI inflation risk? A: Not necessarily reducing, but reviewing. The question is not whether to hold equities. It is whether your equity exposure is sufficiently diversified or heavily concentrated in a theme that has driven the majority of recent index returns. Concentration risk and inflation risk are two separate problems that currently overlap in AI-heavy portfolios.

Q: What are TIPS and are they accessible to expats outside the US? A: TIPS, Treasury Inflation-Protected Securities, are US government bonds where the principal adjusts with inflation, providing a guaranteed real return. They are accessible to non-US investors through UCITS ETFs domiciled in Ireland or Luxembourg, which are available to most expats in Southeast Asia without the tax complications of direct US-domiciled fund ownership. Tax treatment varies by your country of residence.

Q: How does the AI capex debt boom affect corporate bond allocations? A: Hyperscaler debt issuance hit $121 billion in 2025, four times the historical average, with J.P. Morgan projecting $1.5 trillion in new investment-grade bonds over five years. That volume of new issuance competes for capital across credit markets, which can push up yields and push down prices on existing corporate bonds. If your portfolio holds significant corporate bond exposure, the duration and credit quality of those positions deserves a review.

Q: What is the practical risk if central banks reverse rate cuts? A: Long-duration bonds lose value as rates rise. A portfolio built around the assumption of continued rate cuts is exposed if that assumption proves wrong. Aviva Investors has named rate-cut reversal as one of the main market risks for 2026. The practical response is to reduce duration exposure and ensure you hold assets, like real assets, commodities, or inflation-linked bonds, that perform differently when rates move higher.

Q: Is the AI revenue gap a real risk or are the hyperscalers just investing for the long term? A: Both things can be true. The hyperscalers may ultimately generate returns that justify the spend. But the current gap between infrastructure cost and revenue, estimated at $500 to $600 billion by Sequoia Capital, means the assumption that AI spending automatically translates into earnings growth is not yet supported by the numbers. For portfolio construction, it is prudent to hold that uncertainty rather than price in the optimistic outcome as certain.

If you are an expatriate professional reviewing your portfolio in light of these shifts, the next step is a straightforward conversation. No pitch, no pressure. Just clarity on where you stand and what your options are. Book a no-obligation call with Ciprian.

This content is for informational purposes only and does not constitute personalised financial, investment, or tax advice. By reading this post, you agree to our disclaimer.