GIC flags AI concentration as the single largest market risk. BlackRock can't sell bonds. Asia fund managers are rotating into laggards. The multi-trillion-dollar AI infrastructure buildout is funded by debt — and the first cracks are showing.
During the week of July 19-26, 2026, three separate signals from three separate corners of the financial world pointed to the same conclusion: the AI capital spending boom is being funded by borrowed money, and the market is starting to price that risk.
The signals came from unlikely sources:
GIC — Singapore's sovereign wealth fund, managing US$700+ billion and widely regarded as the world's longest-horizon investor — reported its 20-year annualised real return fell to 3.4%, the lowest in six years. The fund explicitly flagged "capital concentration in AI" as the single largest market risk. GIC has doubled down on AI investments, including stakes in Anthropic and Eli Lilly, but cautioned that "the sheer scale of AI capex could create valuation bubbles."
BlackRock — the world's largest asset manager, with US$9+ trillion under management — encountered soft demand for a bond sale, directly attributed to the "AI debt sell-off." When the entity that normally absorbs any paper struggles to place bonds, it's a market signal, not a headline.
Big Asia equity funds — the largest regional fund managers — began rotating out of AI exposure and into laggard sectors. Not a correction. A deliberate de-risking move.
These aren't bearish calls from contrarian commentators. They're positioning changes from the world's largest capital allocators. When GIC, BlackRock, and Asia's top fund managers move in the same direction simultaneously, it's worth understanding why.
The Business Times ran the headline that crystallises the week's theme on July 22: "AI is running on borrowed money." The analysis was blunt — big tech firms are turning to the bond market for staggering sums to finance data centres, because organic cash flows can't keep pace with the capital requirements of the AI infrastructure buildout.
The numbers are staggering. Google raised its capex guidance by US$15 billion to a US$195-205 billion range during its July 23 earnings call. Alphabet CEO Sundar Pichai defended the spending during the call, acknowledging delays in Gemini 3.5 Pro but highlighting rapid progress in Gemini Flash models. Cloud revenue surged 82% versus 64% estimates. Shares still fell 3% post-earnings.
The market heard: "We need to spend even more to stay in the race."
Nvidia signed a US$1.5 billion deal with Amkor for advanced chip packaging capacity — vertical integration to secure supply chains. Intel's forecast crushed estimates as the AI boom boosted chip demand, with shares jumping and the company boosting spending plans. AMD unveiled its Helios AI rack-scale system, a direct competitor to Nvidia's data centre offerings. Every player is spending more, and the gap between capex and revenue is widening.
| Company | Capex Signal | July 2026 Action |
|---|---|---|
| Google (Alphabet) | AI infrastructure | Raised guidance by US$15B to US$195-205B |
| Nvidia | Chip packaging capacity | US$1.5B deal with Amkor |
| Intel | AI chip manufacturing | Boosted spending plans, forecast beat |
| AMD | Rack-scale AI systems | Helios launch, direct Nvidia competitor |
The pattern is clear: every major player is accelerating capex while the revenue side of the equation remains uncertain. The debt markets are starting to notice.
Compounding the AI debt question is the macro environment. During the same week, oil prices surged above US$100 per barrel — driven by escalating US-Iran tensions, with 12 consecutive nights of US strikes on Iran and Houthi attacks on Saudi oil tankers in the Red Sea.
The impact was immediate and broad:
Wall Street fell on July 23 as tech earnings raised AI spending concerns and oil hit US$100. The Nasdaq sank more than 2%, the S&P 500 fell over 1%, and losses were broad-based. Investors were unimpressed by Q2 results from both Alphabet and Tesla.
European shares posted their biggest drop in two weeks on the oil surge and growing Fed rate hike expectations. The selloff hit cyclical and rate-sensitive sectors hardest.
Bond yields spiked to levels not seen since the 2008 crisis. UK gilts climbed above 5%. Germany's 10-year yield hit its highest level since 2011. The Bloomberg Global Treasury average yield reached 3.68% — the highest since 2008. Two-year US Treasury yields hit 17-month highs.
The Fed is being forced to reconsider. Fed Chair Warsh signaled a hawkish stance. Bond markets are actively pricing in rate hikes — the opposite of the rate-cut narrative that had been supporting equity valuations through much of 2026.
Why does this matter for AI capex? Because higher rates make debt more expensive. The entire AI infrastructure buildout is predicated on cheap capital. When bond yields rise, the cost of financing those US$195 billion capex plans goes up. The math that justified the spending at 4% debt service looks very different at 6%.
GIC's latest report deserves closer attention. The fund's 20-year annualised real return of 3.4% is the lowest in six years — a meaningful decline for an institution whose mandate is generational wealth preservation, not short-term outperformance.
GIC attributed the decline to "rising rates, geopolitical conflicts, and trade disruptions" turning global markets volatile. The fund launched a new investment framework to navigate the turbulence. But the specific language around AI is what stands out.
GIC reiterated its bullish stance on AI investments — it has stakes in Anthropic and is clearly positioned for AI upside. But it simultaneously flagged "capital concentration in the sector as the single largest market risk." This isn't a bear thesis. It's a risk management thesis from the world's most patient capital allocator.
Translate this into plain language: GIC believes AI is real and will generate returns. But the sheer amount of capital flowing into the same set of assets has created a concentration risk that could amplify losses if the AI investment cycle encounters headwinds. The headwinds, as GIC sees them, are rising rates (more expensive debt), geopolitical conflict (oil at US$100), and trade disruption.
When the entity managing Singapore's sovereign reserves — with a 20-year time horizon and no need to generate quarterly alpha — flags AI concentration risk, it's not a signal to exit. It's a signal that the risk-reward calculus is changing.
For investors with AI exposure, the convergence of these signals creates a specific set of risks and opportunities:
The debt maturity wall is real. Much of the AI infrastructure spending in 2024-2025 was funded by bonds issued when rates were falling and the growth narrative was pristine. Those bonds have maturity dates. When they come due in a higher-rate environment, refinancing costs will be significantly higher. Companies that built their AI thesis on cheap debt need to demonstrate that AI revenue can service more expensive debt.
Concentration risk is a portfolio construction problem. If AI represents a disproportionate share of your equity allocation — whether through direct holdings, index funds, or leveraged ETFs — the GIC warning applies to you. The Korea leveraged chip ETF carnage from last month (45% drawdown from debut, over 60% from peak) was a preview of what concentration + leverage looks like in a rotation.
The oil-inflation-Fed triangle changes every asset class. Oil at US$100 means imported inflation. Imported inflation means the Fed stays higher for longer (or hikes). Higher rates mean discount rates rise, which means future AI cash flows are worth less in present value terms. The entire AI valuation model is sensitive to the risk-free rate — and that rate is moving the wrong direction for growth stocks.
Asia fund managers are already rotating. The fact that big Asia equity funds are moving into laggard sectors suggests the smart money is positioning for a broader market regime shift, not an AI-specific correction. This is a defensive move, not a tactical trade.
The bull case for AI capex remains powerful. Google's cloud revenue surged 82%. Intel's forecast crushed estimates. AI chip demand is undeniable — every major player is racing to build capacity because the revenue potential is enormous.
Moreover, the alternative to building AI infrastructure now is being left behind. The strategic imperative to invest may outweigh the financial cost. Companies like Google and Nvidia aren't just building data centres — they're building defensible moats. The question isn't whether AI spending will continue; it's whether the returns will justify the debt.
Historically, infrastructure buildouts have always been cyclical. The telecom boom of the late 1990s was funded by debt and ended in a spectacular bust. But the internet survived — and the companies that survived the bust (Amazon, Google, Meta) became the most valuable in the world. The lesson: the technology is real, but the timing and financing matter enormously.
The current AI cycle shares similarities with that era: massive capex, heavy debt financing, and a technology that is genuinely transformative. The difference is that today's investors have the benefit of hindsight — they know how the last debt-funded infrastructure boom ended.
Three indicators that will tell us whether the AI capex reckoning is a blip or a structural shift:
1. Bond market pricing for AI debt. BlackRock's soft demand was a single data point. If AI-linked debt consistently trades at wider spreads or encounters placement difficulties, it signals the bond market is repricing AI risk. Watch the secondary market for tech corporate bonds.
2. Sovereign wealth fund positioning. GIC's 3.4% real return is a lagging indicator — its annual report comes once a year. More timely signals: watch GIC CEO remarks at investment conferences, and monitor whether other SWFs (Temasek, PIF, ADIA) echo similar AI concentration warnings. When sovereign wealth funds move in concert, it's a regime signal.
3. Oil price trajectory. The US-Iran conflict is the wild card. If oil sustains above US$100, the inflation-Fed-rate dynamic forces a repricing of all growth assets. If it retreats below US$85, the pressure on AI capex financing eases. The Strait of Hormuz remains the key chokepoint — any disruption there sends oil higher regardless of supply fundamentals.
The AI story isn't over. But the easy-money chapter might be.