July's tech sell-off tested the AI boom — Singapore's infrastructure bet is still running
Source: The Straits Times
Singapore's economy has spent the last two years riding the AI investment wave — the government this week raised its 2026 growth forecast to 4.5-5.5% partly on the strength of AI-driven electronics exports. But July gave global tech investors their sharpest scare in two decades.
Singapore's economy has spent the last two years riding the AI investment wave — the government this week raised its 2026 growth forecast to 4.5-5.5% partly on the strength of AI-driven electronics exports. But July gave global tech investors their sharpest scare in two decades, and for a city-state whose growth story is now tightly wound around AI capital spending, the question is whether that wobble was a blip or the start of something bigger.
In the last week of July, institutional investors — including the US hedge funds that had led the rally — unwound tech exposure at an unprecedented pace. The Nasdaq fell 3.2% for the month, its worst July in 20 years, and by July 29 the index was down 10% from its June peak, officially in correction territory. The rebound since has done little to convince analysts the volatility is over. Yet the infrastructure build-out underneath the market noise shows no sign of stopping: TrendForce projects the top nine cloud providers — Google, Amazon, Meta, Microsoft, Oracle, ByteDance, Tencent, Alibaba and Baidu — will spend roughly US$886.7 billion (S$1.14 trillion) on AI infrastructure in 2026, up about 90% from last year, heading toward US$1.3 trillion in 2027.
The disconnect is the point. AI-related debt issuance has already reached US$450-490 billion in visible bonds and loans, with cumulative off-balance-sheet commitments across the tech giants estimated at US$1.65 trillion. Morgan Stanley's chief US equity strategist argues the earnings and infrastructure story will survive the unwind, but warns the market "is no longer rewarding AI investment on faith alone" — execution and monetisation now separate leaders from spenders. Hyperscalers are also reframing the economics in ways that matter directly for Singapore: Microsoft is extending data centre lifespans from 15 to 25 years, and Amazon claims its computing campuses can generate returns for 30-plus years. As Tiger Brokers' Singapore-based strategist James Ooi notes, chips may become obsolete within a few years, but the land, buildings, power and cooling around them support multiple generations of hardware — exactly the long-lived infrastructure Singapore builds.
For Singapore the stakes are concrete. Micron, which has invested tens of billions of dollars here, remains one of three suppliers dominating the high-bandwidth memory chips AI data centres need, and the global memory shortage is expected to last until at least 2030 — the sell-off in memory stocks after Chinese rival CXMT's July IPO debut looks like an overreaction. But the deeper message is that Singapore's AI boom now comes with market scrutiny attached. T. Rowe Price's David Eiswert warns AI is unlike software: it becomes more capital-intensive as it scales, with every new model and inference request demanding more compute, memory and energy. Investors will now judge AI-linked economies on returns, not promises.
Why it matters for Singapore: The July correction was a recalibration, not a reversal — the spending cycle that drives Singapore's data centre builds, chip packaging ventures and export growth is still accelerating into 2027. What changed is that Singapore can no longer bank on AI hype alone: its edge now rests on being the physical infrastructure hub — the power, land and supply chains that outlive any single chip generation — and on proving that AI investment here translates into productivity and earnings. That is a stronger position than the market noise suggests, but it means watching whether cloud capex commitments hold through the next earnings season.
