​Singapore’s sovereign wealth funds are betting big on AI, but the central bank isn’t so sure

[SINGAPORE] Within three weeks in July, three of Singapore’s most consequential financial institutions delivered very different verdicts on the future.

​Temasek and GIC, the twin sovereign wealth funds that manage the bulk of the Republic’s reserves, used their annual results to explain why they intend to lean into artificial intelligence.

​Then the Monetary Authority of Singapore (MAS) weighed in and – in the flattest language a central bank can manage – wondered whether anyone had checked the odds on this bet.

​What’s at stake?

​Temasek celebrated a record S$518 billion portfolio and a 10.5 per cent annual return, while signalling plans to more than double its AI exposure by 2031 to keep that growth engine firing.

​GIC showed the same appetite. It raised its equities weighting to 56 per cent, up from 51 per cent, and named AI a core investment theme spanning enablers, monetisers, and adopters of the technology.

​Then came MAS managing director Chia Der Jiun. At the regulator’s briefing, he flagged the sustainability of the AI boom as the single biggest uncertainty facing an otherwise calm financial system. ​Global growth and equity valuations, he pointed out, now depend heavily on a handful of firms making an enormous bet pay off. If it fails, hyperscalers will pull back and markets will reprice – and probably not gently.

​Up to 50 per cent of the long-term expected returns from these three institutions flow into Singapore’s national budget through the Net Investment Returns Contribution (NIRC).

​That steady payout makes the NIRC the government’s single largest revenue source – footing the bill for schools, healthcare subsidies, and public infrastructure so taxpayers are spared the full tab.

​When these entities talk about where the global economy is heading, or explain how they plan to hit their targets without blowing up the portfolio, it pays to listen.

​The believers

​Temasek’s AI exposure stands at 6 per cent of portfolio value, with a target of up to 15 per cent by 2031. The fund frames this near-trebling as embedding AI into how it invests and operates, rather than a speculative punt.

​Tellingly, its headline deal of the year was digital infrastructure, rather than a chatbot: a S$6.6 billion sale of ST Telemedia’s data-centre stake to KKR and Singtel. Temasek cashed in on a decade-long bet on the plumbing beneath the boom, rather than the boom itself.

​GIC’s version was headline-grabbing, yet more caveated than it first appears. Its FY2025/26 report retires the 65/35 reference portfolio it used for 13 years, replacing it with three buckets: growth, income, and inflation protection.

​Instead of simply buying obvious tech giants, GIC has split its AI playbook across enablers building the plumbing, monetisers selling the software, and adopters using it to cut costs.

​Named holdings span the entire food chain: frontier lab Anthropic – which GIC has backed three times in the past year, most recently co-leading a US$65 billion round in May – alongside AI-native fintech Ramp and pharmaceutical giant Eli Lilly, which uses supercomputers to hunt for new drugs.

​At the same time, GIC has committed another US$30 billion to hedge funds. That move comes on top of a decade spent trebling its hedge fund book worldwide, and is pitched as a way to diversify away from the concentration its AI bet represents.

​Still mindful of the risks

​MAS’ caution is no vague nod to prudence. ​Chia has repeated versions of this warning since May – taking it from the UBS Asian Investment Conference to Shanghai’s Lujiazui Forum, and finally to MAS’ annual briefing.

​AI-related capital spending now drives the bulk of US investment growth and roughly half of US GDP growth, exposing the global economy to a narrow bet. Energy and chip costs keep rising, raw-material supply remains tight, and cheaper open-weight models are squeezing the pricing power of the leading labs.

​If revenue fails to catch up with the capital committed, MAS argues, hyperscalers will slash spending and markets will reprice.

But both GIC and Temasek aren’t just plunging headlong into AI.

​Even as GIC goes out to bat for AI, it is well aware of the risks. Ahead of the fund’s own report, group chief investment officer Bryan Yeo described the flood of capital into AI as one giant, correlated risk position, ranking it above any single country or industry as the biggest risk in markets today.

GIC noted that AI is widening the gap between winners and losers. Frontier models are clustering around a handful of players, while the rest of the value chain turns into a commodity.

What this means is ​somebody is going to lose money on this boom.

But GIC is clear-eyed about the risks. And with its new flexible investment framework, it will be aiming to position itself as one of the leaders instead of one of the many hopefuls trailing behind them.

After all, this comes as GIC’s 20-year annualised real return slipped 0.4 percentage point to 3.4 per cent – its weakest showing since 2020.

With this in mind, it’s not surprising that chief executive Lim Chow Kiat pointed to choppy markets and geopolitical tension, while acknowledging a deliberate call to dial back risk-taking this cycle.

​Temasek felt the same drag. Its five-year annual return of 4.6 per cent, against 6.8 per cent over twenty years, reflects years of China-related headwinds. Its portfolio value was also trimmed roughly 2 per cent in the final month of the year alone from the Middle East fallout.

On its part, Temasek also said that it will remain selective and disciplined about risks, especially pertaining to the AI segment.

It is also seeking to hedge its bet on AI’s potential for growth by also building shock absorbers alongside it – targeting a five-fold jump in core-plus infrastructure and more than doubling its private credit allocation to lock in steady cash yields.

​Chief investment officer Rohit Sipahimalani made no bones about why Temasek loves private credit and core-plus infrastructure: they pump out steady cash yields and offer a “narrower range of outcomes” in a shaky market.

​That is a telltale sign. The fund is happy to back the AI boom, yet it is spending real money on boring assets that will not blow up if the market turns.

​The risks for Singapore

​None of this means Temasek and GIC are wrong to lean into AI. They cannot be expected to sit out the largest capital reallocation in a generation because a central banker looked uneasy.

​Singapore’s sovereign funds are hardly ignoring the risk MAS flagged. Both see the danger, both have started hedging their portfolios, yet neither is backing away from the table.

Temasek plans to treble its AI exposure by 2031, which sounds ambitious but will be capped at just 15 per cent of the total portfolio. The risks are further mitigated by boring, cash-yielding infrastructure. Meanwhile, GIC is writing cheques for frontier AI models while parking another US$30 billion in hedge funds. ​Both are betting big while paying up for downside protection.

​That divergence comes down to a simple difference in job descriptions. While all three institutions feed the national budget through the NIRC, MAS is a central bank first and foremost – tasked with defending the currency and policing systemic risk, not chasing growth stocks.

​GIC and Temasek, on the other hand, cannot ignore a generational tech shift given a mandate essentially to multiply the country’s reserves over decades. Risk is unavoidable in their line of work, and sitting on cash is its own kind of gamble.

​In the end, having a central bank that worries out loud and two reserve funds that hedge while they buy might be the most sensible set-up available: one foot on the accelerator, one foot hovering over the brake – and hoping the road ahead is straight.