Temasek Puts $400B Behind AI, Keeps Crypto "Off the Table"
Singapore's state investor says it holds zero direct crypto exposure four years after the FTX collapse, while aiming to more than double its AI allocation by 2031.
By TRAGenX Desk
A record portfolio, and crypto still out
Temasek Holdings closed its fiscal year with a net portfolio value of roughly S$518 billion (about $400 billion) — a record — and used the moment to reaffirm a hard line under crypto. "We don't have directly any, any investment in crypto," Nagi Hamiyeh, president of Temasek Global Investments, told CNBC on July 9. Four years after FTX's collapse cost Temasek roughly $275 million, that position hasn't softened.
The FTX write-off is still doing work
The FTX loss wasn't just a line item. It exposed how little consumer protection existed around Temasek's crypto exposure at the time and pushed Singapore's Monetary Authority to tighten licensing and compliance for digital-asset businesses. For a fund managing $400 billion, avoiding crypto now reads less like ideology and more like a straightforward risk-adjusted call: regulatory uncertainty plus compliance overhead, with no clear offsetting return, is an easy pass.
AI goes from 6% to a 15% target
The contrast with AI is stark. Temasek's AI exposure sits at 6% of the portfolio as of the fiscal year ended March 2026, and management is targeting 15% by 2031 — a jump of more than double in five years. Hamiyeh's framing: "it's all about the applications, and it's all about the companies that embrace AI and build a moat." The fund's longest-term wager isn't on foundation models — it's on physical AI: automation, robotics, and industrial process optimization. Its China-focused AI bets follow the same pattern, tilted toward physical AI and applications rather than frontier model training.
Blockchain infrastructure isn't crypto, in Temasek's book
One nuance worth sitting with: Hamiyeh didn't rule out blockchain outright. Temasek says it continues to evaluate blockchain technology for what it can do for the real economy — infrastructure and tooling, not token exposure. That's a distinction institutional allocators are drawing more sharply post-FTX: blockchain as developer infrastructure is fundable; crypto as a speculative asset class is largely still radioactive for the biggest sovereign-linked funds.
- Capital is rotating hard toward applied AI — physical AI and enterprise applications, not raw crypto exposure.
- Where institutional interest in blockchain persists, it increasingly routes through infrastructure and dev tooling rather than holding tokens.
- Regulatory clarity (or its absence) is now a direct input into capital allocation, not a footnote to it.
Why this matters beyond one fund's balance sheet
Temasek is one allocator, not a proxy for the whole market — plenty of institutions still hold direct crypto and token exposure. But a $400 billion fund publicly separating "blockchain infrastructure we'll fund" from "crypto we won't touch" is a useful signal for anyone building in Web3 or AI-adjacent fintech: the capital that matters most for scaling a company is increasingly chasing applications and infrastructure with real users, not speculative token upside.
FAQ
Frequently asked questions
- Does Temasek's stance mean institutional crypto investment is dying?
- No — it reflects one large allocator's post-FTX risk calculus. Other institutions still hold crypto and blockchain exposure, but wariness of direct token exposure is common among the largest sovereign-linked funds.
- What's the difference between Temasek avoiding "crypto" and still watching "blockchain"?
- Temasek frames crypto as speculative asset exposure it won't hold directly, while blockchain is infrastructure it continues to evaluate for real-economy applications — a distinction between the asset class and the underlying technology.
- What kind of AI is Temasek actually betting on?
- Its stated longest-term wager is physical AI — automation, robotics, and industrial process optimization — alongside applications and companies it believes can build a durable moat, rather than a bet concentrated in foundation-model training.
Sources