Malaysia's Employees Provident Fund (EPF) concentrated 48.7% of its U.S. equity portfolio in just 10 holdings in Q4 2025. NVIDIA, Microsoft, and Meta lead the pack. Nearly half of a sovereign pension fund's U.S. equity allocation in 10 names.
For a sovereign pension responsible for 15+ million Malaysian workers' retirement, 48.7% concentration tells you exactly how the fund views the current market.
The top-10 breakdown
| Rank | Approximate holdings |
|---|---|
| 1-3 | NVDA, MSFT, META — AI/tech core |
| 4-5 | GOOGL + others — mega-cap tech |
| 6-10 | Mix including MU and other names |
| Top-10 total | 48.7% of portfolio |
48.7% concentration: where it sits
| Manager | Top-10 weight | Category |
|---|---|---|
| Berkshire | ~70% | Extreme conviction |
| ARK Invest | ~50% | Thematic concentration |
| EPF (Malaysia) | 48.7% | High conviction for a pension |
| Jennison | 47.6% | Growth conviction |
| CalPERS | ~47% (with VOO) | Index + overlay |
| S&P 500 | ~35% | Market baseline |
| MFS | ~30% | Diversified active |
| DFA | <20% | Anti-concentration |
EPF at 48.7% is MORE concentrated than CalPERS and significantly above the S&P 500 baseline. For a sovereign pension, this is aggressive positioning.
Why 48.7% from a sovereign pension is different
Pension fund constraints
Sovereign pensions typically operate under:
- Conservative risk mandates (protect retirement savings)
- Diversification requirements (don't put eggs in one basket)
- Regulatory oversight (government accountability)
- Multi-decade time horizons (stability over excitement)
What 48.7% means in this context
EPF's top-10 concentration exceeding the S&P 500's natural concentration means:
- Deliberate overweighting: Someone approved this concentration level
- AI thesis at the governance level: The EPF board signed off on heavy AI exposure
- Deviation from benchmark: This isn't closet indexing — it's an active bet
The NVDA-MSFT-META trio
The top 3 (NVDA, MSFT, META) represent EPF's AI thesis across three layers:
| Stock | Role in AI ecosystem | EPF thesis |
|---|---|---|
| NVIDIA | AI compute hardware | GPU monopoly for AI training |
| Microsoft | AI platform + enterprise | Azure AI + OpenAI partnership |
| Meta | AI consumer + infrastructure | AI-powered ads + Llama models |
This isn't random mega-cap exposure. It's a deliberate bet on three companies that each control a critical layer of the AI value chain.
What 48.7% top-10 means for risk
Upside scenario
If the AI trade continues and NVDA/MSFT/META outperform, EPF's U.S. equity returns will significantly outperform the benchmark. Malaysian retirees benefit from being early to the AI thesis.
Downside scenario
If the AI trade reverses (valuation compression, capex deceleration, regulatory backlash), 48.7% concentration amplifies the losses. A 30% decline in the top 3 would cost the portfolio ~15% from those positions alone.
The risk management question
EPF's total AUM is ~$250B+, with the vast majority in Malaysian government bonds, real estate, and domestic equities. The $13.6B U.S. equity portfolio is ~5% of total assets. So the 48.7% concentration is within a small sleeve — the actual fund-level risk is managed.
Cross-reference with other sovereign pensions
Are other sovereign pensions similarly concentrated?
| Fund | Top-10 weight | AI conviction |
|---|---|---|
| EPF (Malaysia) | 48.7% | Very high |
| CPPIB (Canada) | ~moderate | Moderate |
| CalPERS (U.S.) | ~47% (VOO-heavy) | ETF-driven |
| Swiss National Bank | ~index | Low (passive) |
| Norway GPFG | ~index | Low (broad) |
EPF stands out as the most AI-convicted sovereign pension in the 13F database.
Originally published at 13F Insight
Top comments (0)