Why published sizes of sovereign funds differ

The same fund can carry very different size figures, and there is usually a plain reason. Of the 167 entities with a comparable headline size in our profiles, 139 also record at least one other size figure. When that other figure is on a different basis, such as equity against total assets, the median gap is 55%. When it is on the same basis, the median gap is 14%. In almost every case, the gap comes from one of four things: what is counted, the basis of the figure, the date, or the currency.

How big the differences are

Each profile has one headline size and may record other size figures for the same entity: a different measure, a different date, or a part of the whole. For every other figure we can put in the same currency as the headline, we divided the larger of the two by the smaller. A ratio of 1.00 means the figures match. A ratio of 2.00 means one is double the other.

Source: our dataset. Basis: headline and other size figures recorded in each profile, leaving out committed capital and targets. The bands overlap: a pair above 2 is also counted above 1.25.

Across the 139 entities, 105 record at least one figure on a different basis from the headline, and 43 record a figure on the same basis at another date (our count). The most common pairings are these:

Source: our calculation from our dataset.

Total assets is usually the largest figure, because it counts everything on the balance sheet, including borrowed money and operating companies. A portfolio value is often much smaller, because it may cover only the investments a fund manages directly.

Reason 1: what is counted

The biggest gaps come from counting different things under the same name.

A holding group against its own capital. Our profile of the Turkiye Wealth Fund records audited consolidated total assets of TRY 18,997bn, about US$442bn, and consolidated equity of TRY 2,729bn, about US$63.5bn, at 31 December 2025. That is a ratio of 6.96. The larger figure includes the banks and companies in its group. That is a balance sheet, not a portfolio value.

One fund against its wider group. Hong Kong's Exchange Fund published audited total assets of HK$4,161.2bn for the fund alone and HK$4,719.7bn for the group at 31 December 2025. Same entity, same date, two correct totals.

One fund against everything a manager runs. The New Mexico State Investment Council manages several permanent funds and some money for other government clients. Its report gives US$74.63bn of managed net assets at 31 July 2026, covering all of them. A figure for a single fund will be much lower: our profile records US$41.01bn for the Land Grant Permanent Fund and US$13.51bn for the Severance Tax Permanent Fund on the same date. The highest and lowest of these figures differ by a ratio of 5.52.

Reason 2: the basis of the figure

Even when the entity is the same, a figure can be on a different basis.

Assets under management against total assets. Saudi Arabia's PIF reports both. Its 2025 report gives SAR 3,396bn of assets under management, about US$906bn at the riyal's 3.75 peg, and SAR 4,541bn of consolidated total assets, about US$1.21tn, at 31 December 2025. The two figures differ by a ratio of 1.34, and both are correct.

Consolidated balance sheets. Some state investors publish only group accounts. ADQ reports audited consolidated total assets of AED 996.777bn at 31 December 2025 and total equity of AED 521.4bn, and the Investment Corporation of Dubai reports consolidated total assets of AED 1.68tn. Both totals include operating companies. Bahrain's Mumtalakat reports audited consolidated total assets of BHD 7.223bn at 31 December 2024 and equity of BHD 3.19bn. Total assets and equity are both correct, and they answer different questions.

Statements and estimates. For Danantara in Indonesia, our profile records a reported figure of about US$900bn to US$1tn, unaudited. It also records reported consolidated assets of IDR 17,960tn for the state companies it oversees, and launch capital of about IDR 320tn (about US$19.6bn). These measure different things: assets under oversight, group balance sheets and capital put in. None of these is an audited fund figure.

Reason 3: the date

Funds report at the end of their own financial year, and some report quarterly. A figure from one year end and a figure from another can both be correct.

Japan's GPIF shows the effect clearly. It held JPY 293.6tn at 31 March 2026, which our profile converts to US$1,841bn. Three months later, at 30 June 2026, it held JPY 317.76tn, or US$1,956bn. A list that uses one date and a list that uses the other will differ by more than US$100bn for the same fund.

Reason 4: the currency

Many funds report in their own currency. Every dollar figure for them is a conversion, and the result depends on the rate and the day.

Norway. NBIM reports the Government Pension Fund Global in kroner: NOK 22,683bn at 30 June 2026. At the ECB reference rate of NOK 9.9267 to the dollar for that day, that is about US$2,285bn on our arithmetic. A conversion made on another date, at another rate, gives a different dollar figure for the same kroner amount.

Botswana. The Pula Fund held BWP 27.865bn at 31 December 2025. At the rate in the Bank of Botswana's own notes that is US$2.05bn. At a market rate of BWP 13.139 per dollar it is US$2.12bn. Same fund, same day, two dollar figures.

Pegged currencies. Hong Kong's Exchange Fund reported total assets of HK$4,463.6bn at 30 June 2026. At the 7.8 peg that is about US$572bn. Pegs make conversion simpler, but the dollar figure is still a conversion. Of the 167 comparable headline figures on our site, 97 are converted to US dollars from another currency (our count).

When a fund publishes no figure

Some of the funds most often named among the largest do not publish their size. Our profiles record that the Abu Dhabi Investment Authority, the Kuwait Investment Authority and the Qatar Investment Authority do not disclose their assets. Their headline figures, about US$1.1tn, about US$1tn and about US$580bn, are estimates (source: our dataset). We label them that way on every page.

Where our headline figures come from

Every size figure on our site carries its basis and source. Across the 231 entities on our list, grouped by where the headline figure in each profile comes from:

Source: our dataset.

So nearly three in five headline figures rest on the fund's own reporting or audited accounts (our arithmetic: 134 of 231, or 58%). Each profile shows the exact basis, with the source and date.

How to read any size figure

Before you compare two funds, ask four questions:

1. What is counted? One fund, a manager's full book, or a whole group?
2. What basis? Assets under management, total assets, net assets or equity?
3. What date? Which year end or quarter?
4. What currency? Is the dollar figure the fund's own, or a conversion, and at what rate?

These questions are also why we never add sizes together across funds. A total that mixes portfolios, group balance sheets, estimates, dates and currencies would not measure anything. Our largest sovereign wealth funds page shows the basis and date next to every figure.

Caveats

1. Figures change. The values behind this study are the figures recorded in each profile, and funds revise and update their figures over time.
2. Ratios, not totals. We compare figures for the same fund. We never add figures across funds.
3. Coverage. Entities with only one size figure, or none, are not in the ratio tables. Other figures in a currency we cannot match to the headline are left out of the pairs.
4. Causes. We name a cause only where the fund's own reports, as recorded in its profile, set it out.

Methods note

Population: 139 entities whose profile records a comparable headline size and at least one other size figure. Committed capital and targets are left out. Ratio: the larger of the headline and each other figure divided by the smaller, in US dollars or in the headline's own currency. Medians and counts are our arithmetic on the profiles. Currency examples use the conversions recorded in each profile. Basis counts group the headline size figure in each profile by its source. Full method: methodology.