Scottish Mortgage after SpaceX and Tesla: Baillie Gifford's growth portfolio vs MSCI World

Executive summary

From 31 March 2007 to 31 October 2026, Scottish Mortgage's shares compounded at 15.64% a year in sterling, against 10.25% for MSCI World net, yet earned slightly less per unit of risk: a Sharpe ratio of 0.65 against 0.68 and a Sortino ratio of 1.02 against 1.04. The trust's interim report, published on 6 November 2026, shows a net asset value (NAV) total return of 31.5% for the six months to 30 September, against 16.0% for its FTSE All-World benchmark.

SpaceX, which listed on 12 June, was 15.0% of total assets at 30 September and 11.1% on 4 November; the managers have been reducing it as lock-ups allow. The last Tesla shares were sold during the half, and gearing fell from 11% to 4%.

Copio's assessment: for shareholders who held throughout, Scottish Mortgage has delivered strong compounded growth while exposing shareholders to deep losses. Its higher growth accompanied about 1.8 times the index’s volatility and a slightly lower Sharpe ratio; this comparison does not prove that taking more risk caused the extra return. Twice its shareholders lost about 60%: 62.6% in 2008 and 60.1% from November 2021 to May 2023, while MSCI World in sterling fell 3.6% over the second period's dates.

The qualification is the entry date. Over the five years to October 2026, which begin days before the November 2021 peak, the shares returned 0.37% a year against 11.00% for MSCI World, and they regained their total-return peak only on 26 May 2026. The half-year’s 31.5% rests heavily on a few positions: SpaceX, Moderna and Anthropic contributed 15.9 percentage points between them.

For a household, the next step is to check the combined NVIDIA, Amazon and ASML exposure across the trust and any world or Nasdaq funds. Then translate a repeat 60% fall in the trust into a loss on the whole portfolio, including any spending that would require a sale.

What is in Scottish Mortgage?

At 30 September 2026 Scottish Mortgage had £18.7bn of total assets: 73.9% in listed shares, 24.8% in unlisted company shares, 0.1% in unlisted bonds and 1.2% in net liquid assets (list of investments, PDF page 20). Shareholders' funds, after £984.3m of borrowings at book value, were £17.7bn. Its ten largest holdings made up 52.3% of total assets.

At 30 September 2026, of total assets: SpaceX 15.0%; TSMC 6.5%; NVIDIA 6.2%; Moderna 4.3%; ByteDance 4.0%; Anthropic 4.0%; Zipline 3.7%; MercadoLibre 3.1%; ASML 2.8%; Amazon.com 2.7%. Source: Scottish Mortgage Interim Financial Report, list of investments, PDF pages 17-20; holdings, returns and locators (CSV). ByteDance, Anthropic and Zipline are unlisted and carried at the directors' valuations.

SpaceX was valued at £2.98bn at 31 March 2026, before it listed, which was 19.3% of total assets by our calculation. The managers write that it later "had grown to more than a quarter of the portfolio, unusually large even by our standards" (interim management report, PDF page 6). By 30 September the holding was £2.80bn, 15.0%, and the chair's statement gives 11.1% on 4 November.

The interim report lists Tesla among eight complete sales in the half, with Netflix, Delivery Hero, Kering, Insulet, HeartFlow, Sana Biotechnology and Uptake Technologies. The trust says Tesla returned 13 times the capital invested since 2013. CNBC reported on 6 November 2026 that the sale was in April and that the trust's July update put the amount invested at about $470m and realised profit at about $6.2bn; we did not retrieve that update.

New holdings included SK hynix, KLA, DISCO, Vistra, EQT and Bending Spoons, plus three private companies: Hadrian, ICON and Chai Discovery. The NVIDIA holding rose from £489.6m to £1,154.1m and the Anthropic holding from £171.5m to £744.4m, through purchases and price changes together (PDF page 17).

The 53 private companies made up 24.9% of total assets, and four of them account for over half of that exposure (private companies summary, PDF page 12). At 31 March, unlisted holdings were 41.4% of total assets; SpaceX's listing moved most of the difference into listed shares. The directors value private holdings under the IPEV guidelines on a three-monthly cycle, with extra reviews after "trigger" events (note 6, PDF page 27).

By the trust's classification, 62.2% of total assets were in the United States, 9.8% in China and 6.5% in Taiwan. Technology was 43.7%, telecommunications 15.0%, consumer discretionary 12.0%, healthcare 11.2% and industrials 9.9% (distribution of total assets, PDF page 11).

Financing changed most in the half. The trust repaid all its bank loans and a £50m debenture, leaving £984.3m of borrowings at book value, mainly unsecured loan notes redeemable from 2036 to 2062 at a weighted average cost of 2.7% (notes 7-8, PDF page 27). Net of £241.6m of cash and settlements, gearing was 4%, against 11% at 31 March.

The long-dated notes were worth £608.7m at fair value. That gap is why NAV per share with debt at fair value, 1,726.6p, is higher than NAV at book value, 1,690.7p. The shares closed the half at 1,590.5p, a 7.9% discount to the fair-value NAV; during the six months they ranged from an 8.2% premium to an 11.0% discount.

The board reissued 40.95 million treasury shares at a premium between April and early June, raising about £602m, then bought back 72.9 million shares for about £1.06bn once they returned to a discount. Ongoing charges were 0.33% in the year to 31 March 2026, and the management fee is 0.30% on the first £4bn of assets and 0.25% thereafter.

These are actual holdings, distinct from the policy limits. The investment policy caps any holding at 8% of total assets at the time of purchase, and new private investment at 30% of total assets, also measured at purchase, plus £250m of additional capacity approved on 10 April 2026. Lender covenants restrict gearing to 35%, and the trust takes on no new borrowing beyond 30%.

Why does Baillie Gifford run it this way?

The managers, Tom Slater and Lawrence Burns of Baillie Gifford, describe a portfolio built to own the few companies that produce extreme returns, held for years and funded partly with long-dated borrowing. The objective, in the 2026 annual report: "The objective is to maximise total return from a portfolio of long term investments chosen on a global basis enabling the Company to provide capital and dividend growth."

The benchmark is the FTSE All-World Index in sterling over five-year rolling periods, and the policy states: "This benchmark is a reference point for considering performance and emphatically is not a portfolio construction tool." The managers' core beliefs explain why: "Our own research demonstrates that in the past thirty years approximately five per cent of stocks have returned at least five-fold in any five-year period" (PDF page 19).

They accept losses on individual positions as the price of those outliers. "We need to be willing to accept loss if there is an equal or greater chance of (almost) unlimited gain" (PDF page 18). On volatility they draw a line between a trust and a whole portfolio: "Limiting volatility can make sense for an overall investment portfolio but we doubt the benefits of using equities to meet this objective" (PDF page 19).

Private companies follow from the same aim. "Access to fluctuating daily quotations for our holdings is more often a distraction than an advantage for a fund with permanent capital", the beliefs state (PDF page 20). The 2023 report adds that measuring the private limit at the time of purchase "protects Scottish Mortgage from being forced sellers of private companies" when values move.

The board treats borrowing as a structural advantage of an investment trust. "The Board is committed to the strategic use of borrowings in the belief that gearing the portfolio into long run equity market returns will enhance returns to shareholders over the long term", and "no attempt is made to time short term market movements through tactical shifts in the level of gearing" (gearing policy, PDF page 47).

In his 2026 review, Slater turns the index argument around: "A passive allocation to a global equity index today is a 63% bet on the United States, a 33% bet on the technology sector, and a position in which over 35% of your capital sits in ten companies." His next sentence: "That is not a diversified default."

He calls volatility "the opportunity, not the risk" for a trust like this one, and concedes: "We recognise that this makes the experience of owning Scottish Mortgage less comfortable than it would be if we managed the portfolio closer to a benchmark" (PDF page 14).

The half-year sales apply the same reasoning to position size and price. On SpaceX: "We reduced the holding as lock-ups allowed, not because our enthusiasm dimmed but because it had grown to more than a quarter of the portfolio, unusually large even by our standards." On Tesla: "Since we first invested in 2013, it has returned 13 times the capital invested, but we felt the share price had come to reflect much of its promise in autonomy and robotics."

The managers also record a sizing error. Of Moderna, which contributed 4.5 points in the half, they write that "the holding we took was too large for what was still an early-stage biotechnology company. That was a mistake" (PDF page 6). The board's own description of the intended shareholder is blunt: "Scottish Mortgage is intended for shareholders who are able to take a long-term view and withstand periods of pronounced market volatility" (PDF page 5).

Historical performance versus MSCI World

Scottish Mortgage publishes its NAV daily, but its reports give official NAV total returns only by financial year, so we compare it with MSCI World on two matched bases. The NAV comparison uses the trust's one-year figures for the ten years to 31 March 2026. The share price comparison uses monthly sterling total returns from April 2007 to October 2026; both use Bank Rate as the cash benchmark.

A hypothetical £100 in the NAV at 31 March 2016 grew to £704.87 by 30 September 2026, against £382.59 in MSCI World net and £641.93 in the shares. Over the ten financial years to March 2026, NAV compounded at 18.28% a year, the shares at 16.98% and MSCI World at 12.77%.

£100 at 31 March 2016 became £704.87 in Scottish Mortgage's NAV and £382.59 in MSCI World net by 30 September 2026; year to March 2021: +111.2% and +38.4%; 2022: -13.1% and +15.4%; 2023: -17.8% and -1.0%; 2024: +11.5% and +22.5%; six months to September 2026: +31.5% and +15.0%. Source: Scottish Mortgage annual reports 2017-2026 and interim report for NAV; MSCI World net GBP month-end levels (index 990100); yearly comparison (CSV), official returns with locators (CSV).

The NAV beat MSCI World in seven of the ten years. The lead came in long runs: from March 2016 to March 2021 the NAV multiplied 4.75 times while the index multiplied 1.95 times. It then trailed for three years running, in 2022, 2023 and 2024, by between 11.0 and 28.5 percentage points a year.

The shares compounded 1.3 points a year less than the NAV over the decade, mainly because a 1.3% premium in March 2016 became a 9.5% discount by March 2026. Shareholders earn the share price, not the NAV.

With ten observations and only two or three years below cash, these annual Sortino ratios are unstable: a single bad year moves them a long way. The Sharpe ratios are the steadier reading, and they favour the index. The monthly share price series has 235 observations and gives the sturdier comparison.

£100 in the shares in March 2007 became £1,721.91 by October 2026, against £676.27 in MSCI World and £139.75 at Bank Rate. The ratios split by decade: from April 2007 to March 2017 the shares had the higher Sharpe ratio, 0.65 against 0.59, and from November 2016 the index did, 0.84 against 0.67.

Scottish Mortgage shares: 24.3% annualised excess-return risk for a 17.6% mean annual return; MSCI World net: 13.2% for 10.7%; Sharpe 0.65 against 0.68. Source: Scottish Mortgage closes from Yahoo Finance, with dividends reinvested on the ex-dates, the MSCI levels above and Bank of England Bank Rate, monthly average (IUMABEDR); frontier data (CSV), monthly inputs (CSV). The curve is restricted to long-only mixes of MSCI World and cash rebalanced each January, without leverage or costs; it is not the trust's own opportunity set.

The trust sits far to the right of the curve because no long-only mix of the index reaches its risk. Its slightly lower Sharpe ratio means a straight line from cash through Scottish Mortgage is a little flatter than the same line through MSCI World. The extra return came with proportionately more risk.

The crash record shows how that risk arrived. We measured three sell-offs from MSCI World's daily peak to its trough, and the trust's own peak-to-trough fall from November 2021. A weaker pound cushioned MSCI World in sterling during 2008 and 2022, which flatters the index in these windows.

Source: daily inputs (CSV) and episode results (CSV). In the 18 months when MSCI World fell more than 5%, it averaged -7.2% and Scottish Mortgage -8.7%; the shares fell in 17 of them. In the 11 months when the shares fell more than 10%, they averaged -14.4% while the index averaged -5.4%.

The trust's own figures show the 2021-2023 fall. NAV per share with debt at fair value peaked at 1,482.8p in the year to March 2022 and bottomed at 771.6p in the year to March 2024, a 48.0% fall (2022 report, PDF page 7; 2024 report, PDF page 22). Across those three financial years the shares traded between a 7.2% premium and a 22.7% discount.

Gearing rose from 6% at March 2021 to 13% at March 2022 and 14% at March 2023, so borrowing amplified the fall (ten-year record, 2026 report, PDF page 25). The 2023 report records the board's reading: the discount widened "With a backdrop of heightened market anxiety, particularly around growth and private company investing".

Reported NAV highs stayed below 1,482.8p in every year to March 2026, the highest being 1,346.7p; the six months to September 2026 brought a high of 1,741.3p. The shares regained their November 2021 total-return level on 26 May 2026, four years and seven months after the peak. After 2008 the shares regained their May 2008 peak in October 2010, while MSCI World in sterling regained its October 2007 peak in March 2010.

NAV per share fell from 130.3p at March 2008 to 76.8p at March 2009, 41.1%, with gearing at 23% and then 26% (2017 report, ten-year record, PDF page 26). Since the half-year end the pattern has repeated on a small scale: from 30 September to 6 November 2026 the shares fell 7.6% and MSCI World 3.8%, and on 5 November the shares lost 4.8% against 3.3% for the index.

Copio's reading

What does the record say about compounded growth?

Scottish Mortgage has delivered higher geometric growth than MSCI World for long-term holders, but the evidence does not show it is the better way to buy that growth. The objective here is expected log wealth for a shareholder over decades, with no withdrawals and the trust's constraints: permanent capital, the 8% and 30% purchase limits, and borrowing within covenants. The observed 15.64% CAGR is a result, not an estimate of what the portfolio will earn from today's prices.

Three tests from the evidence:

• Is the concentration paid for? Over the full period, yes: the shares compounded 5.39 points a year faster than the index. Volatility cost about 1.9 points of compounding a year, the gap between their 17.58% mean and 15.64% CAGR, against 0.4 points for MSCI World. The Sharpe ratio was still slightly lower, and over the last five years the concentration earned 0.37% a year against 11.00%.
• Does any protection or diversification improve growth after its cost? The cited reports do not identify a dedicated hedge against a broad equity crash, and its diversification between holdings did not prevent 60% losses. Long-dated fixed-rate debt reduces refinancing pressure, and the managers credit the purchase-time private limit with avoiding forced sales of private companies. These features do not eliminate all forced-sale risks. Shareholders absorbed the loss and the discount instead.
• Do fees, financing and cash flows change the result? Ongoing charges of 0.33% are low for active management. Financing cost £25.1m in the half, more than the £24.4m management fee. Gearing of 13-14% during the 2022-2023 losses deepened the fall, and the move from premium to discount cost shareholders 1.3 points a year against NAV over the decade.

Half-year results depend on a handful of names: SpaceX, Moderna and Anthropic added 15.9 of the 31.5 points. The managers' own Moderna admission shows that position sizes are judgements that can be wrong. The decisive comparison still missing is the expected return of the current holdings at current valuations against the index, after the volatility drag; neither the reports nor the price history can supply it.

Is it crash aware and protected?

It is built to survive crashes as a company, not to protect its shareholders from them. The board reviews leverage and liquidity stress tests, repaid all bank debt in the half, and holds borrowings that mature from 2036 to 2062, so no refinancing falls due for a decade. The 2023 report credits the purchase-time private limit with keeping the trust from being a forced seller of private companies, and it held them through 2022.

For the shareholder, losses were larger and longer than the index's: 62.6% in 2008, 60.1% from 2021 to 2023, and four years and seven months to recover the second time. The managers say plainly that they do not try to limit volatility inside an equity portfolio; that job falls to the rest of a shareholder's assets.

Today's 4% gearing and £241.6m of cash reduce the leverage. The chair attributes the fall in gearing to portfolio gains and to repaying borrowings partly with SpaceX proceeds, and the gearing policy says borrowing is not timed to markets.

A 60% loss needs a 150% gain to break even; our crash-recovery study sets out that arithmetic and how long recoveries take. A household drawing on the holding during such a fall would sell at the widened discount as well as the lower NAV.

Is it diversified without sacrificing growth?

Its 53 private companies and several dozen listed ones retain a common exposure to growth equities, much of it in technology and the United States. That common exposure can outweigh differences between individual businesses. By the trust's own classification 62.2% of total assets were in the US and 43.7% in technology, close to or above the shares of the index that Slater calls "not a diversified default". Over the full period, its monthly returns had a correlation of 0.76 with MSCI World and a beta of 1.40.

November 2016 to October 2026: Scottish Mortgage shares' correlation with the Nasdaq-100 ETF 0.75, with MSCI World 0.66, with its listed top holdings 0.36-0.63; MSCI World with the Nasdaq-100 0.86. Source: Yahoo Finance closes converted to sterling at month-end GBP/USD, and MSCI World net GBP; correlation matrix (CSV). These are observed monthly correlations over ordinary and stressed months together; they do not estimate crash correlations.

The trust moves more with the Nasdaq-100 than with the world index, and its listed top holdings are only moderately correlated with each other, from 0.11 for TSMC and MercadoLibre to 0.58 for TSMC and ASML. That spread did not hold in the 2022 rate shock: from January to June 2022 the shares fell 47.9% while MSCI World fell 14.6%. The common driver was the valuation of distant growth, which the private holdings share too.

The private quarter adds access to companies such as ByteDance and Anthropic. It does not, by itself, establish an independent source of returns or protection against a growth-equity sell-off. In 2021-2023 the NAV, which includes the directors' private valuations, fell 48.0% while the shares fell 60.1%. Directors' valuations update less often than market prices, so the smaller NAV fall is no evidence that the private holdings carried less economic risk. Our guide to counting independent bets explains why correlation, not the number of holdings, measures diversification.

Copio does not grade Scottish Mortgage in its portfolio checker, because a quarter of its assets are private companies and the rest are concentrated single stocks held with borrowings, which no UCITS ETF replica can reproduce.

A household holding it next to a world or Nasdaq-100 ETF owns NVIDIA, Amazon and ASML twice; measuring the combined weight in each company shows how concentrated the whole portfolio is. A hypothetical 10% position in the trust losing 60% subtracts 6% from the starting portfolio if everything else stays flat. That is a loss scenario, not a forecast; other equity holdings can fall at the same time. Compare the resulting balance with money needed for planned withdrawals.

In copio’s private beta, use fund look-through to inspect overlapping listed companies and crash replay to examine the rest of your portfolio’s historical losses. Use the trust’s own reports for its private holdings and borrowing. A replay cannot reproduce future private valuations or discount movements.

For other concentrated growth portfolios in this series, see Cathie Wood's ARK Innovation ETF, which held Tesla at 9.4% at 30 September 2026, and Leopold Aschenbrenner's Situational Awareness fund. Terry Smith's Fundsmith is another sterling-based concentrated fund compared with MSCI World, and our explainer on why a higher Sharpe ratio need not mean faster compounding covers the gap between the two measures used here.