Tesla's Expected Return at a Third of the Volatility — The Math of Diversification

In the first half of 2026, the overseas stock Korean retail investors held most was — once again — Tesla (No. 1 by custody value at the Korea Securities Depository). For anyone who wants Tesla's returns but dreads its −60% roller coaster, Markowitz supplied the answer seventy years ago: you can shave off the swings without lowering your return target. To see if that holds, we built a portfolio with the same expected return as Tesla in DIVA Quantizer.

Feed the last six years of Tesla (July 2020 – June 2026, 72 months) into DIVA's Markowitz analysis and you get: expected return +49.83%/yr, volatility 67.14%/yr, Sharpe 0.69, 95% VaR −60.61%. DIVA's personality label: “The Romantic Beast.”

The problem is the price paid for that return. 67% volatility is more than four times the S&P 500 over the same window (15.8%/yr), and a normal-distribution 95% VaR of −60.6% means roughly one year in twenty could cut the account to less than half. The moment you can't stomach that swing and sell at the bottom, the return in the statistics table is no longer yours.

A portfolio's risk is not the weighted average of its holdings' risks. As long as correlations sit below 1, portfolio volatility lands strictly below that average — assets moving to different rhythms cancel out each other's swings. So what matters is not the number of holdings but what you mix: the less an asset moves like Tesla, the more of Tesla's swing it cancels.

Our candidate pool is Tesla plus nine household names. Matching Tesla's ~50%/yr bar takes some high-return candidates, so the pool spans AI semiconductors through aerospace, healthcare, financials, energy, staples and gold. Over the same 72-month sample, here is each asset's return and volatility alongside how closely it moved with Tesla (correlation):

In months when Tesla plunged, Eli Lilly (correlation −0.05), ExxonMobil (−0.07) and gold (−0.01) moved independently or the other way on average. Even Micron — a fellow “AI winner” — correlates at just 0.12. This is the pantry for what Markowitz called the free lunch.

Simple diversification first

Hold all ten names at an equal 10% and volatility drops to 19.9% — but expected return falls with it, to 40.3%. Risk went down and return went with it. That's not what we asked for. The goal: keep Tesla's return, cut only the risk.

Markowitz optimization with the return pinned

Two constraints: ① the window-average return must equal Tesla's (49.83%/yr), and ② no single name may dominate — the tightest per-asset cap that still reaches the target (18%). Minimizing variance under those constraints, the optimizer converges on seven names:

The result: the same +49.83% expected return at 24.43% volatility instead of 67.14% — roughly a third. Excess return per unit of risk (Sharpe) rises 2.7×, from 0.69 to 1.89. On the efficient-frontier chart, the Tesla dot (σ 67.1) has slid horizontally left to land exactly on the curve (σ 24.4).

How much of this is the “magic of diversification”?

Let's split it honestly. Of the drop from 67.1% to 24.4%, correlation's share is the final leg: held separately, these seven names average 41.2% volatility; held together, 24.4% (a 40.7% diversification benefit). The first leg (67.1% → 41.2%) isn't diversification at all — it's swapping into assets that happened to earn more per unit of risk over the past six years. The magic of correlation below 1 is free; choosing what to hold is still estimation.

“But I don’t want to sell my Tesla”

The optimum above cuts Tesla to 1.9% — effectively a “sell your Tesla” answer. So we added a constraint holding Tesla at 30% (lower bar in the weights figure). The result: the same expected return at 30.0% volatility (45% of standalone Tesla), Sharpe 1.54. You can keep a large core position and still shed more than half the risk. All four scenarios side by side:

One lesson survives any change of sample, though: as long as correlations are below 1, mixing pulls risk below the weighted average — that math works on any six years you cut. A 100% single-stock account tends to sit inside the frontier at every return level, buying the same return with more risk. The real question is “what to mix,” and the answer starts not with forecasting returns but with checking how your holdings move together.

Where your portfolio sits on the frontier, and how your holdings correlate, is exactly what DIVA Quantizer's Markowitz analysis shows. Enter your tickers and weights, and a few clicks return your expected return, volatility, diversification benefit and your spot on the efficient frontier. Envying Tesla's return is one thing; engineering that return at a third of the swings is another — and the latter is the tool Markowitz left us.

Sources: prices are yfinance month-end adjusted close (Jul 2020 – Jun 2026, 72 months); risk-free rate is the US 13-week T-bill (^IRX, 3.73%/yr). Optimization: mean-variance minimization (SLSQP) with no short selling, weights summing to 1, and an 18% per-asset cap. Retail holding rank: Korea Securities Depository, H1 2026 foreign-securities custody value. All metrics are DIVA Quantizer Markowitz model output.

This article was written with the help of AI.