Published on: 2026-07-29
AI chip shares are falling while companies with steadier earnings are attracting money. A 90/10 barbell would not try to predict which group wins next. It would protect most of the capital first, and confine concentrated growth exposure to a smaller sleeve that can fail without pulling down the whole portfolio.
A 90/10 barbell places 90% of capital in a defensive core and limits speculative exposure to the remaining 10%.
The loss boundary works only when leverage cannot push losses beyond the speculative sleeve.
A volatile investment belongs in the risky side only when its potential upside is large enough to justify failure.
The 28 July AI sell-off tests position sizing and concentration, not the basic barbell structure.
A barbell places capital at two clear ends rather than filling the portfolio with positions that offer neither strong protection nor exceptional upside.
The 90% side focuses on liquidity and capital preservation. The 10% side accepts a high probability of loss in pursuit of gains large enough to affect the entire portfolio.
The ratio is an illustration, not a fixed formula. The central rule is more important than the exact percentages. The 10% sleeve is allowed to fail. The 90% sleeve is not.
The defensive core should contain assets whose main job is to preserve capital and remain available during market stress. Cash, Treasury bills and short-duration government securities generally fit that role more closely than ordinary shares.
A company such as Coca-Cola may have steadier earnings than a semiconductor manufacturer, yet it remains an equity and can fall during a broad sell-off. Lower volatility is not the same as capital preservation.
Start by identifying the money that cannot depend on a market recovery. That amount belongs outside the speculative sleeve.
Assume a $100,000 portfolio places $90,000 in the defensive core and $10,000 in speculative positions. If the core earns 4% and the risky side falls to zero, the portfolio finishes near $93,600 before fees and taxes.
If the defensive core earned 4% over the year and the risky side fell to zero, the portfolio would finish near $93,600 before fees and taxes. The structure breaks when the $10,000 sleeve can create a larger loss.
Borrowing, margin, uncovered options and some short positions can extend losses beyond the capital initially assigned. Leveraged products can also create far more market exposure than their initial cash requirement suggests.
Conviction does not set the correct position size. The maximum acceptable loss does.
A speculative position needs more than an exciting industry or a fast-moving price. It should have a downside that can be contained within the assigned allocation and a realistic chance of returning several times the amount at risk.
Suppose $2,000 is allocated to an early-stage company. Failure to secure funding, launch a product or win a major contract could destroy most of the position. A successful outcome might produce several times the original investment.
That opportunity may fit the speculative sleeve when the downside cannot escape the $2,000 allocation and the possible reward is large enough to absorb several unsuccessful attempts.
Five questions expose weak candidates.
What is the maximum plausible loss?
Which event could produce a large gain?
How large is that gain under a realistic outcome?
What would prove the original case wrong?
Would one successful outcome be large enough to offset several failed attempts?
A stock with a plausible downside of 70% and expected upside of only 30% offers poor asymmetry, even when its price moves violently. High volatility creates movement. It does not create a favourable payoff.
Different names can hide the same exposure. Holding Nvidia, a semiconductor ETF and an AI-focused fund may look diversified while leaving the portfolio dependent on the same spending cycle, interest-rate environment and valuation assumptions.
Fund overlap can also create concentration without one obviously oversized position. FINRA notes that concentrated exposure to one company, asset class or market segment can amplify losses.
List the largest holdings and ask what could make them fall at the same time. If the answer is weaker AI spending, higher interest rates or a broad equity sell-off, the positions may share more risk than their labels suggest.
The same test applies to the defensive side. Coca-Cola may reduce exposure to the AI cycle, although both Coca-Cola and a chip stock can decline during a market-wide liquidation.
The practical question is: Would the defensive core still work if the speculative thesis failed?
A barbell needs rules before prices begin moving. Review the allocation at fixed intervals or when either sleeve crosses a predetermined threshold. A speculative sleeve that grows from 10% to 18% has changed the portfolio’s risk, regardless of how strong the winning position appears.
Trimming restores the original boundary. A fall from 10% to 4% does not automatically justify replenishing the sleeve. Add capital only if the original case remains valid, the expected payoff still justifies the risk and the total loss budget allows another attempt.
The original case remains valid.
The prospective payoff still justifies the risk.
The total loss budget has room for another attempt.
Repeated replenishment can turn a supposedly capped 10% loss into 20%, 30% or more over several failed cycles. Replacing every failed bet eventually turns a loss limit into a habit of draining the safe side.
The allocation should show what each part is supposed to accomplish before any securities are selected.
Portfolio role |
Allocation |
Main purpose |
|---|---|---|
Defensive core |
$90,000 |
Preserve liquidity and reduce dependence on speculative outcomes |
Speculative sleeve |
$10,000 |
Pursue concentrated upside within a fixed loss budget |
Before placing the first position, write down the maximum loss for the entire speculative sleeve, the condition that would invalidate each idea and the allocation level that would trigger rebalancing.
Without those rules, 90/10 describes where the money started rather than how the risk is controlled.
Nassim Nicholas Taleb popularised the modern barbell approach as a way to avoid relying on precise forecasts. Mark Spitznagel founded Universa Investments in 2007, with Taleb contributing to the development of its tail-risk work.
Universa describes its approach as protecting client portfolios against severe market losses, including during the 2008 crisis. The broader lesson was not to predict the exact crisis. It was to remain resilient while retaining limited exposure to an extreme payoff.
A barbell can contain the damage from a failed speculative idea, preserve liquidity and retain exposure to unusually large gains. It cannot guarantee positive returns, eliminate inflation risk or make weak speculation attractive.
The structure also fails when both sides depend on the same market condition or when leverage allows losses in the speculative sleeve to spread beyond its allocation.
No. The ratio is a common illustration. The allocation depends on liquidity needs, the acceptable portfolio loss and the risks carried by the speculative positions.
Yes. Treasury, broad-market and focused-sector ETFs can supply different exposures. Their duration, leverage, holdings and overlap must still match the role assigned to them.
CFDs can provide tactical exposure in the speculative sleeve, although leverage can magnify gains and losses. Maximum loss should be calculated from the full market exposure rather than the initial margin alone, alongside the broker’s product terms and account protections.
Yes. The defensive core may lag during a prolonged rally, while several speculative positions may fail before a large winner appears. The outsized payoff may never arrive.
A 90/10 barbell does not tell investors whether AI stocks will recover or fall further. It sets a boundary before the market moves: most capital is kept in assets chosen for stability and liquidity, while a smaller allocation is used for higher-risk opportunities.
The strategy depends on the boundary remaining intact. If leverage, overlapping positions or repeated replenishment allow speculative losses to consume more than the original allocation, the portfolio is no longer operating as a barbell.
The useful question is not whether AI shares are cheap after a sell-off. It is whether the position still fits the loss limit set before the trade was opened.
