Why Shariah-Compliant Stocks Are Heavily Tilted Towards AI

Are Shariah-Compliant Stocks an AI Bet You Never Chose?

On 9 November 2026, Sarah Hunter, Assistant Governor (Economic) at the Reserve Bank of Australia, spoke at the UBS Australasia Conference in Sydney. Her subject was what happens to the economy as "intelligence" gets cheaper. She said the share prices of AI-related companies in the United States have roughly quadrupled since the end of 2022. Reserve Bank staff estimate that about 5% of Australian households' financial wealth is now in AI-related shares, held directly or through superannuation. If the returns on AI investment disappoint, she warned, some asset prices could face "sharp repricing and possibly disorderly adjustments".

Her speech does not mention Islamic finance. It still raises a question for a Muslim saver who holds a halal equity fund. How much of that fund is a bet on artificial intelligence?

The screen leans towards technology

Quite a lot, and more than a conventional fund. S&P Dow Jones Indices compared its Shariah and sustainability screens on the S&P 500 at 31 August 2026. The Shariaah screen removed 237 of the 503 companies. What was left was 49.6% information technology, against 37.9% in the S&P 500 itself. Financials fell to 2.8%, against 12.3%. Over the three years to August, the Shariah index was also more volatile than its benchmark, by 2.15 percentage points a year. S&P puts that down to sector concentration, not to any risk choice.

Global indices show the same shape. On 30 October 2026, the MSCI ACWI Islamic Index was 44.31% information technology. Its ten largest holdings made up 34.78% of the index, against 10.41% in the conventional MSCI ACWI. Microsoft alone was 12.19%, against 3.65% in the parent index. One large US halal ETF shows the same pattern. On 11 November 2026, the SP Funds S&P 500 Sharia Industry Exclusions ETF held Apple at 13.66%, Nvidia at 10.98% and Microsoft at 10.63%. These three names were 35.27% of SPUS.

The reason is mostly mechanical. The screen takes out conventional banks and insurers. Its debt limits then favour companies with a very large market value and modest borrowing. In this cycle, those are the technology companies building and selling AI. The AAOIFI standard dates from 2004, long before this boom. The tilt comes from the rules, not from a plan to buy AI.

The tilt has worked in both directions. In 2024, the MSCI ACWI Islamic Index returned 5.10%, against 18.02% for MSCI ACWI. In 2026, through the end of October, it returned 21.74%, against 11.53%. Both figures are gross returns from the same MSCI factsheet. An investor tracking this index has done well this year. That investor also carries more of the risk Hunter described.

The debt ratio moves with the share price

The debt screen does not compare a company's debt with the size of its business. In most index families, it compares debt with the market value of the company's shares. AAOIFI's Shariah Standard No. 21 caps interest-bearing loans at 30% of market capitalisation. S&P uses an average market value over 36 months, and the Dow Jones Islamic Market indices use 24 months. Two well-known rulebooks take a different path. MSCI's core Islamic Index Series divides by total assets. So does Malaysia's Securities Commission, which sets a limit of 33% of total assets.

The choice matters in a boom. When share prices rise, the bottom of the ratio grows and the debt ratio falls, even if the company borrows more. A 2019 study by Boudt, Raza and Wauters found that Dow Jones Islamic screens had become less restrictive over time for this reason, as market values grew faster than accounting figures. The authors also cite an earlier proposal by Obaidullah to use accounting measures, so the screen would not follow the booms and busts of stock prices.

The same mechanism works in reverse. The screening service Zoya notes that a falling share price raises the debt ratio while the debt stays the same. Averaging over 24 or 36 months slows this down. S&P also lets a stock stay in if it is within two percentage points of the limit, and treats it as non-compliant only after three breaches in a row. I think these rules soften the effect but do not remove it.

AI is now being built with borrowed money

The debt side of the ratio is growing too. The IMF's Global Financial Stability Report of October 2026 says the share of global investment-grade bond issuance taken by hyperscalers, the largest cloud and AI infrastructure companies, has risen since 2024 to 10%, from less than 1%. A BIS Bulletin of 27 October 2026 puts gross bond issuance by the five largest hyperscalers at US$222 billion by September 2026. Their capital spending for the year is expected to pass US$800 billion. Some of the borrowing sits outside the companies' own accounts. The BIS notes that Meta does not consolidate Hyperion, a joint venture that placed US$27.3 billion of bonds backed by Meta's leases.

A debt screen reads the company's balance sheet, so debt held in a separate vehicle may never reach the ratio.

Oracle shows how unsettled this already is. On 3 November 2026, Zoya put its debt at 37.51% of market capitalisation and marked it not compliant under the AAOIFI method. Another service, Halal Terminal, reports that Oracle fails the AAOIFI, FTSE and MSCI screens but passes the Dow Jones and S&P screens, which use trailing averages. On 11 November, SPUS still held it at 0.62%. These are calculations by screening services, not index announcements. Still, they show one company on both sides of the line, depending on the rulebook.

Same company, different verdict

S&P says this openly. Its commentary states that "the same company can be compliant under one index family and noncompliant under another", because thresholds and denominators differ even when the business has not changed.

The rules also change. On 20 August 2026, S&P announced that from 21 September its Global BMI Shariah indices would cut the leverage limit from 33% to 30% and add a new 30% cash test. S&P estimated that the share of eligible stocks that pass would fall from 54.7% to 46.1%. Its November commentary, which uses August data, still describes the 33% test.

Each of these is a Shariah governance decision. A Shariah board and an index provider choose the threshold, the denominator and the averaging window. A fund manager chooses the index. The saver inherits all three choices. AAOIFI's standard adds one more duty. If the conditions can no longer be met, the investor is required to give up the investment. When a ratio rises because prices have fallen, that duty can arrive at a bad time to sell.

Back to Sydney

Hunter was careful in her speech. She said the size and timing of AI's effects are very uncertain, and that the largest gains from new technologies often take longer to arrive. None of this means the AI boom will end badly. A halal fund that tracks these indices holds more of this risk than a conventional one, and the rules that put it there move with the market.

The saver in a halal equity fund did nothing wrong. Every holding passed the screen. Yet if AI returns disappoint and prices fall the way Hunter described, the debt ratios of some of those holdings, measured against market value, will rise after the fall.

So who decided how much AI risk a Muslim saver should carry, and did anyone ask the saver?

Disclaimer: The views expressed in this blog are not necessarily those of the blog writer and his affiliations and are for informational purposes only.

Every FinFormed letter is free, and it will stay that way.

I write these letters in my own time. Hosting, email delivery and keeping the website running cost money every month. If a letter helped you understand something, prepare for an interview, or see an issue differently, a coffee helps me keep going.

Prefer to give monthly? Become a Supporter for $5 a month or $50 a year. Supporters read exactly what everyone else reads. They can also request a one-on-one chat with me on Islamic finance or careers.

Know someone who would find this letter useful? Forward it to them, or follow FinFormed on LinkedIn.

Thank you for reading.