The ZAR Brief

You Get The AI Risk. You Don't Get The AI Reward.

Asia's AI selloff hit the rand harder than anything that happened in South Africa this week.

16 October 2026·The ZAR Brief

Japan's Nikkei index fell 5% overnight. Trading in South Korea was halted. The JSE opened into that storm holding almost no AI shares at all, and the rand still fell.

That gap is the whole story. South African investors are taking the pain when AI shares fall, without owning much of the gain when they rise. When investors get nervous in Tokyo and Seoul, funds that invest in emerging markets sell our currency first. They are not selling a view on South Africa. They are just getting out of risk, and the rand is one of the easiest things to sell.

The comfortable version of this story is that the JSE escaped because it never joined the AI boom. The honest version is harder. The JSE did not escape. Its price was set by money moving around the world, money we have no say over. And it happened in the same week Ray Dalio warned that the bigger adjustment has not even started.

The Big Signal

Nothing Local Changed This Week

That is the point.

South African bond yields moved because American bond yields moved. The rand moved because global nerves moved. And the petrol relief that looked certain when the oil price fell on Trump's "energy ceasefire" claim will mostly disappear, because we pay for petrol in rands, not dollars. Oil can drop 2%, and it counts for nothing if the rand drops 2% in the same week.

Three different headlines. One cause. None of it decided in Pretoria.

The Problem Nobody Wants to Say Out Loud

Allan Gray told clients this week to look at Nigerian shares instead of South Korean ones, to reduce their exposure to AI. Read that again. A Johannesburg fund manager went to Lagos to get away from an AI bubble.

The honest reading is that there was never a local AI trade to get away from. Take out Naspers and Prosus, which both ride on Tencent, and the JSE Top 40 is banks, mines and shops. Our AI exposure is one share, not a sector. You cannot hold a lot of something that is not listed here.

So this is where a South African investor sits in 2026. Fully exposed to global AI nerves through the rand. Barely exposed to global AI profits through the JSE. You carry the risk, and you miss the reward.

What Most Investors Miss

They think they fixed this by investing offshore. BlackRock passed $15.5 trillion in assets this week. That is one company managing money worth roughly half the size of the US economy. A large share of South African retirement savings reaches global shares through a BlackRock iShares tracker fund.

Here is the catch. Those tracker funds put more of your money into the biggest companies. So the "global" part of a typical South African retirement annuity is mostly seven American shares: Nvidia, Apple, Microsoft, Amazon, Meta, Alphabet and Broadcom. Savers believe they spread their money across the world. They put it into seven companies and called it global.

Which means both halves of the average South African portfolio now move on the same thing. When AI shares fall, the local half loses on the rand and the offshore half loses on the shares. The protection people thought they were buying is not there.

The Bigger Picture

Dalio's warning that the selloff in US government bonds has "a long way to go" matters more to our bonds than the next Budget speech. Load shedding, local politics, the deficit: all of it matters less while US interest rates are climbing. We spend enormous energy arguing about our own problems and very little on the US rates that set the floor under what it costs us to borrow.

Meanwhile, local asset managers are already feeling it. PSG Financial Services reported 5% earnings growth this week, but its performance fees fell 74%. Performance fees are what a manager earns when it beats the market. A flat JSE means there is nothing to beat. The market read the result as good news. It is really a warning for every other JSE manager earning that same kind of fee.

The uncomfortable conclusion: JSE returns increasingly come from abroad, while the way managers earn their fees still assumes those returns come from here.

Stock Spotlight

Telkom SA

If this week's theme is that South African returns are decided offshore, Telkom is the test case, because almost nothing about Telkom is offshore. All its money is made here. It is listed only on the JSE. It earns no dollars to cushion a weak rand. Government and the PIC together own close to 40%. No global money is coming to save it or to sink it.

What the Market Believes

That Telkom is an old fixed-line phone business dragging along a mobile business nobody bothers to value properly. It trades at about 3.5 to 4 times EV/EBITDA, a rough measure of what you pay for each rand of operating profit. Vodacom trades near 6 times and MTN near 4.5. On expected earnings, Telkom sits near 8 times against Vodacom's mid-teens. In plain terms, the market treats Telkom as the cheapest and the weakest of the three.

What the Market May Be Missing

Telkom Mobile now has more than 19 million subscribers, and its mobile service revenue is growing close to 10% while the whole group grows 2% to 3%. Openserve is the wholesale fibre network that BCX, Telkom's own consumer arm and outside internet providers all pay to use. That is real infrastructure, in a market where almost no infrastructure of that kind is listed at all.

What Actually Matters

Whether Telkom actually sells assets, not whether it keeps talking about selling them. The Swiftnet cellphone tower sale brought in close to R6.75 billion. Talk of selling or separating Openserve comes up regularly and never happens. Telkom trading at 4 times against Vodacom's 6 only makes sense if BCX keeps losing money and the asset sales stall. Right now the market is expecting both.

What to Watch from Here

Whether the next sale actually closes, and what the price war with Rain and Cell C does to profit margins in mobile, the one part of the business that is working. Until a deal closes, this stays a cheap share with a good story attached rather than a share the market has decided to value more highly. Which raises the wider point this week: when market sentiment is imported, the only way a JSE share gets valued higher is by fixing itself.

Read the full analysis for Telkom here

Quick Signals

Three things SA investors should watch this week

  • BlackRock passes $15.5 trillion. One firm now manages money worth about half the size of the US economy, mostly in tracker funds that buy more of the biggest companies. Why it matters: your offshore money and your rand are now moved by the same seven American shares.
  • PSG performance fees fall 74%. Overall earnings still grew 5%, which is what hid the drop. Why it matters: Coronation, Sanlam and every other manager earning fees for beating the market face the same flat JSE. PSG simply reported first.
  • Oil falls on "energy ceasefire" claims. Oil down, rand weak, fuel levies unchanged. Why it matters: no real relief at the pump means no real drop in inflation, which gives the Reserve Bank less room to cut rates, not more.

AI + Finance

Every South African bank now has an AI story. Assistants for staff, chatbots for customers, "AI-powered" language in the annual report. That is the top layer of AI, and it is the layer losing its value fastest, because Capitec, FNB and Standard Bank can all buy very similar tools from the same few suppliers. If everyone buys the same thing, nobody gains an advantage.

The real money sits one level down, in chips, computing power and data centres. Salesforce and Adobe are fighting over who sells AI software to businesses, while the companies supplying them the hardware underneath collect the profits. Almost none of that layer is listed in Johannesburg in any meaningful size.

Which is this week's theme again, showing up in company spending instead of in the currency. South African companies are buying the part of AI with the weakest lasting profits, and calling it strategy.

The real local advantage will show up quietly, in catching fraud and deciding who gets credit, years before it appears as a product promising to beat the market. A former theatre director beat Goldman Sachs and JPMorgan on US inflation forecasts from his garage this year using no model at all. The advantage was never the label.

Leave a review

Reviews are approved before appearing. Featured reviews may appear on the homepage.

0/1000