AI fears come roaring back, and the IMF warns the AI boom is adding fragility

AI fears returned in October led by graphics processing unit (GPU) providers (CoreWeave, Nebius and Elice) and AI memory makers (SK Hynix, SanDisk and Micron), which fell between 20% and 40% in the month. Investors are concerned that the trillions being poured into AI won't see sufficient returns. The International Monetary Fund's (IMF's) October Global Financial Stability Report identified three risks that could tighten financial conditions, all of which trace back to the AI investment boom.

AI-driven market correction

Rise in sovereign bond yields

Cyber threats

Compute demand and the capex cycle

The current shortage of computing capacity is keeping prices elevated, with GPU rental prices up 40% year to date (see chart). Lower token prices are spurring demand. There has been a 250x reduction in token costs for the same model since 2023, making it the fastest cost reduction for any new technology in history (BCA), and token demand has increased by 330x over the same period, making it the fastest adopted technology. Agents, not humans, now generate most of the token demand. Agents surpassed humans in token use in February 2026, and they now consume 14 times as many tokens as they did then, which will support demand; the hyperscaler compute cloud backlog remains huge, nearly $2.4tn (Calcbench, September 2026). In addition, early evidence of AI productivity is showing up, with S&P 500 revenue per worker at record levels. We remain overweight global equities and are watching compute prices and capex announcements closely.

The fixing of fiscal policy: Has the MTBPS delivered?

In the early post-1994 period, the Medium Term Budget Policy Statement (MTBPS) was released with very little fanfare or attention. Most market watchers and investors kept an eye on the "real deal" of the Budget, which followed in the February of the new year. This is no longer the case, however, as the National Treasury now effectively runs a twice yearly fiscal signalling process - with both the MTBPS and the Budget offering crucial information about the state of expenditure and revenue allocation in the country.

The headline of this year's MTBPS for me is that five years of fiscal consolidation in the wake of State Capture are finally beginning to bear fiscal fruit. A glance over the major numbers, for example, shows that debt, expenditure make-up and revenue trends all indicate progress. Llower debt levels, declining wage bill growth rates, a slight increase in infrastructure expenditure (read investment over consumption) and a focus on increasing public sector efficiency are all very strong signs that Treasury is finally seeing positive returns on its fiscal stance. This is no small feat in a very tough global environment brought about by war in the Middle East, higher oil prices and borrowing costs, and a 2026 growth target that has been revised downwards from 1.6 to 1.2%.

The seriousness of the budget is nowhere more apparent than in the strong statement of intent around the four principles underpinning fiscal strategy. The Minister of Finance noted in his speech that these four principles support growth through prudent fiscal policy and more capital investment; improving value for money by cutting waste; reducing debt-service costs while managing the debt burden responsibly; and entrenching fiscal sustainability through a principle-based fiscal anchor. This is music to an economist's ear - and very good for bondholders. Most heartening, though, is not the principles per se - any finance ministry can parrot these - but that they are now backed by a credible track record under Minister Godongwana and his team (led by Duncan Pieterse).

The most important development arguably lies in the fourth principle. Draft legislation, to be tabled in 2027, will write a fiscal anchor into the Public Finance Management Act, defining fiscal sustainability in law for the first time in South Africa's history. Crucially, this anchor will not be a rigid numerical rule: instead, after each general election, the incoming Cabinet must approve a fiscal plan, report against it at every Budget and MTBPS, and explain any material deviations to Parliament - with the Fiscal and Financial Commission (FFC) independently assessing its consistency. Fiscal rules are not unusual: over 100 countries have them, from the European Union's (EU's) '3 and 60%' deficit and debt ceilings to Germany's constitutional 'debt brake'. South Africa has wisely opted for a principles-based route, closer to New Zealand's or Australia's, where transparency rather than hard numbers will provide the discipline. Given how often numerical rules are suspended in a crisis, this seems the smarter choice - but, as always, the bite of the fiscal anchor will depend on how seriously Parliament, the FFC and non-ruling parties engage with their new roles.

The debt dynamics tell the story of a decade of fiscal drift finally being attenuated, although the figure below show that the road to debt freedom will be a long, slow slog. As the data show, after gross debt-to-GDP rose relentlessly from 67.7% in 2021/22 to close to 79% in 2025/26, it has finally stabilised and even reached a turning point. For the first time since 2008, debt levels have stopped rising - and are projected to fall to 76.6% by 2029/30. One idiosyncratic but not unimportant point reflected in this graphic is that the Budget 2025 estimates all show a lower debt ratio. This is not a loss of fiscal credibility update that comes through in this MTBPS, but rather a reflection of lower inflation, which in turn lowers nominal GDP values and mechanically makes the debt-to-GDP ratio worse than predicted in the higher-inflation environment of Budget 2025. Underpinning the improved MTBPS debt outlook is the primary surplus, which is expected to rise from 1.0 to 2.9% of GDP by 2029/30, while the all-important deficit-to-GDP ratio will narrow from 3.9 to 2.5%. Our deficit position suddenly looks seriously credible, leading to whispers of a ratings upgrade. Debt-service costs remain stubbornly high, however, driven largely by muted revenue amidst anaemic growth. These service costs are projected to stabilise at around 20.6% as share of revenue, but at R441.5bn this year it still exceeds our spending on health or the entire social development function.

On the expenditure side, the broad contours have barely shifted. Spending is expected to grow at 3.6% a year to R3.02tn by 2029/30, with the social wage still absorbing about 60% of non-interest spending. Additions are at the margin, where there are new allocations to municipalities and provinces, gender-based violence and policing. An important and highly welcome shift is that capital spending is now the fastest-growing item, at 5.3% a year.

As the data below show, revenue is also better than expected, revised up by R17.1bn this year, with corporate income tax alone up R34.2bn (but almost all lost through the R30bn fuel levy reduction). Given the recent commodity price surges, most of the heavy tax lifting was done by the mining sector, followed by finance and manufacturing. The fuel levy reduction announced by the Minister cuts the general fuel levy by R1.50 a litre on petrol and R3 on diesel until 1 December, amounting to a reduction of about 37% for petrol and 76% for diesel.

The public-sector wage bill has again showed impressive discipline. Its share of non-interest spending has fallen from a peak of 36.1% in 2018/19 to 33.4% - or as a share of GDP, from 10.2% to 9% by 2029/30. Wage bill growth is slowing from around 6% to 3%, aided by the ghost worker audit and the Early Retirement Programme saving R4.4bn a year. The current wage agreement only runs to 2027/28, but Treasury's hard work in moderating wage demands will likely hold into the new agreement.

Ultimately, though, fiscal policy has done as much heavy lifting as it can in a heavily constrained growth and revenue environment. With growth forecast at only 2% in 2029, debt reduction is dependent on squeezing spending relative to revenue year after year. The only sustainable way out is faster growth. Higher levels of economic growth will lift revenue organically and not via volatile, short-term commodity windfalls. In doing so, this enlarged denominator of the debt ratio will pull down both debt-to-GDP and the share of revenue swallowed by interest. This is the true dividend of fiscal discipline: space to tilt spending away from consumption and towards capital investment. The MTBPS has fixed the plumbing of fiscal policy - the task now is to turn on the tap of economic growth!

Top 5 funds

Defensive funds lead the way over one month, with the Berkshire Hathaway fund in top spot and healthcare next in line. The FTSE fund is also relatively defensive, with low technology and high energy weights.

Emerging markets (EMs) still lead the pack over 12 months, with Japan and healthcare in the following positions, but it is noticeable that 12-month returns have become much lower as the impact of higher oil prices and interest rates is felt.

1-month absolute performance as at 29 October 2026

12-month absolute performance as at 29 October 2026

Global bonds a "once-in-a-generation" opportunity

JPMorgan is describing current elevated bond yields as a "once-in-a-generation" opportunity. It argues that bonds are a competitive asset class for both income and hedging. BCA's research supports the hedging case (see chart) after it examined 16 episodes in which a country entered recession with government interest expenses above 4% of GDP. In 14 of those cases, bonds provided protection through the recession, even though fiscal balances deteriorated in most cases. In the two cases where bonds failed to hedge, currency weakness was the cause. BCA argues this is unlikely for the United States (US), where imports are only 14% of the economy - and 93% of those imports are invoiced in dollars.

Yields may still rise before normalisation is complete, but the adjustment is largely behind us. The Bloomberg consensus is for nominal GDP growth to decelerate to 4.7% in 2027, and expectations of at least three more hikes by the end of 2027 have pushed yields well above the long-term equilibrium. Alpine Macro research shows that similar past overshoots were followed by a 9% return over the next year.

The most likely driver of a short-term reversal is an oil price reversal, but real growth would need to crack for Treasury yields to decline meaningfully. A growth disappointment could be driven by a slowdown in tech capex or residential real estate. The 30-year US mortgage rate jumped from 6.2% in January to 7.4%, and US pending home sales fell to a record low. While we have been buying global bonds, we remain slightly underweight until we see a decline in oil or a slowdown in nominal growth.

US presidential disapproval is at levels last seen before a loss

At 63%, Trump's presidential disapproval is close to previous peaks over the last 80 years. Each of those peaks was followed by the incumbent president's party losing the White House at the next election: the Korean War and scandals under Truman (1952); Vietnam and unrest under Johnson (1968); Vietnam and Watergate under Nixon (and then Ford) (1976); the Iraq war and recession under George H W Bush (1992); and the global financial crisis under George W Bush (2008).

Of course, anything could happen over the next two years, especially if the Democrats take both the House and the Senate and make tough and unpopular decisions - despite a strong economy, corporate tax receipts are plunging as companies avail themselves of generous tax breaks under 2025's Big Beautiful Bill Act. A Democratic sweep of Congress in the midterm elections could open the door to corporate income tax hikes.

Alpine Macro has analysed market performance around midterm elections. Since 1970, over 14 cycles, the S&P 500 has generated an average price return of 13.8% in the nine months following midterms and has posted positive returns following every single midterm. The average price return over the nine months leading up to midterms is 1.6% - this year the market has gained 11%, one of the strongest run-ups to Election Day since 1970, which may temper the positive post-midterm rally and, combined with the prospect of tighter regulations across technology under a Democratic sweep, could see a smaller rally than history suggests.

China's export engine runs hot while stimulus gathers pace

Third-quarter GDP growth came in at 4.5% year-on-year (y/y) (see chart), raising the likelihood that China will meet its annual growth target of around 5%. Exports are underpinning this growth and surged the most since February, with shipments abroad up 36% from a year earlier and imports up 31%, pushing global imbalances to new highs. The trade surplus has forced the European Union (EU) to negotiate a cut in allowances for Chinese electric vehicles entering Europe, and Germany is looking for more curbs on Chinese imports.

Domestic demand remains muted, and Chinese authorities have unveiled a mini stimulus package that includes:

Chinese stocks now trade at a historically wide valuation discount to developed markets and even to other EMs, measured by forward-cyclical-adjusted personal expenditure. Despite the attractive entry point, we are waiting for policies to translate into housing demand and robust consumer sentiment.

Outlook: Strong activity, but oil, policy and AI-funding risks remain

The IMF raised its 2026 global growth forecast to 3.1%, up from 3.0%, for two reasons. First, the energy shock has been absorbed by rerouted supplies and inventory drawdowns, which would otherwise have cut growth significantly. Second, heavy AI-related spending is underpinning growth, particularly in the US.

This is supported by activity data, which remain strong. Broad output growth across sectors lifted purchasing managers' index (PMI) sentiment on employment and trade. The acceleration in global PMIs to multi-year highs in Q3 points to earnings growth of around 20% in the EU and the US, according to JPMorgan. Taiwan's export orders, a leading indicator of earnings, are also accelerating. As such, we remain overweight global and local equities.

The global oil market faces a two-speed recovery, as crude availability is recovering but refined product supply is being hampered by capacity constraints. Diesel and other middle distillates remain under the greatest pressure, and crack spreads and product prices are likely to stay elevated in the near term, which is feeding into inflation. Oxford Economics estimates that the energy shock from the Iranian conflict and the AI buildout are the two biggest contributors to sticky inflation, and they have pushed their peak in US core personal consumer expenditure (PCE) out to Q4 2026. Politics remains a further risk, with US and SA elections in November. This has kept us underweight global bonds, overweight SA bonds and neutral overall bonds.

According to BCA, approximately one-fifth of US equity market capitalization is attributable to multiple expansion driven by investors pricing in substantial long-term earnings growth from AI. Strong growth is battling high oil prices and AI uncertainty, and volatility is likely to remain elevated in the short term.