AI meets the ballot box: the week bonds went back to work

Last Thursday the Nasdaq 100 fell 6.36%, its worst day since March 2020. It recovered some ground on Friday and finished the week down a little under 5%.

On Monday, US markets bounced hard. The S&P 500 rose 2.3%, the Dow 2.2% and the Nasdaq Composite 2.7%. Two things did most of the work. Brent crude fell almost 6% after President Trump said renewed strikes on Iran were “highly unlikely” if talks progressed, and buyers came back to the chip stocks sold off last week. Nvidia was the exception, down about 4% on company-specific news. Markets are also positioned for this week’s US inflation figures.

A swing like that shows how fast sentiment can turn. It also tells you something about last week. Most of the damage was in AI hardware, a lot of the selling was on autopilot, and US government bonds rose while shares fell. The bigger change may be in the AI debate itself. The question is shifting from whether the demand is there to who pays for the build.

We don't take any one view as gospel. We read the research, talk to people managing money and check what we hear against the numbers. This week that meant Jason Coggins' weekly summary of global and Australian broker research, and Nicholas Crown's Crown Macro Letter.

What happened

The US midterms delivered a Democratic sweep of both houses of Congress. As Crown points out, that has not traditionally been a blanket negative for markets. What mattered was where some of the key wins landed.

Goldman Sachs estimates that state-level pauses could potentially affect about 45% of planned US data centre capacity. Texas alone accounts for about a quarter of that planned capacity nationally. JPMorgan counts 42 of 45 local ballot measures against data centres as passed or leaning that way. The AI build-out, which has driven a large share of US market gains, now has a political problem.

Crown's take: the semiconductor trade was already crumbling, and the result gave it a reason to blow up.

What they say

The research falls into two camps, and both have numbers to back them.

Sources: Martin Wolf (FT), Goldman Sachs, JPMorgan, South Korea's trade ministry and the FT's Unhedged column.

Two details sit behind that table. Oracle's newest bonds pay coupons of 7% to more than 8.5%, up from 5% to 7% in February. And Goldman's construction analysts read the election as a delay, not a cancellation.

Crown sits somewhere in between. In his view the customers held up while the suppliers were repriced. The big cloud companies have other businesses to fall back on; hardware makers need the next order. He also thinks credit is starting to force some spending discipline, with funding costs rising just when the build-out needs more money.

What the data says

Three things stood out to us in this week's research.

1. Much of the selling was mechanical.

A lot of that was computers and margin calls, not investors deciding each company was suddenly worth less.

2. The damage was concentrated. On the day, the selling spread well beyond AI, and Crown warns that's how an AI sell-off spills into the whole market. But the real damage was in hardware. As of Thursday, Deutsche Bank had the Philadelphia Semiconductor Index more than 38% below its June peak, and the cost of insuring CoreWeave, a specialist AI cloud provider, against default jumped. The market looks to be separating the stronger AI businesses from the weaker ones, not walking away from the theme.

3. Bonds protected portfolios again. Over the month to last week, US Treasuries gained close to 3% while the S&P 500 lost about 5%. In recent years, when inflation was the main worry, bonds and shares often fell together. Last week the old pattern came back. As the AI trade sold off, yields fell and bond prices rose.

It won't always work. US payrolls rose 165,000 in October, well above expectations, and markets moved back towards pricing a December Fed rate rise. If inflation, rather than an AI scare, drives the next move, bonds and shares could fall together again.

Closer to home

Australian investors felt this too. Most hold global shares through super, index funds or managed funds, and the largest US technology companies make up a significant share of global indices. You can end up concentrated without ever deciding to be.

Australia also has its own build-out under way. Deloitte Access Economics counts 37 data centre projects worth A$141 billion in its national investment pipeline.

The rate backdrop is different here. The RBA raised the cash rate to 4.85% on 3 November, its fifth rise this year and the highest since 2008. Capital city home prices are 8.1% below their March peak. Thirty of 32 economists polled by the AFR expect the next move to be down, though not before late 2027, while market pricing still gives about a one-in-three chance of another rise by February.

The questions we're asking

A few questions we're keeping an eye on:

• Is this 1998 or 2000? Goldman Sachs' Rich Privorotsky sees a 1998-style accident, the kind that turned out to be a buying opportunity, rather than the start of a slump like March 2000. His condition is that AI demand holds. Nomura expects about a month of choppy trading first.
• Who funds the next stage? If borrowing costs keep rising for the weaker builders, spending may be cut by the lenders rather than by choice. The pricing of the large AI listings now being marketed will be an early test.
• Does capital simply move? JPMorgan expects developers and capital to shift towards more receptive locations, including sites abroad. Whether any of it ends up in Australia's pipeline, we'll be watching.
• Do bonds keep doing the job? This week's US inflation data and Treasury auctions, and the Fed's decision on 9 December, will show whether the rally in bonds holds.

Our thoughts

Interestingly, the bulls and bears disagree on where this goes but largely agree on what's driving it. AI has gone from a demand story to a funding story. Who can pay for the next stage, and at what cost, may matter more from here than who makes the best chips.

For families, none of this is about picking the bottom in chip stocks. Two things matter more:

• Know what you own underneath. A global shares fund, a technology fund and a "quality" fund can all hold the same handful of companies. A genuine defensive allocation is what gave portfolios a cushion last week.
• Have cash to fall back on. For someone drawing an income, a 6% day is a reason to check the cash reserve. For someone with twenty years ahead, it may be noise. Cash or bonds to draw on means you can deal with a fall calmly, not in a hurry.

We view these frameworks not as prediction tools, but as a way of understanding how capital is moving through markets. Performance appears increasingly determined not by what markets do, but by where capital is positioned.

If last Thursday made you wonder how much of your portfolio rides on the same few companies, that's a question worth answering before the next one. We offer a complimentary second opinion on existing portfolios. There is no cost and no obligation, and nothing is recommended unless we are formally engaged.

Enjoy.

Murdoch Gatti, Founder and Private Wealth Adviser, York Wealth Management

Sources: broker research summarised by Jason Coggins (week ending 7 November 2026), The Crown Macro Letter (9 November 2026), South Korea's trade ministry (October trade data, via Yonhap). Third-party views are the authors' own and are not York's views. No representation is given, warranty made or responsibility taken about the accuracy, timeliness or completeness of information sourced from third parties. Because of this, we recommend you consider, with or without the assistance of a financial adviser, whether the information is appropriate regarding your particular circumstances.