The Weekender is my bi-weekly take on macro shifts and emerging themes. It’s not investment advice — or even our firm’s official view. I aim simply to inform, challenge, and maybe entertain. If you’d like this in your inbox every other Saturday morning via Northern Trust, subscribe to The Weekender.

Running Towards the Flame

Sometimes the news flow overwhelms, and it's hard to find signal in a cacophony of mostly negative headlines. Now feels like one of those times. If AI doesn't get me, a cybercriminal, a French bond villain or a Russian plague outbreak probably will. Pick your poison from there: diesel prices, policy hawks, datacenter protestors, hurricanes, narrow leadership in markets (none at all in some politics), too much capex, not enough return. Add a few beautifully articulated bear cases, usually from middle-aged men selling books or a broker selling low transaction costs (I speak as a former broker), and I'm urged to act.

Pessimists are so damn convincing.

Then I remember two things.

First, negativity bias is an evolutionary hangover. A useful defense mechanism once, but I've moved on from my scantily clad, cave-dwelling ancestors and am reliably informed that I now possess a prefrontal cortex (for deeper thinking, I think).

Second, a younger contemporary — with fewer biases and bad habits than me — recently shared on the Compound and Friends podcast (listen from 01:04:50 onwards) that an investor who missed the ten best days in the S&P 500 Index each year since the 1950s would have seen annualised returns fall from around +8% to roughly -12%. Minus twelve! Hindsight is perfect, of course, and the best days tend to live next door to the worst. But the cost of flinching is high.

So I'm not flinching.

In fact, if you're reading this on Saturday, I'll be on a plane, running towards the flame and heading for one of the places the headlines suggest I avoid. Not out of bravery, but for clients (and a healthy respect for bonus season). That, and I'm curious. As John le Carré warned, a desk is a dangerous place from which to view the world.

Who Else Isn't Flinching?

The market, for one. The Federal Reserve raised rates last month for the first time in three years, the ten-year yield has traded above 5%, and equities have refused to buckle. That may not be irrational. Corporate America remains in relatively good shape. Much of the S&P 500's debt was termed out at low fixed rates, interest coverage remains healthy and higher yields have yet to meaningfully impair earnings across the index. Higher real yields are painful for housing, utilities and other bond proxies, but they may also reflect a higher economic speed limit, a higher equilibrium, especially if AI is raising productivity and returns on capital. Growth is supposed to be bad for bonds. The curious development would be equities failing to clear the higher hurdle. So far they have cleared it, and in a midterm election year at that, historically the weakest of the cycle.

Digging In

We talk a lot about signals in this missive, and pay attention when things that should happen don't. Stocks rising on bad news has tended to be a good sign. An asset refusing to fall on bad news is a subtler one. Take gold. It yields nothing, real yields have made new highs, and still it has failed to make a new low for the move. It is digging in, perhaps anticipating further debasement, or financial repression, or simply betting that the Fed's hawkishness will fade. If last month's hike proves "one and done," expect comparisons with March 1997, when Fed chair Alan Greenspan tapped the brakes once in a raging bull market, then stood back and trusted technology to do the disinflating. We all know what followed in equities — the S&P 500 returned more than 30% for the year. Fewer remember what happened to bonds. It might pay to look, for risks have two tails%tail%.

Faster Towards the Flame

AI insiders aren't flinching either. If anything, they are speeding up. As discussed last time, the lead indicators for AI capital expenditure were accelerating, and they still are. South Korean exports rose 83.5% year on year in September, the fastest pace since 1974, as chip shipments more than tripled. Memory makers continue to report robust results and sign multi-year supply agreements, much to the chagrin of those crying "peak." Graphics processing unit (GPU) pricing remains remarkably resilient, even for older generations, which strengthens the collateral behind AI credit.

SpaceX may have sent an even stronger signal, one that supports Nvidia CEO Jensen Huang's contention that compute is becoming revenue. Having raised $86 billion at its IPO in June, SpaceX is reportedly in talks to borrow as much as $40 billion more to buy chips. The implication is intriguing. If compute shortages persist and pricing stays firm, existing datacenters may be able to refinance against growing cashflows, while operators with spare capacity could re-let it at higher rates.

This cannot last indefinitely. Nothing does. But it may last longer than the skeptics expect. The first phase of AI was funded largely by equity. The current phase is increasingly led by credit. The next may be funded by the cashflows of the infrastructure already in place.

Imagine that headline. You'll have to, for it is far too positive to attract attention, and attention is what sells advertising.

Birdseye, Not Edison

We should be careful not to confuse the cost of compute with the cost of intelligence. The cost of compute remains expensive. The cost of intelligence is falling fast. If AI follows the path of previous general-purpose technologies, the price of the technology itself will trend towards marginal cost, while value migrates to whatever is built on top. Regular readers will recall my favorite example. Electricity’s biggest impact arguably came not from Thomas Edison's utilities but from what others built with it. Elisha Otis remade the city with the elevator. Clarence Birdseye remade the food chain with the freezer.

So who are the Birdseyes this time? We've discussed some candidates already, cybersecurity and biotech among them (I'm fascinated to see what emerges from Anthropic's new biology "wet lab" in the Bay Area). Here is another. Could one of the first killer apps of the agentic world be not compute, but digital money?

No Ancestors in the Agentic Economy

Which brings us to the next thing that doesn't flinch. It has no negativity bias (I hope), no emotional baggage and no ancestors at all.

The AI agent.

Agents are proliferating. Grok Bot came first, Meta's Muse followed, and OpenAI's Dots arrived at the end of September. As they multiply, they will not only consume more compute, they will also need a financial system designed for machines as much as for humans.

Consider how the internet was paid for. Subscriptions worked for some, but most of the web relied on advertising. Human attention became the product. But bots don't click on ads, as Cloudflare's Matthew Prince likes to say, and he should know. His network sits in front of roughly a fifth of the web, and in May automated traffic on it overtook human traffic for the first time. The internet economy monetized attention. The agentic one will have to monetize data. So who pays?

The agent does.

Cloudflare now lets websites charge machines for access, and in August it gave agents wallets of their own. Data has always been valuable. What it never had was an efficient mechanism to charge machines for using it. Now it is becoming a toll, paid per use, by machines.

Welcome to the Agentic Economy. If it develops as its proponents expect, the payment rails will presumably be on-chain, given the sums (tiny) and the speeds (milliseconds). An agent can't open a bank account. It can hold a wallet.

I will leave others to draw the investment conclusions, but one interested party is the U.S. Treasury, because dollar stablecoins are backed largely by short-dated government debt. More agents could mean more stablecoins, and more stablecoins, more buyers of that debt. The marginal buyer of American debt may increasingly be found not in a vest in Greenwich but in a digital wallet in Silicon Valley. The Guiding and Establishing National Innovation for U.S. Stablecoins (GENIUS) Act takes effect on January 18, and another small piece of finance moves on-chain.

Open All Hours

Digital money may be only the beginning. Where cash goes, collateral tends to follow, and regulators are clearing the path. Last month, the U.S. Securities and Exchange Commission opened a five-year, conditional path for <span class="highlight">tokenized%token%</span> stocks to trade on-chain, on venues that need not register as exchanges. The U.S. derivatives regulator said brokers and clearing houses may hold customer funds in tokenized versions of assets they were already allowed to own, Treasuries and government money-market funds among them. And the Depository Trust & Clearing Corporation, which sits at the center of U.S. securities settlement, is due to launch its own tokenization service this month.

If that sounds like plumbing, trust your instincts. It is plumbing. But plumbing matters, and these are big pipes. By one industry estimate, there are some $300 trillion of high-quality liquid assets in the world, and only a tenth or so is in use as collateral at any given time. Tokenized assets, stablecoins aside, amount to about $39 billion. That leaves a long way to go, and it will take years, but the direction of travel seems clear. More of finance is likely to move on-chain.

So what? Well, for one thing, on-chain markets don't close. They trade continuously, because their newest participants don't sleep, don't eat and don't take weekends off. A market that never closes never really opens, which means my long-held ambition of ringing the opening bell is fading fast. It also raises uncomfortable questions for a publication called The Weekender.

You Can't Embargo the Sun

Machines may not eat or sleep, but they are hungry all the same. For electrons. Which is how clean power, having stalled as climate politics, found three less controversial names: energy security, affordability and AI. Google's deal this week to buy 3.6 gigawatts of power from Constellation, a quarter of it new nuclear, is the latest reminder.

The war supplied another. Hearing Ed Conway discuss Trade World in Henley on Thursday reminded me that consumers don't buy crude oil. They buy refined products. Crude is flowing through the Strait of Hormuz again, yet product shipments from the Gulf are by some measures barely half their pre-war level, and by others far less. Europe, having let its refining base shrink, took half its seaborne diesel imports from America in August, twice the share of a year ago. This is climate policy meeting thermodynamics: outsource the dirty work and you outsource the leverage too. I doubt many new refineries get built. But renewables may get a rebrand, because if this conflict has taught us one thing, it is that you can embargo fossil fuels, but you can't embargo the sun. Solar's greatest ally may turn out to be not Greta Thunberg, but Sam Altman. Another headline unlikely to trouble the front page.

The Moth and the Firefighter

Of course, not flinching cuts both ways. An agent with no negativity bias has no fear either, and fear, for all my mockery of the cave, kept my ancestors alive long enough to produce me.

The risks are real. Even the labs warn about the risks associated with increasingly autonomous systems. Bridgewater Associates founder Ray Dalio calls AI a classic bubble, and the build-out leans ever more heavily on debt. The stablecoin law arrives in January with parts of its rulebook still unwritten. Markets that never close have no time to think. And machines that won't pause in a panic could turn a run into something measured in milliseconds, not days. These are genuine left tails.

So are the right ones. Gold has absorbed plenty of bad news and refused to break down. Equities have done much the same. AI insiders are spending more, not less. And a new economic actor is arriving with a wallet, an objective function and no ancestors.

The moth and the firefighter both head towards the flame. The difference is that one of them has read the signals.

Which is why I'm on the plane.

Curious, not brave.

And still not flinching.

Have a great weekend.

Gary