Welcome to the first World According to Rob, a new article series here on X in which I discuss things that are interesting me in the world of markets, tech, AI, infrastructure or anything else that comes to mind.
But first, a brief introduction.
My name is Rob Nunn. I'm 32 years old, and for the best part of 15 years I've made a career out of connecting people, putting deals together and building businesses.
It has taken me around the world, into some rather extraordinary rooms and given me the opportunity to meet people from just about every walk of life. From entrepreneurs and investors to scientists, politicians and heads of government. I've worked across infrastructure, technology, finance and international affairs, often finding myself involved in subjects I knew absolutely nothing about 12 months earlier.
I'm a husband, father of three young kids, voracious reader, occasional magician and distinctly average golfer. I'm fascinated by history, economics, politics, science, technology and, perhaps most of all, people. I have a habit of asking too many questions and disappearing down intellectual rabbit holes with alarming regularity.
My day job today is running Park West, my family office and its associated holding company, through which we originate opportunities, structure deals and establish businesses across multiple sectors and markets. Our interests span infrastructure, technology, financial services and strategic advisory, operating across six continents. Much of what we do involves bringing together people, capital, ideas and institutions to turn opportunities into commercial realities. We tend to focus on creating and structuring the opportunity, leaving the business of actually delivering it to people far better equipped to do so. We also, as you would expect, invest into plenty of industries including banking, insurance and technology companies in public and private markets.
Now, onto the main article... if you found it interesting, please do reply, share and contact me.
________________________
How Hard Can It Be? (About 6%, Apparently)
That inconvenient constraint that everyone's ignored and will become the central theme of the next 6-12 months.
Every investment presentation I've sat through this year has a slide or section or mention about constraints. They don't call it that but if you've sat through even a college start up pitch you know the one. It comes after the chart called revenue expectations that is always shaped like a ski jump. Constraints.
For AI, the constraint is chips, then power. For defence, it's factories and people who know how to weld. For the grid, it's transformers and a planning system that moves at roughly the speed of continental drift.
All very true.
But there's one constraint that rarely, if ever, makes it to the presentation, because for the first half of my career it was very nearly free. The cost of capital. Or, as normal people call it, the price of money.
The three great buildouts of this decade, compute, power and re-armament, are being paid for largely with borrowed money. And all three have turned up at the bank on the same morning, asking for it at the highest price in a generation. That, rather than any shortage of chips or steel, may decide which of them gets built on time, and who ends up owning it.
1. Price:
In September the Federal Reserve raised rates for the first time since 2023, to 4%. The US ten-year yield then went above 5.6%, a number nobody had seen since 2002. It has calmed down a bit since, closing Friday just under 5.1%, in the way a man calms down a bit after being told the price of a new gearbox.
In London, the 30-year gilt hit 6.05% on 8 October, its highest since early 1998.
Every long-dated project in the world is now being priced against those numbers. Most of them were planned when the numbers were half that which is hardly a rounding error, as one presentation for a nameless company tried to imply in a presentation to my firm earlier this summer!
2. Compute:
The riskier end of AI infrastructure debt is already trading at distressed yields, which is the bond market's way of saying "we'll see". But the really interesting bit is further up the ladder, among the borrowers who are supposed to be the safe ones.
At the start of the year, the best-rated data centre bonds yielded below 5%. Now it's close to 7%. One bank's analysts reckon things get properly difficult at 8%. So a three-point cushion has become a one-point cushion in under ten months.
AI-related debt issuance went from a record $113bn in June to $23bn in September. That isn't a slowdown. That's someone yanking the handbrake on the motorway. Big tech had borrowed heavily all year and then, all of a sudden, in September it stopped picking up the phone.
So the total clear of risk, and not at all under any scrutiny, private credit funds are being asked to step in. One large manager reckons it could be called on to fund around $1tn of AI compute, more than half of everything the industry manages! Think about that for a second. A private, relatively unregulated (well under regulated) money market may bare 50% of the entire industry's credit exposure in this scenario.
Meanwhile, nine of the fifteen non-traded funds Fitch tracks had more investors asking for their money back than the 5% quarterly limit allows.
So there's a queue of money trying to get in, and a queue of money trying to get out? Through the same door. At the same time. Which, if you know your near term history begs a question. Why? Why are so many flying for safety as others are gunning for exposure?
Increasingly, I fear that it's less of a strategy and more of an indicator that perhaps no one knows what they're doing.
3. Power:
US utilities are forecast to spend around $1.3tn between 2026 and 2030. One bank puts America's data centre power shortfall at 57 gigawatts by 2028.
For context, that's nearly eighteen Hinkley Point Cs (The World's Largest nuclear power station being built in the UK). Which has been under construction since 2017 and isn't due online by 2030. Impressive, yes. Going to resolve a shortfall in 18-24 months, no.
Utilities pay for all this by borrowing for decades and then ask the regulator to let them send the bill to customers. At 4%, with customers broadly on side, that works beautifully. At 6%, with customers asking why their electricity bill is paying for somebody else's chatbot, it works rather less well.
Bankers are already talking about convertibles and backup project finance. When the bankers start reaching for the creative options, it's usually because the simple ones have stopped working. Which is fine when it's backing a SaaS business but not when it's powering my house!
4. Defence:
NATO's European members and Canada spend an estimated 2.53% of GDP on core defence. The target is 3.5% by 2035. Nobody has quite explained where the difference is coming from, but I think we all know. It's being borrowed. The EU's SAFE programme alone offers €150bn of loans.
In Britain there's talk of defence bonds to plug a reported £15bn gap which is likely to be £30bn in reality. And the situation on the European economy is far worse. Fund managers expect them to price at a spread over gilts. Which is to say, the same debt, but more expensive, with a frigate on the front.
Governments do not stop buying frigates or missiles or drones because yields rise. They borrow anyway and cut something else. And that nudges the price of money up for everyone else in the queue and, in the medium, that becomes sticky with entrenched inflation.
5. We've seen it all before: She sells she shells on the seashore...
I haven't seen many cycles. I read about the ones I missed.
In the mid-1840s, the Bank of England cut rates, government bonds became about as thrilling as a beige estate car, and Britain went completely mad for railways.
In late 1845 the Bank raised rates again. Money went scurrying back to government bonds. Railway shares started to fall, and new investment stopped almost overnight.
Nobody told Parliament. In 1846 it passed Acts for 9,500 miles of new line, the peak of the whole mania, for railways the money had already started to leave.
About a third of the lines authorised were never built. But 6,220 miles were, from schemes approved between 1844 and 1846. Many of the failed ones were snapped up by bigger, richer companies for a fraction of their value.
So the railways got built. It's just that the people who paid for them often didn't end up owning them.
This month, investors put almost $14bn into US government bond funds, after a record $59.3bn in the third quarter.
Feels rather familiar.
6. What I'm watching
Wednesday 28th October. The UK Budget in London and the Fed's decision in Washington, on the same day, a few hours apart. Two rooms full of people setting the price of money on either side of the Atlantic. The first time that that has aligned since, wait for it, 2007. What could possibly go wrong?
After that: whether investment-grade AI borrowing starts again or stays quiet. Fourth-quarter redemption figures at the semi-liquid credit funds. And how many power and defence projects quietly change how they're paid for before they admit to changing when they'll be finished.
None of this means the buildouts stop. The need for compute, power and security is real, and largely indifferent to the bond market.
It means the cost of capital decides the order of the queue, and quite often who owns the thing at the end of it.
So here's the slide I'd add to every deck this year.
The constraint: money.
The solution: time.
The problem: time now costs 6%!
The necessary disclaimer: Nothing in this essay is investment advice, a financial promotion, or an offer or solicitation to buy or sell any security.
