The day two countries lost power
On 28 April 2025, at 12:33 Spanish time, the power went off across the whole of mainland Spain and Portugal.
I moved to Portugal after it happened. On the day, trains stopped and card machines went down. Portugal's grid wasn't fully back until 00:22 the next morning.
Europe's grid operators later called it the most severe blackout on the continent in 20 years (ENTSO-E, March 2026). I'd been reading about grid problems for a while. Seeing two whole countries lose power at once made it a lot less abstract.
If you've read my free report, The $106 Trillion Trade, you'll know where this is going. If you haven't, there's a link to it further down.
This letter carries on where the report finishes. It's about following a theme this big every week, including the weeks when the headlines get loud and you start doubting yourself.
The AI pyramid
In May 2026 I ran a FINK group call called Hidden Themes That Will Define 2026. Afterwards, the thing people asked me about most was something I call the AI pyramid.
Everyone wants AI and AI needs chips. Chips need memory. The data centres full of those chips then have to be connected up, which takes optical kit and a lot of cable.
At the bottom of the pyramid is the layer people tend to forget about, which is power. Every data centre needs a lot of it and it needs it all the time.
This is what I said on the call.
"We have to rebuild the world basically in terms of infrastructure… And the existing grid cannot at all cope with this new power demand. It cannot."
Me, FINK group call, 13 May 2026
I still think that's right. The numbers in the report back it up.
McKinsey puts the bill at $106 trillion
McKinsey reckons $106 trillion has to go into infrastructure by 2040 to keep up with what the world needs (McKinsey, September 2025).
Obviously nobody writes that as one cheque. It gets spent on thousands of specific jobs, like a substation, a subsea cable or a data centre with its own power station on the side. Each one has someone paying for it and someone getting paid to build it. I'm interested in the people getting paid.
The big money has already worked out where the urgency is. Energy and digital make up about 40% of the $106 trillion, but they're already taking about 75% of recent private deal value (McKinsey, March 2026).
Private infrastructure funds held about $1.5 trillion in 2024, up from about $500 billion in 2016 (McKinsey, September 2025). In 2025 they raised nearly $200 billion, which was a record (McKinsey, March 2026). So there's no shortage of money looking for projects.
The catch is that private money follows government decisions. A fund can't lay a cable until someone has approved the route and the grid connection. So a lot of what happens over the next few years comes down to governments deciding to get on with it.
Heathrow and a 57-year-old substation
You don't have to look very far for an example of what old infrastructure costs.
Late on 20 March 2025 a transformer caught fire at the North Hyde substation in Hayes, west London. Within hours Heathrow had lost power to some of its critical systems and the airport stayed shut for most of the next day.
More than 1,350 flights and almost 300,000 passengers were disrupted. The airlines put the damage at £80m to £100m (the Guardian, July 2025).
The substation was 57 years old. Back in July 2018, oil samples had shown a worrying moisture reading in the part that later failed. The official review found the action that reading called for was never taken (NESO, July 2025). That's nearly seven years of nobody getting round to it.
Britain has now started paying for the upgrade. Ofgem has approved £28bn of initial investment in gas networks and electricity transmission for 2026 to 2031 and expects it to grow to about £90bn.
Ofgem's own figure is £108 a year on a typical dual fuel bill by 2031. Take off the savings it expects and it's about £30 a year (Ofgem, December 2025). So whether you invest in any of this or not, you're going to be helping pay for it.
Follow the build-out every week with us. £209 a year for the first 50, kept for as long as you stay subscribed.
Take a founding place →Education, not financial advice. Capital at risk.
Why it's urgent now
Infrastructure used to be a fairly dull policy story. Data centres are what made it urgent.
The IEA expects data centres to use about 950 TWh of electricity in 2030, up from 485 TWh in 2025 (IEA, April 2026). That's roughly double in five years.
A typical AI-focused data centre already uses as much electricity as 100,000 households (IEA, April 2025). Some of the ones being built now will use 20 times that.
The biggest tech companies spent more than $400 billion on capital projects in 2025 and the IEA expects that to rise by another 75% in 2026. Just five of them now spend more than the whole world invests in producing oil and gas (IEA, April 2026).
They can afford all of it, but getting it plugged in is much harder.
In rich countries a new power line can take four to eight years to build. Waiting times for transformers and cables have doubled in three years. The IEA reckons around 20% of planned data centre projects could be delayed by grid problems unless something changes (IEA, April 2025).
In the US, Wood Mackenzie estimated that the supply of power transformers would fall about 30% short of demand in 2025. About 80% of that supply is imported (Wood Mackenzie, August 2025).
The queue is bigger than the grid
The queue to connect is the number I keep coming back to. At the end of 2025 there were 2,061 GW of US power projects waiting to connect to the grid. Everything the US has actually installed comes to 1,374 GW (Lawrence Berkeley National Laboratory, June 2026).
So the waiting list is bigger than the whole system.
On the same Berkeley Lab data, projects that switched on in 2025 had waited a median of 61 months, which is more than five years.
It shows up in order books too. One maker of engines and generators ended June 2026 with a record $72.1 billion of orders waiting to be delivered. On its results call it said some power buyers were already placing orders as far out as 2030.
People ask me if this is a bubble. I can't tell you where share prices go next. But a lot of these orders are already in the system and what governments approve over the next few years is what we'll all be running on in the 2030s.
Signal tracks the queue, the order books and the stocks every week. £209 a year for the first 50, then £299.
Join as a founding member →Education, not financial advice. Capital at risk.
Why I follow this so closely
I spent years as an execution broker for small and medium-sized hedge funds, then led growth at TradingView. Alongside FINK I work with Impact Sync, which arranges investor funding for public infrastructure. So I spend a lot of time in the rooms where these projects get decided. They rarely stall because nobody needs them. It's usually funding, procurement or an argument over which department owns the decision. When they do get going, someone has to sell them the kit. Those are the companies I follow. I'd rather you knew all that up front.
Seeing pros and retail investors up close surprised me. Retail investors spend their time hunting for the perfect stock. The best pros I worked with spent theirs making sure no single bad call could wreck them.
Where people go wrong with a theme like this
Most people will miss this one because it's boring. Transformers aren't exciting and nobody's ever bragged at a barbecue about owning a switchgear company.
The people who do spot it often get it wrong. I've done that myself, in public. In April 2026 I named five grid stocks on X. They ran hard into June and by 29 July they'd fallen 31.2% from that peak (FINK calculation, Yahoo Finance data in pounds with dividends).
Three of the five kept reporting strong results and raising their guidance the whole time. So the theme was fine and the share prices had simply got ahead of it.
Anyone who piled in near the top probably sold near the bottom. That's the usual result when there's no plan.
Nobody knows how big AI's appetite for power really is. McKinsey's estimate for data centre spending to 2030 runs from $3.7 trillion to $7.9 trillion (McKinsey, April 2025). The IEA's 2030 range runs from 833 TWh to 1,008 TWh (IEA, April 2026).
If AI lands at the low end, some firms that built for the high end will find out what overcapacity feels like. The grid still needs fixing either way, but share prices won't wait around for it.
So spotting the theme is the easy part. Having a plan for a fall like the one my grid basket had is a lot harder.
Themes and stories
I try to keep a theme and a story separate in my head.
A story is usually one company with a good headline. Quite often it's a small firm with a great press release and no profits. That's usually where the headline chasers pile in late.
A theme is money that has to be spent whether any one company wins or loses. The picks and shovels tend to sit on that side.
A theme only helps you if you're still around while it plays out. With my grid basket the theme carried on and the hard part was having my size and exit set before the fall. That's the problem I rebuilt Signal to help with.
What FINK Signal is now
Signal is our paid research letter. It lands every Friday, UK time. It follows this build-out as it happens.
The report explains the bill. Each issue of Signal looks at where the money's going that week and which kinds of company are getting paid. It also flags anything that looks like it's getting ahead of itself.
It won't tell you what to buy. You'll see what we're watching and why we're watching it, plus what would make us change our minds.
There are five parts to it.
1. The weekly infrastructure update
Each week we cover what's actually happened in the build-out, from new orders and grid upgrades to the projects that got signed off or stalled. Then we look at how the stocks in the theme are behaving.
2. The long-term theme portfolio
This is a portfolio of ten or more companies we hold for the long term, all tied to the build-out. We track it in public in US dollars against the S&P 500. You'll see the scorecard every week, bad weeks included.
3. Stock in Focus
Each time we pick one company we're watching closely. We go through what it does in this theme, why we're interested and what would prove the idea wrong. It's there for you to think about and it's never a call to act.
4. Process in Focus
This alternates with Stock in Focus. It's about how investors actually approach a theme like this, so you'll see different strategies with their pros and cons.
We also cover how to think about building a portfolio around a big idea without betting the house on it. We'll give you the questions to ask yourself and leave the answers to you.
5. The momentum idea and the small-cap idea
Every week we look at a bigger company where the money's moving now. Once a month we look at a smaller company that gets far less coverage. Smaller companies are higher risk and can move hard in both directions, so the bear case is always included.
The kind of companies in the portfolio
The report sorts the build-out into types of business. That's grid kit, cables, on-site power, the firms that build it all and the materials that go into everything.
The portfolio holds companies that sell the kit rather than the ones that own the wires. I'm not naming them here, but below is the kind of business each one is.
- A company that sells the engines and generators data centres are buying because they can't wait for the grid
- The firm whose networking kit sits inside the data centres themselves
- A maker of the electrical gear that moves and manages power
- A chipmaker whose chips manage electricity in everything from grids to servers
- A European name behind a lot of the switchgear and software inside grids and data centres
- A business that puts power on site with fuel cells
- A British engineering name in power systems and engines
- One that insures and manages risk for big projects, data centres included
You'll see every name, entry price and weight in the first issue.
See every name, entry price and weight in your first issue. £209 a year for the first 50, then £299.
See the portfolio: £209/year →Education, not financial advice. Capital at risk.
After six months
Most investors I speak to are stuck in the same loop. They hear about a theme late and buy after it's already run. The first big drop shakes them out and they swear off the whole thing.
Signal is meant to help with that. By then you'll have a good idea which layer of this build-out is getting the money and which one's running ahead of itself. You'll also have watched a real portfolio go through some bad weeks.
So the next time there's a headline about a grid failure or a data centre delay, you'll know which kinds of company it touches and why.
Who it's for
It's for you if a lot of your money sits in cash or UK shares and you'd like a clearer view of this build-out than the headlines give you. It's also for you if you'd like to see how a theme behaves while it's happening.
It isn't for anyone who wants tips to copy. We show you what we hold and what we're watching, but what you do with your own money is your call.
What you get
- The Friday Signal issue. Every week, all of it behind the paywall
- The weekly infrastructure update. What's moving in the build-out and in the stocks
- The theme portfolio scorecard. Ten or more companies, against the S&P 500, in US dollars
- Stock in Focus and Process in Focus. Taking turns from one issue to the next
- The weekly momentum idea and the monthly small-cap idea. What we're watching and why
- The full archive. Every past issue
- The $106 Trillion Trade report. All 16 pages, with every number sourced
The price
Signal is £69 a month or £299 a year.
We'd much rather you took the year. A theme like this takes years to play out and monthly makes it too easy to drop out after one slow month. Twelve months at £69 comes to £828, so the year works out a lot cheaper.
The first 50 people to join get it for £209 a year. That's our founding price and you keep it for as long as you stay subscribed. Once those 50 places have gone, it's £299 a year for everyone who joins after.
There's no trial and no waitlist. You won't find a countdown timer on the page either, because the only limit is the 50 places.
Kept for as long as you stay subscribed. Once the 50 places have gone, it's £299 a year for everyone who joins after.
Join as one of the first 50 at £209 a year →Education, not financial advice. Capital at risk.
The standard price once the 50 founding places have gone.
Join for the year: £299Education, not financial advice. Capital at risk.
If you'd rather pay £69 a month, you can do that here. That's £828 over twelve months.
Join monthly: £69Education, not financial advice. Capital at risk.
Education, not financial advice. Capital at risk. Signal shows what we hold, what we watch and why. It's not a personal recommendation to buy or sell anything.
Investments can fall as well as rise and you may get back less than you put in. Past performance isn't a reliable guide to future results.
Quick answers
Who's Signal for?
What lands each Friday?
I'm a FINK Community member. Do I get Signal?
Can I cancel?
Is there a free trial?
My read
$106 trillion has to be spent and the grid can't cope as it is. The queue to plug in is more than five years long and the order books are filling up.
None of that means share prices have to go up. They can run ahead and drop more than 30% while the theme carries on. I've been through exactly that.
You don't have to take my word for any of it. Signal lets you watch the money get spent one week at a time.
There are 50 founding places at £209 a year. After that it's £299 a year.
£209 a year, kept for as long as you stay subscribed. First 50 only.
Take one of the 50 founding places →Education, not financial advice. Capital at risk.
David & Tim
P.S. The scorecard shows you what we hold, but it won't show you how we decide. That's what FINK OS is for.
Tim and I put every idea through the same five layers. We call them Filter, Enter, Size, Manage and Exit. Signal shows how we read the market and FINK OS teaches you to run the process yourself.
P.P.S. The founding price is only for the first 50 people. After that, it's £299 a year for everyone who joins.