Before I get going today, I want to mention that something big is happening this week…

On Thursday, 20 August, at 4pm GMT.

It’s something I’m personally really proud of.

Proud because I’ve spent the last few years working away in the background on something I had originally intended to keep to myself.

After a lot of discussion with my publisher and the team here at Southbank, we came to the same conclusion:

This was simply too useful to keep to myself.

And thanks to the extraordinary advances we’ve seen in AI this year, I’ve just put the finishing touches to it…

This is the algorithm that is at the core of what I call, HYPERION.

Hyperion is a custom AI engine I’ve built from the ground up. It’s been in planning, development, and construction for around three years now.

I’ve taken my 30+ years of knowledge and experience in the markets, more than a decade of research into AI, and the most advanced frontier models available today to build it.

I consider it the most advanced investment tool Southbank Research has ever had access to.

It gives me institutional-grade decision-making power to help identify some of the most explosive stocks across the US and UK markets.

I won’t go into much more detail today, because on Thursday at 4pm I’m going to show you exactly how it works.

I’ll demo HYPERION live, show you what it can do and explain how you can use it.

Make sure you register for the reminders here , then tune in Thursday at 4pm. Just note that clicking the link automatically adds you. You won’t need to fill in any of your details.

Now onto today’s essay…

I’m always the first to be sceptical when I see a strange-looking chart.

The one below is no exception…

Then I looked a little closer… and realised it’s not all that wild after all.

In fact, it’s showing something I’ve known about for a while. But seeing it mapped out like this by a16z really puts into perspective just how significant the opportunity in “neocloud” companies could be over the next 12 to 24 months.

That chart shows the acceleration in quarterly revenues from “neoclouds.” What’s fascinating about two of the three you see here, Applied Digital and CoreWeave, is that they used to be cryptocurrency miners.

The sold out Nvidia GPUs

In the early 2010s, AI “compute” in the mainstream vocabulary wasn’t even a thing.

But just because AI compute didn’t really exist like it does today, that doesn’t mean there wasn’t an insatiable appetite for Nvidia GPUs.

Today, Nvidia GPUs are the backbone of modern AI compute. A decade or so ago, they were also the backbone of bitcoin mining.

Nvidia GPUs ended up being really really good at two things: powering the most advanced video games, and solving Bitcoin’s proof-of-work algorithm to mine bitcoin.

Of course, not every GPU had the same power or hashrate. But the high-end GPUs became so popular with bitcoin miners that demand far exceeded supply and prices exploded.

Eventually, some savvy entrepreneurs realised there was a bigger opportunity.

Raise a heap of cash, fill a warehouse with computing hardware, and point the whole thing at Bitcoin.

Server rack upon server rack. Thousands of machines running around the clock, processing algorithms, and chasing bitcoin rewards.

And out of this “digital gold rush” came companies like Riot Blockchain and HIVE Blockchain.

After the dot-com bubble burst, what happened to thousands of miles of fibre-optic cable laid during the boom?
A. It was dug up and recycled
B. Governments bought most of it
C. Much of it sat unused waiting for demand to catch up
D. It was converted into electricity cables


Find out what your friends and family think as well. Share this email with them so they can answer today’s poll and subscribe as well.

It turned out to be so profitable that we ended up with dozens of Bitcoin – and eventually other crypto – mining companies. Applied Digital (which was Applied Blockchain), Hut 8, Cipher Mining, Bit Digital, CoreWeave… all bitcoin and crypto miners.

They were called “miners.” But practically speaking, they were large compute centres, with bitcoin the preferred target.

Until something else began to pay more handsomely…

Artificial Intelligence.

What was very quickly realised by all of these players is that the demand for their compute was no longer coming solely from Bitcoin. Frontier AI companies that needed immense amounts of compute to train and to allow their AI models to “think” were increasingly a source of demand as well.

And they were willing to pay more for it.

So, tell me this…

If you’re a business that’s getting £100 from one thing, but there’s another thing that’s willing to pay you £1,500 for that exact same thing – a 15x increase in revenue – what are you going to do?

Yup! You’re going to pivot to the thing that’s going to pay you more.

And that thing is AI compute.

Earnings blow the roof off

Then last week, earnings landed for several of these neoclouds, and they were outstanding!

CoreWeave (Nasdaq: CRWV) reported on 11 August. Revenue hit $2.58 billion for the quarter, up 112% on a year ago.

The contracted backlog is now US$104 billion. And that doesn’t include the US$25 billion of new commitments it signed in the first few weeks of this quarter.

Nebius (Nasdaq: NBIS) followed the next day (although it’s not a former bitcoin miner) with a 454% lift in revenue to US$582 million. It has US$8 billion in cash and is cash flow positive to the tune of US$2.3 billion.

It also quadrupled its total contract value quarter on quarter.

Applied Digital (Nasdaq: APLD) had reported a little ahead of both of those, but it’s the same story. Revenue ballooned up 407% to US$258 million.

It also signed two 15-year take-or-pay leases with a new US based, high investment-grade hyperscaler for 300 megawatts at its Delta Forge 1 and Delta Forge 3 AI factories.

Another one you’ve probably not heard a lot about, WhiteFiber (Nasdaq: WYFI), also saw a healthy jump in  earnings. The company was carved out of bitcoin miner Bit Digital (Nasdaq: BTBT) last year. It lifted revenue 54% to US$28.8 million.

It saw a 43% jump in cloud services revenues and a 173% jump in colocation revenue with an almost 60% margin..

Riot Platforms (Nasdaq: RIOT) did US$174.2 million, up 14%, of which only US$23.2 million came from data centres. The rest is still bitcoin mining. But… that’s because RIOT is a little later to the AI compute party.

The bigger deal in its earnings release was that Riot said it had signed a 20-year lease covering 191 megawatts at its Rockdale campus in Texas, running to June 2048, with a world leading frontier AI lab.

It didn’t say who. Bloomberg did, naming Anthropic as the tenant.

It’s a massive deal, worth nearly as much as RIOT’s market cap. And if it is Anthropic, it’s wild to think it’s paying up big for a bitcoin miner turning into an AI factory provider.

To me, these emerging neoclouds are shaping up as one of the best trades of 2026 and into 2027.

The most scarce resource in the world, believe it or not, is still Nvidia GPUs.

But this time around, it’s not because of Bitcoin. It’s because of compute.

And the companies that have the most Nvidia GPUs in the world are all former bitcoin miners (at least most of them).

When something is in such short supply, with such incredible demand, that translates into prices heading higher.

There will be risks, and big bumps along the way, but I’m absolutely looking very closely at opportunities here… with my HYPERION engine fired up and ready to go!

Don’t forget about Thursday at 4pm.

Until next time,


Sam Volkering
Investment Director, Southbank Investment Research

PS The neocloud opportunity is exactly why I built HYPERION. There are simply too many stocks, too much data and too much happening beneath the surface for me to catch everything on my own. On Thursday at 4pm GMT, you’ll watch HYPERION scan the market live, rank the opportunities it finds and single out its number one stock. Grab your free viewing pass to Ignition Day here.  (Again, when you click this link, you’re automatically agreeing to receive emails about Hyperion.)