The whole trick was building before the demand arrived.

He spent twelve years turning a six million pound company into a FTSE 250 platform by building capacity nobody had ordered yet. He now argues that the nerve which built it would not work today, because the thing that used to be scarce no longer is.

Michael Tobin CBE
Michael Tobin CBE
Founder, Tobin Ventures · formerly chief executive, TelecityGroup
Questions put 12
Resolved into data 06
His framing score 88/ 100
Attribution Named, by request

Ask what changed and he starts at the wall socket.

Most answers to that question start with AI. His starts somewhere less fashionable, and it is the more useful answer.

Michael Tobin CBE spent twelve years building RedBus and then TelecityGroup from a six million pound market capitalisation into a FTSE 250 company worth more than two and a half billion pounds. Asked to score the distance between the era he operated in and the one he now advises on, he puts it at 88 out of 100 toward a different game entirely. The reason is not the workload. It is that the binding constraint has moved.

That matters because his entire operating method depended on where the constraint sat. Building ahead of the curve was a decision an operator could make alone, funded by conviction and a balance sheet. Building ahead of a grid queue is not. The confidence that built TelecityGroup would today run into a connection date set by somebody else.

He answered twelve questions for this report. Six of them resolve into something the desk can put a number against, and those are set out below as instruments, each paired with an independently sourced figure that Entelligencia has put to the same question. The evidence is ours, not his. He was not shown it before he answered. The remaining answers are judgement rather than measurement, and they are printed exactly as they arrived.

Entelligencia desk · The A Roll · edition 02 of a continuing series

Twelve years, one nerve question.

He has one method and he has run it twice. Identify who will want to own this asset in five years, then build the thing that buyer will need, and buy the capacity to build it before anyone has ordered it. What follows is not a career. It is that method put to five tests, in the order he met them, ending with the one it has not yet passed.

Eight frames
The European colocation market, before the wave

The method starts by ignoring what the market wants now.

Before hyperscale, the European colocation business had the opposite problem to the one it has now. Power was procurable. Space was procurable. What nobody could be sure of was whether anyone would take the racks.

Every judgement in this edition traces back to that asymmetry, and to what happened when it inverted.

Test one: can you run it when there is nothing to hide behind?

He takes on a business with a six million pound market capitalisation. At that size there is no cushion: capacity bought early is a bet the company can be killed by, and capacity bought late is a customer lost to whoever moved first.

GBP 6mMarket capitalisation at the start
12yrsFrom that seat to the exit

Test two: are you selling the building, or what connects to it?

The combination turns a set of buildings into a platform: one interconnection story, one balance sheet, one listed currency to buy with. That is the moment the business stops being priced as property and starts being priced as infrastructure.

It is also the moment the method becomes repeatable, which is what makes the next decade legible rather than lucky.

Announced · from the contributor’s submission

Test three: will you buy the capacity before anyone orders it?

Asked to name the inflection point of that period, he does not name an acquisition or a listing. He names a discipline, and it is commercial rather than technical.

“Catching the wave of demand and being in the position to service it required the confidence to invest in new capacity upfront. Noone wants to wait a year for capacity.” Michael Tobin CBE · reproduced exactly as submitted

What the method compounds to when it holds.

Twelve years of building ahead of the order book, compounded through a listing, produced a FTSE 250 company worth more than two and a half billion pounds. The curve alongside is the shape of that decision repeated. It is not a share-price series.

400xOrder of magnitude, start to exit
FTSE 250Index position reached
Announced · endpoints as stated, curve indicative

Test four: does it work twice, on a different asset?

TelecityGroup is sold. Then Itconic, the Iberian platform, delivers a second return on the same logic: identify who will want to own this in five years, then build the asset that buyer will need.

Then again, on another continent. With Permira he bought Teraco in South Africa for around $100m in 2015 and sold it for about $1bn in 2019. Same logic, a market with none of Europe’s connectivity density, and roughly ten times the money in four years.

Three exits on one method is what separates a good operator from a transferable thesis, and the thesis is what he now sells.

Contributed · Teraco figures as stated by the principal

Test five: can you still see it once you are not running it?

As an investor and adviser he sits above hyperscale, enterprise colocation, inference and the construction industry that serves all three at once. He is precise about what that buys: patterns a single operator cannot see.

A dedicated chief executive, in his account, is structurally blind to some of them. Not through any failing. The job supplies exactly one vantage point, and he now has several.

And the test the method now has to pass.

Hyperscale demand dwarfed every previous curve, and in doing so moved the bottleneck somewhere an operator cannot simply choose to be brave about. Conviction and a balance sheet no longer buy a connection date.

“Demand is no longer the constraint to growth. Power is.” Michael Tobin CBE
The Instruments · six of twelve answers

Six answers that resolve into a number.

Six answers that resolve into a number.

Each tile carries one of his points as the heading. Open it and you get the answer in his own words, then an independently sourced figure the desk has put to the same question. The evidence is ours, not his. He was not shown it before he answered, so where the two agree it means something, and where they part it means more.

Instrument file · A02-I
Six of twelve answers
Graded against the record
Grades apply to the paired evidence, not to the contributor’s opinion. An opinion is not a claim about the world until a figure is put beside it, and the figure is what carries the grade. Full taxonomy in the methodology.

The interview roll. A Q and A with Michael.

These are the answers no chart improves. They are printed as a question and an answer, in the order they were asked. Where the submission is quoted it is reproduced exactly, including the typing; where the desk has summarised, it says so and does not use quotation marks.

Transcript · A02-V
Written submission, July 2026
Unedited within quotation marks
Answered on a scale of 0 to 100
0 · The same game, bigger100 · A different game entirely

88 out of 100. The number is the answer, and it is the quietly radical thing in this submission. A practitioner with twelve years of operating scar tissue is saying that most of what he learned is now context rather than method. The instrument for this answer sits above, paired with the hyperscaler capex series that puts a scale on what he is describing.

Contributed · his own scoring
Reproduced exactly as submitted
“hyper scale demand has dwarfed all previous demand curves in DC capacity and has brought new challenges around the availability of power in all regions. Demand is no longer the constraint to growth. Power is.”

Four sentences, and the last two are the thesis of this entire report stated by someone with no reason to flatter it. Building ahead of the curve was a decision an operator could make alone. Building ahead of a grid queue is not.

Reproduced exactly as submitted
“Catching the wave of demand and being in the position to service it required the confidence to invest in new capacity upfront. Noone wants to wait a year for capacity.”

He was asked for a moment and answered with a discipline. Note what the second sentence assumes: that a year was the worst case. In the markets this report tracks, the equivalent wait is now measured in grid cycles rather than build cycles, which is precisely why he no longer thinks the method transfers unmodified.

Summarised by the desk from the submission

He traces demand through four distinct phases, each with a different buyer and a different requirement, and notes that today’s buyers are generally sophisticated enough to understand the differences between them. Set out in sequence, it is a useful corrective to the habit of treating the customer as one thing.

Phase 01

Telco colocation

High-connectivity space, bought for the carriers and the cross-connects rather than the compute.

Phase 02

Cloud, on and off premise

The enterprise migration. Scale and standardisation start to matter more than adjacency.

Phase 03

Hyperscale training

Large language models. Density, power and speed to energisation displace every previous criterion.

Phase 04

Enterprise inference and edge

Regional, latency-bound and closer to the user. A different buyer again, and the one he is watching.

Read against the report’s own market chapters, phase four is the one that most changes where capacity should sit, and it is the least represented in announced pipelines.

Summarised by the desk from the submission

He does not name the grid, the turbine queue or the transformer shortage. He names the public. The risk he identifies is NIMBYism, and specifically the prospect of populations refusing to accept small modular nuclear reactors sited in their neighbourhoods to serve data centres. Without that acceptance, in his read, the power to service human demand for AI consumption does not exist.

Since the submission it has stopped being abstract. In July 2026 residents of West Cumbria attended a consultation on Pioneer Park, an SMR-powered AI data centre sited next to Sellafield, where Radiation Free Lakeland campaigners contested the clean-energy framing and raised the River Ehen. It is the exact objection he described, on a real site, with a date. It is the most contrarian answer in the submission and the only one no amount of capital solves. It reframes the build as a question of consent rather than engineering, which is the conclusion this report reaches independently in the Unrest Tracker and in Losing the Room, from an entirely different direction.

Contested · a forward risk, not a recorded one
Summarised by the desk from the submission

For an industry that talks in gigawatts and global platforms, the answer is strikingly local. Having people in region and in country is paramount, he says, and the reason is not headcount. They need to understand local dynamics, work with local planners and utility providers, and carry what he calls local empathy.

On investability he refuses a single ranking. It differs region by region, driven variously by demographic growth, data sovereignty, strategic location, energy costs, fibre and subsea landing stations, or latency. As he puts it, every region is different. That is an inconvenient answer for anyone selling a global playbook, and it is the reason this report is organised as ten market chapters rather than one index.

Where to go next.

Strata · Capital Capital His six points read against the financing question: who pays for the build, and how. Read the chapter → The A Roll · Edition 03 Jérôme Totel The next edition. He says power is now the constraint; Totel runs a platform that buys it years before there is a customer to charge for it. Read the edition → The series The A Roll Every edition, the format, and who is sitting for the next one. Open the series →

What is his, and what is ours.

In conversation with Michael Tobin CBE, founder of Tobin Ventures and formerly chief executive of TelecityGroup. Contributed to The Next Hotspot via the Entelligencia briefing survey and attributed by name at the contributor’s request.

His positions are his own. The surrounding framing, the paired evidence, the grading and the editorial selection are Entelligencia’s. He was not shown the evidence in The Instruments before he answered, and he has not been asked to endorse it.

Quoted passages are reproduced from his written submission without alteration, including the typing. Where the desk has summarised rather than quoted, the answer is marked as summarised and no quotation marks are used. The career arc uses figures stated in the submission. The value curve is no longer indicative: it is plotted from six dated points in the public record, listed with their sources in The Instruments.

Imagery in The Arc is atmospheric rather than documentary. It illustrates the market and the period under discussion; it does not depict the contributor’s own sites.

ConductedSummer 2026, in writing
EditionA02 · The A Roll
ContributorMichael Tobin CBE
SeatFounder, Tobin Ventures
SubmissionWritten, July 2026
AttributionNamed, by request
PortraitPermission granted
Contributor sign-offPending
MethodHow claims are graded →