Annual Letter 2026 · LightSpeed Holdings

London · 17 October 2026

This is the fourth letter I have written to you, and the first one I have enjoyed writing. The group made $412 million of net asset value this year against $387 million last year, and for the first time none of that increase came from something we bought. It came from things we already owned, getting better.

What follows is eleven minutes of reading. There are no appendices. If you only have two minutes, read the last chapter, which is the only one that asks you for anything.

I have tried to write the parts we got wrong as plainly as the parts we got right. There are three of each.

A word on how to read the numbers. We mark the book once a year, at the close of the financial year, and we mark it the same way whether the year has been kind or unkind. Three of the thirteen positions are held at less than we paid for them. Two of those three are held at less than they were worth twelve months ago. None of them has been quietly moved into a different line to make the total read better.

The businesses themselves had an ordinary year, which after four extraordinary ones is the outcome I would have chosen. Northbeam grew slower than it promised and faster than it did in 2024. Kestrel spent the year fixing a supply chain it should not have broken. Vela did the thing we bought it to do, eighteen months later than planned. None of that is dramatic. Most of what compounds is not.

What has changed is the shape of the work. Four years ago the group was a place where things were bought. It is now a place where things are run. That is a slower business and a duller one to describe at a dinner party, and it is the reason this letter has fewer numbers in it than the last three.

One

The year, honestly

We began 2026 with three operating companies, two buildings and a habit of describing ourselves as a technology group. By June it was clear that description was flattering. Technology is 31% of the book. Food is 17%. Healthcare is 15%. We are a holding company that owns a good software business, and pretending otherwise cost us a year of thinking clearly.

Naming that changed three decisions. We stopped looking at software acquisitions we could not have won. We put money into the Kestrel packaging line instead, which returns less and returns it sooner. And we hired a second seller in Europe rather than a third engineer in Tokyo.

A holding company that cannot describe itself in one sentence is a holding company that will buy the wrong thing.

The lesson of the first half

The second half was quieter and better. Bookings finished at $4.82 million for the third quarter, the best the group has had. Win rate went up two points while average deal size went down three per cent, which is exactly the trade we wanted: more customers, smaller commitments, faster decisions.

Two

Three businesses, three different years

It is tempting to write about the group as one thing. It is not one thing. Each business had a different year and each one needs something different from us next year.

Harbour Freight had a boring year Revenue up four per cent, margin flat, no incidents. Boring is what we bought. The temptation is to go looking for growth here. We are going to resist it and take the dividend instead.

Kestrel Foods had a hard year and won it Wheat moved against them for seven months. They held price, lost two customers, and finished with margin down only one point. That is the single best piece of operating work anyone in the group did this year, and almost nobody outside Kestrel noticed.

Vela Health had a year we cannot yet judge Bookings up twenty-five per cent on a small base, and a clinical result due in the spring that makes the number either irrelevant or the most important thing in this letter. We have extended the bridge rather than price the round in the dark.

Quarry Road, Newark · the only building we bought this year

We also bought one building. Quarry Road is eighteen thousand square metres of logistics space that Harbour Freight was already renting from someone else. Paying rent to ourselves is not a strategy, but it is arithmetic, and the arithmetic was good.

Three

What we got wrong

Three things, in order of how much they cost.

We were slow on Corvus. Footfall has been soft for four consecutive months and we have been waiting for a turnaround plan since June. Waiting was a decision and we made it by not making it. The plan now arrives at the November board, five months late, and the mark is down 3.4% for the year.

We over-concentrated in North America. Forty per cent of new bookings came from one region and two accounts. Every individual decision that produced that was correct. The sum of them was not.

We under-invested in the thing that is working. Tenpoint Labs is up thirty-one per cent and we own less of it than we should. That is not a mistake we can fix by wanting to.

Every individual decision was correct. The sum of them was not. That is the hardest kind of mistake to see from inside.

On concentration

Four

How we work now

A smaller thing, but the one people ask about most. Every document this group produces is now a single file that opens in a browser and needs nothing else — no login, no subscription, no network. The board pack, the daily brief, the sales figures, the sign on the meeting room door. All of it is the same handful of colours and the same typeface, and all of it can be read on a plane.

Including the doors The panels outside the meeting rooms run the same file format as this letter. It cost nothing to make them match, and the first time a visitor said the building looked like the company, I understood why it mattered.

I do not think this is a competitive advantage. I think it is hygiene, and I mention it only because we spent four years without it and did not know what it was costing.

Five

What we are asking you for

Three things, and only three.

Approve the Halvorsen renewal at the revised terms. Three years, $24.0 million, with the service credit clause reopened. Legal thinks conceding it is worth doing and I agree. It is our largest single contract and our most fragile.

Extend the Vela bridge by ninety days. Four million dollars, maturing in eleven. Converting now means pricing the round before the clinical result, which is a bet neither of us has enough information to make.

Fund a second seller in Europe. A hundred and eighty thousand a year. This is the smallest number in the letter and the one I care most about, because it is the only one that fixes the concentration problem in chapter three rather than describing it.

The next year will be less interesting than this one, and that is the plan.

In closing

We have no acquisitions in the pipeline. We have one clinical result, one contract renewal, one turnaround, and a hiring plan of two people. If I write to you next October and the most interesting thing I have to report is that Kestrel held margin again, we will have had a very good year.