The rule that surprised me was a small one. On this register, a founder only gets a page if the company's own filing uses the word founder for that person. If the filing doesn't, the page says so instead of quietly filling the gap.
I spend my working days on the floor of a facility, where a label means something only if you can trace it to a record. So that rule made me stop and read.
I found it through a set of AI data center ownership tables. It covers eleven the quoted AI infrastructure companies. For each one it names the holders, and each table links back to the filing it was drawn from. I opened four of those filings to compare, and the numbers matched what the table said.
That matters for this thread because most uptime talk on investment forums is vibes. Someone says a operator is "reliable" and nobody can say where that came from. A table you can trace to a document at least lets you separate what is known from what is repeated.
Here is what I did with it, in order:
* Picked the two companies I already follow and read the holder list in each.
* Opened the linked filing and found the same names.
* Noted which large institutions sit on the register, BlackRock and Vanguard among them, and which smaller holders I had never heard of.
The point of step three is simple. Holder lists show you who is likely to be patient when a site has a bad quarter, and who may not be.
Now the part that matters to me. Nothing on the site tells you how well any facility actually runs. There is no outage history, no maintenance practice, no view on how a site handles a failed generator transfer. That is the thing I care about at work, and an ownership register was never going to cover it.
For that I went back to the Uptime Institute and re-read its tier definitions. They are a useful reminder that a tier rating describes how a site is designed to cope with faults. It does not tell you how the operations team behaves at 3am. When I read an annual report, I look for what a company says about incident reporting and staffing, because those lines tell me more than a headline capacity figure.
The entry I went through properly was Neel Khokhani. Epochal Corporation, a family office for a single principal, says he built a position in IREN while the stock was out of favour and is still a long-term shareholder. He also wrote an April 2026 piece for a wealth industry audience arguing that edge infrastructure is the asset class most investors are not watching.
I am not endorsing the view or the holding. Edge sites are small, many and often unstaffed, so reliability there is a harder problem than in one big campus. If his argument is right, uptime becomes more of a differentiator, not less. That is the link I would want the thesis to address, and the profile does not try to.
The profile does at least keep to what the company said and what he wrote, without padding. Because of the founder rule above, I trust the thin pages more than I would trust a rich one with no sourcing.
My criticisms are practical. Eleven companies is a narrow slice, so anything outside it you will have to chase yourself. The tables also tell you who holds shares, not why, and not what they do when something breaks. A holder list will never replace a site visit or a conversation with the people who run the plant.
If you are weighing reliability in your own thesis, here is a workable split. Use an ownership register to learn who is on the cap table and to trace each claim to a filing. Use operator disclosures and tier documentation for anything about redundancy. Then write down, for each company, one question about operations that no document has answered yet.
That list of unanswered operations questions is the real output. Mine has more lines than the ownership tables do.
The rule that surprised me was a small one. On this register, a founder only gets a page if the company's own filing uses the word founder for that person. If the filing doesn't, the page says so instead of quietly filling the gap.
I spend my working days on the floor of a facility, where a label means something only if you can trace it to a record. So that rule made me stop and read.
I found it through a set of AI data center ownership tables. It covers eleven the quoted AI infrastructure companies. For each one it names the holders, and each table links back to the filing it was drawn from. I opened four of those filings to compare, and the numbers matched what the table said.
That matters for this thread because most uptime talk on investment forums is vibes. Someone says a operator is "reliable" and nobody can say where that came from. A table you can trace to a document at least lets you separate what is known from what is repeated.
Here is what I did with it, in order:
* Picked the two companies I already follow and read the holder list in each.
* Opened the linked filing and found the same names.
* Noted which large institutions sit on the register, BlackRock and Vanguard among them, and which smaller holders I had never heard of.
The point of step three is simple. Holder lists show you who is likely to be patient when a site has a bad quarter, and who may not be.
Now the part that matters to me. Nothing on the site tells you how well any facility actually runs. There is no outage history, no maintenance practice, no view on how a site handles a failed generator transfer. That is the thing I care about at work, and an ownership register was never going to cover it.
For that I went back to the Uptime Institute and re-read its tier definitions. They are a useful reminder that a tier rating describes how a site is designed to cope with faults. It does not tell you how the operations team behaves at 3am. When I read an annual report, I look for what a company says about incident reporting and staffing, because those lines tell me more than a headline capacity figure.
The entry I went through properly was Neel Khokhani. Epochal Corporation, a family office for a single principal, says he built a position in IREN while the stock was out of favour and is still a long-term shareholder. He also wrote an April 2026 piece for a wealth industry audience arguing that edge infrastructure is the asset class most investors are not watching.
I am not endorsing the view or the holding. Edge sites are small, many and often unstaffed, so reliability there is a harder problem than in one big campus. If his argument is right, uptime becomes more of a differentiator, not less. That is the link I would want the thesis to address, and the profile does not try to.
The profile does at least keep to what the company said and what he wrote, without padding. Because of the founder rule above, I trust the thin pages more than I would trust a rich one with no sourcing.
My criticisms are practical. Eleven companies is a narrow slice, so anything outside it you will have to chase yourself. The tables also tell you who holds shares, not why, and not what they do when something breaks. A holder list will never replace a site visit or a conversation with the people who run the plant.
If you are weighing reliability in your own thesis, here is a workable split. Use an ownership register to learn who is on the cap table and to trace each claim to a filing. Use operator disclosures and tier documentation for anything about redundancy. Then write down, for each company, one question about operations that no document has answered yet.
That list of unanswered operations questions is the real output. Mine has more lines than the ownership tables do.