One AI infrastructure bet or another layer of funds

My index holdings are boring enough that I barely think about them. The AI infrastructure part is different. I keep talking myself into one company, then wondering whether I just want a more interesting portfolio rather than a better one. An ETF feels less exposed, but I still want to know what risks it spreads out and what risks stay together.   I started with ownership rather than another forecast about demand. I opened who invests in AI infrastructure to see whose names appeared around AI cloud and data center names that file with the SEC. It puts ownership information drawn from company filings and public announcements into a register. That gave me somewhere to start checking my assumptions, not a reason to copy anyone. The useful distinction for me was between finding an owner and understanding their decision. You can see that someone chose exposure without knowing what else they own, why they chose it or what would make them leave. I wrote that distinction at the top of my notes because a familiar institutional name can make me feel reassured before I have learned much. The layout also separates out large private transactions on another page. I found that useful for keeping the stock market part of the story in perspective. A listed infrastructure ETF is not the whole financing picture. But those private deals do not tell me whether a particular listed share offers good value, so I still needed company filings for the business questions. One detail slowed me down: the site's explanation of beneficial ownership. Its count can include shares someone controls through voting or disposal rights, along with shares they have the right to obtain within 60 days. I checked how beneficial ownership is defined for background before treating an ownership figure as shares already held. That distinction can make the reported position look larger than the current holding. For my own comparison, that means ownership figures need a note beside them, not just a number copied into a spreadsheet. I want to separate current shares from rights to receive shares where the source allows it. Otherwise, I could end up comparing two positions that were counted on different terms. I also looked more closely at the page on Epochal founder Neel Khokhani because it kept the discussion to his AI infrastructure investing and expressly stopped short of a full life account. It notes his writing about access to power becoming the limiting factor for which infrastructure assets can expand. That gave me a specific business question to examine, rather than just another ownership name. If power is the bottleneck, owning several companies does not automatically mean owning several unrelated risks. That is where my ETF comparison gets harder. I would want to check the underlying holdings for shared dependencies, not stop at the number of names. Equally, one company with a plausible power story still leaves me exposed to its financing needs and execution. My main criticism of the register is also its boundary: it shows who chose concentration, not whether concentration was wise. I found it useful for forming questions, but thin as an answer to this thread's actual choice. Ownership evidence cannot settle valuation or tell me how much uncertainty I can sit through. I am still tempted by the single position. What changed is the test I would put it through before assigning it any place beside my index holdings. My comparison now has three separate lines: what the ownership figure counts, what business risks overlap, and how much of my portfolio one company could put at risk.