You track every generation of GPU process nodes, compute capacity, and memory bandwidth, and you know large model parameter counts and training cluster sizes by heart. But when it comes to putting real money into compute power, you freeze — chips, optical modules, data centers, operations and maintenance... the supply chain is so long that betting on any single segment feels like a risk of misjudgment. For people who understand the technology, the hardest part is often turning a technical call into an actual position.

In September 2026, the first wave of compute power themed ETFs, including the ChiNext Compute Power ETF E Fund (158050), began listing in succession. For someone who truly understands compute, what makes this product interesting is that it simply bypasses the question of "which segment to bet on" — start by looking at the index it tracks, the ChiNext Compute Infrastructure Index. This index targets the very supply chain you know well, and its keywords are actually more familiar to you than anyone.

The first keyword is "ChiNext" — the sample comes from companies on the ChiNext board engaged in new technologies, new industries, and new business models. The second is "compute infrastructure" — note that it does not focus on a single chip, but rather the "water, electricity, and network" of the AI era, stringing together the entire chain of computing equipment, network transmission, data storage, data center support, and operations and maintenance services. In other words, it is exactly what you always say: compute power has never been just about GPUs.

The index first screens ChiNext companies that do compute, networking, storage, and operations and maintenance work, removing those ranked in the bottom 10% by average daily trading value over six months, then takes the top 50 by average daily total market value over the same period. More critically, it enforces weight caps — for segments like processors and computing equipment, any single stock is capped at 10%, while all other segments are capped at 3%. This effectively builds in "diversification" at the index level, preventing the entire index from being held hostage by any one segment. Source: National Securities Index Website.

From an industry composition perspective, it covers a combination of "underlying hardware plus infrastructure services" — including both hardware devices that carry computing and data transmission, as well as IT services, data center support, and operations and maintenance that keep systems running stably. For someone focused on the technology stack, this is like taking the entire chain from chips to racks, from networking to cooling, and folding it all in, rather than betting on a single stretch. Data source: iFind, 2026.08.31, Shenwan Level 3 industry classification.

The top ten holdings account for a combined 43.96%, close to half the index. Xiechuang Data carries a relatively higher weight, while the rest is spread across printed circuit boards, IT services, communication equipment, data center cooling, and other areas — highlighting the core strength while keeping coverage across multiple niche segments. Data source: iFind, 2026.08.31, Shenwan Level 3 industry classification.

Turning what you know into a position is exactly where the ChiNext Compute Power ETF E Fund (158050) and similar ETFs that track the ChiNext Compute Index come in. They package a batch of strong companies across the compute power track into a single product — buy one ETF, and you get simultaneous exposure to the entire compute supply chain, eliminating the need to agonize over whether to bet on chips or optical modules. Just as you would never load all your inference tasks onto a single model, your allocation does not have to be staked on just one segment either.

For someone who understands compute, the most satisfying outcome is this: the technology direction you believe in can finally become a real position. Its net value will fluctuate in the short term, much like a model still in tuning phase, so do not rush to conclusions. The key is to not let all that technical judgment in your head stay stuck at the step of "want to invest but do not know what to buy."