Meta and Mercury announce deals as nuclear generation sets a record

Meta announced a deal with Apex Clean Energy on Oct. 1, and Mercury announced one with NZ Clean for 118 megawatts the same day, two corporate agreements between named parties and neither with a price in the day's data. Alongside them came a figure from the other end of the same trade: nuclear generation set a 2,702 terawatt-hour record, and ETF Trends is citing it to argue the buildout case for the NUKZ nuclear fund, putting operating capacity at 423 gigawatts-electric across nearly 440 reactors at an 83.7% average capacity factor.

Neither announcement describes something an advisor can put in a client account. A fund holds securities, and an agreement negotiated between a buyer and a seller is not a security, any more than a fleet's output record is; the two deals announced that day are agreements between the parties that struck them. That is the shape of the AI-power trade as it reaches a portfolio, with the physical and contractual layer negotiated in private and the investable layer consisting of whatever the public market has issued around it.

Take the record first, because it is the base-load case in numbers. A capacity factor of 83.7% describes a fleet running at a high fraction of what it can produce, 2,702 TWh is the volume that comes out, and 423 GWe across nearly 440 reactors is the plant behind both, all of it per ETF Trends' account. These are physical measures of performance, and they do not say who gets paid for the electricity. Neither Oct. 1 agreement's terms appear in the coverage.

ETF Trends is making the case for NUKZ on that foundation, and the case rests on what the fleet does next rather than on what it has already produced. The coverage identifies no project or timeline behind the buildout argument, and nothing connects the fleet in the record to either of the day's deals. What a fund of that kind gives an advisor is a basket of nuclear-linked equities, priced daily and exposed to the flows of the ETF market. It does not give the 2,702 TWh, the 118 megawatts Mercury contracted for, or any share of the terms Meta and Apex agreed. The client owns the equity, and the output belongs to whoever contracted for it.

That gap is not an oversight in product design. A bilateral agreement is written for the two parties who negotiate it and priced against their circumstances, which makes it a poor fit for a vehicle that creates and redeems shares daily. The listed fund is the general-purpose version of the same idea, and it is sold to anyone with a brokerage account.

One more distinction sits in the day's data and tends to vanish in a headline. The tracking labels Meta's deal with Apex Clean Energy and Mercury's with NZ Clean as announced; Globeleq's and Hut 8's carry as closed. An announcement describes a deal the parties have chosen to make public, while a closing describes something that has happened, and a trend line built out of announcements is a claim about intent. Anyone sizing the contracting wave should be clear about which of the two they are counting.

The financing side has the same structure, and it is the half already sitting in ordinary client accounts. Five AI-capex issuers borrowed more than $132 billion in the first half, an issuance pace heavy enough to push tech past banks as a weight in investment-grade bond indexes and to lift tech's share of pure corporate bond ETFs by about 300 basis points. Corporate bond indexes weight by debt outstanding, so a borrower that issues more becomes a bigger part of the benchmark, and the benchmark passes through to every fund tracking it.

The consequence is easy to miss because nobody decided anything. A client holding a broad investment-grade corporate bond ETF owns a slice of that borrowing, and the tech weight inside the fund moved by roughly 300 basis points in the first half without an order being placed in an AI fund or a policy statement being written about data centers. Set the two halves of the trade side by side and they look lopsided in an instructive way: six months of borrowing by five issuers was enough to reorder sector weights in investment-grade indexes, while the power agreements that decide whether the electricity gets delivered are struck one at a time, between two companies, in volumes measured in megawatts.

In practice the two exposures land in different sleeves and get reported as different things. A nuclear fund arrives as a satellite position — small, deliberately sized, easy to explain at a client review. A broad corporate bond fund arrives as core fixed income, held for ballast, its sector weights somebody else's decision. The position an advisor chose gets the scrutiny; the position that chose itself gets a line on a holdings report.

Put the sleeves beside each other and a familiar problem surfaces. A client who owns nuclear-linked equities and a broad corporate bond fund holds two positions that both depend on the same buildout continuing, one through the operators and one through the borrowers. Advisors spend a great deal of time on equity concentration and much less on whether the bond sleeve, bought for ballast, has quietly become a second bet on the same theme.

What the client owns there is a claim on the companies doing the borrowing. It carries credit risk, index concentration and whatever the next half's issuance brings, and it carries no operating risk or operating upside, because a lender to the buildout is not a participant in what gets built.

Which leaves the question an advisor actually has to answer: three exposures, one headline, and how much of the story a client owns. The nuclear fund runs on utilization and the buildout case. The corporate bond position runs on issuance pace and credit conditions. The agreements between Meta and Apex, and Mercury and NZ Clean, run on terms the coverage does not disclose. A portfolio holding the first two has no participation in the third.

The practical work is look-through, and it is not exotic. Ask the fund provider for holdings detail and read the sector weight rather than the label, because the label on a core bond fund says nothing about how much of the AI buildout the fund has absorbed, and the label on a thematic fund says nothing about which operators it holds. Then decide whether the exposure belongs in the portfolio on purpose. An advisor who wants AI power as a deliberate sleeve can size it; one who discovers it inside a ballast holding has taken the position by default, and default positions are the ones that get explained badly after a poor quarter.

For family offices and larger RIAs with access to private vehicles, the contracted layer may be reachable another way. The coverage does not say whether any vehicle is being raised around these two agreements, and neither deal's terms are public, so the route, if one exists, runs through relationships rather than a platform.

The rest of the day's activity points the same direction. Globeleq closed a deal, Hut 8 closed a deal, Vertiv opened an office, and Bain Capital is in talks with data-center developer Edged US, according to PWD's tracking. The day produced closings and negotiations rather than products a client can subscribe to.

There is a view worth stating plainly for anyone who spends a morning comparing fund lineups. The durable economics of AI power sit in the contracted layer, where a buyer and a seller agree to terms, and that layer carries no ticker. What the shelf offers instead is the proxy: nuclear-linked equities on one side, the issuers' bonds on the other. Both are legitimate holdings, and neither of them is the contract.

The thing to watch is whether the contracting migrates toward something a fund can hold, or whether the shelf keeps selling proxies while the agreements stay private. On Oct. 1, both deals went out with no ticker attached to either of them.