Three electrical equipment makers reported the June quarter as if they were reporting the same trade. Schneider Electric's North American sales rose 23% and carried the company to record global revenue. Vertiv raised its full-year 2026 net sales growth guidance to 31%, with operating profit up 44% and adjusted operating profit up 51%. Orders at Eaton's Electrical Americas unit rose 41%, and the electrical backlog across the business rose 43%. All three quarters ended 30 June 2026, and all three are downstream of the same forecast: McKinsey expects global data center investment to reach $7 trillion by 2030.
The forecast stack behind those results is worth laying out in order of how much of it is already committed. S&P Global puts the combined capital spending of Amazon, Microsoft, Alphabet, Meta, Oracle and SpaceX above $1.3 trillion by 2027 — near-term, concentrated in six balance sheets, and the least speculative number in the set. McKinsey's $7 trillion runs three years further and covers the whole market. NEMA's contribution is the constraint rather than the opportunity: data centers are expected to account for 38% of net growth in US electricity consumption through 2037. Access to power, not to capital, is the thing that now rations expansion.
The money is landing on suppliers and on the reseller channel behind them. Insight Enterprises reported consolidated net income up 65%; Connection, another solutions provider, reported net income up 33.8%. That is the tell that this is a buildout rather than a chip cycle — margin is showing up in the companies that procure, integrate and deploy, not only in the ones that manufacture.
Schneider Electric sits directly in the bottleneck, selling power distribution, switchgear and cooling — the equipment required to run high-density compute at all. Steve Carlini, the company's chief spokesperson for data centers and AI, framed the shift in scale plainly: a decade ago data centers were far smaller, and large sites today need their own medium-voltage switchgear. His estimate is that installed global data center capacity reaches roughly 200 GW by 2030, about double today's 100 GW.
Set that beside McKinsey, which expects capacity to nearly triple by 2030 as the largest cloud operators and other technology companies build out AI infrastructure. The vendor selling into the boom is forecasting roughly two-thirds of what the consultancy forecasts. That inversion is unusual enough to be worth holding onto.
The second source of demand is not new buildings but denser ones. Carlini expects rack power density to nearly double every year, and cites Nvidia's infrastructure roadmap as the example: rack draw rising from about 227 kW to 400 kW as new generations of AI systems ship. Past Rubin Ultra, he says, the industry will have to move to 800 V high-voltage direct current distribution. The knock-on effects run through the whole facility. Schneider has already deployed more than 4 GW of liquid cooling capacity, a figure that should climb quickly as Blackwell Ultras and Vera Rubins are liquid-cooled by design.
Those two density numbers do not quite sit together. Nearly doubling every year is the sales line; 227 kW to 400 kW across multiple hardware generations is a 76% step, and a slower curve than the annual doubling implies. Which of the two a utility or a developer plans around changes the capital plan materially, and only one of them is anchored to a published roadmap.
The load-bearing claim in all of this is Carlini's last one, and it is the one that deserves the most scrutiny. He says Schneider's data from large internet companies worldwide rests on customer commitments rather than long-range projections — that these firms are handing over purchase orders, not forecasts, and that there is no speculation about who will occupy the sites. If that holds, the $7 trillion and the $1.3 trillion stop being sell-side arithmetic and become a backlog. Eaton's 43% backlog growth is the closest thing here to independent corroboration, and backlogs, unlike guidance, are hard to revise upward on enthusiasm.
What none of these numbers describe is supply. Every figure on offer — 200 GW, 400 kW racks, 38% of consumption growth, 4 GW of liquid cooling — measures demand or the equipment that serves it. Nobody in this account quantifies the generation being added to meet it. The suppliers benefit either way: switchgear and cooling get sold against a signed site whether or not the power behind it arrives on schedule.
Which is the tension buyers are actually underwriting. Schneider, Vertiv and Eaton sell into the constraint rather than around it, and the purchase orders Carlini is counting commit a customer to a building, not to the electricity that runs it.