TL;DR · 30-second read
The Short Version
Giant computer warehouses built for artificial intelligence are drawing political pushback in the US, and proposed limits often apply only above a certain size. Equinix, which runs such buildings for large businesses, says its sites usually fall below that line.
The line sits around 100 megawatts, roughly the electricity a small city uses. A typical Equinix site uses about a third to just over half of that.
The catch: these limits are still being debated. Being small only protects Equinix if the line stays where it is.
CNBC reported on November 6 that Equinix sees its smaller data centers as less exposed to potential US construction restrictions. Asia-Pacific president Cyrus Adaggra said its facilities typically range from 30 to 60 megawatts (MW), below the roughly 100 MW level that Adaggra said US proposals generally target. Adaggra also said capital spending has risen from about $2 billion a few years ago to $7 billion.
Equinix’s third-quarter earnings release, filed with an 8-K on October 28, reported record annualized gross bookings of $533 million, up 35% from a year earlier, and full-year 2026 capital expenditure guidance of $5 billion to $6 billion.
Executive Summary
Equinix is positioning its footprint as a partial answer to a growing political risk. As scrutiny of large US data center projects increases after the midterm elections, the company argues that its typical site is too small to be caught by rules aimed at about 100 MW and above.
The argument matters because the pressure is concentrated elsewhere. BMI, a Fitch Solutions company, said about 53% of projected growth in the US hyperscaler pipeline sits in states where newly elected governors could impose restrictions. Hyperscalers are the giant cloud and AI operators whose campuses dominate that pipeline. Equinix, by contrast, attributes its growth largely to large enterprises.
The size advantage is real on the numbers Equinix gives, but it depends on rules that are not yet settled. Demand looks well supported in Equinix’s filed results. The $7 billion spending figure is harder to place, because it exceeds the company’s own 2026 guidance.
A 100 MW Line Is a Proposal, Not a Moat
The arithmetic behind Equinix’s claim is simple. A 30 to 60 MW site has 40 to 70 MW of headroom beneath a 100 MW threshold. Adaggra put it this way on “Squawk Box Asia”: “We will invest in markets where we see customer demand, and because we have a footprint that is typically less than 100 megawatts, we tend not to fall and be caught by the legislation or the regulation, which is typically at about 100 megawatts in the U.S.” The words “typically” and “tend” are doing real work. They describe a usual outcome, and they do not rule out exposure.
Three things could narrow that headroom. Lawmakers could set a lower threshold, since the restrictions under discussion remain potential and Adaggra’s figure describes a general pattern. Rules could also measure capacity by campus or operator instead of by building, in which case several mid-sized facilities might be counted together. Finally, power use per site may rise as equipment gets denser. Equinix’s new Enhanced Performance Innovation Centers are liquid cooling environments for high-density AI infrastructure, a sign of where some customer demand is heading, though the company has not linked them to larger site sizes.
The people affected are Equinix’s planners and customers. Enterprises choosing a location for AI inference, the work of running already-trained models, may value a site that avoids a contested approval process. That benefit lasts only if the threshold and its measurement method survive the legislative process. The claim is best read as a reasonable description of current proposals, and not yet as a durable regulatory exemption.
The Bookings Support the Demand Story, but the $7 Billion Needs Context
Equinix’s filed results give the demand argument solid footing. Annualized gross bookings, a measure of the yearly value of newly signed business, reached a record $533 million in the third quarter, up 35%. Revenue rose 11% to $2.571 billion, interconnection revenue grew 10%, and revenue from Fabric, its software-defined connectivity service, rose 24%. The company also plans to deliver nearly 40% more cabinets, the racks customers rent for equipment, in 2026 than it first anticipated.
The spending figure is less clear. Adaggra cited $7 billion, while Equinix’s October 28 release guides to total 2026 capital expenditures of $5 billion to $6 billion. That gap is at least $1 billion. The higher number may reflect a different period or scope, but no such definition accompanied it. Even the guided range is substantial, at roughly half or more of expected full-year revenue of $10.325 billion to $10.365 billion.
Against the backdrop of AI spending jitters, the mix of demand matters as much as the total. Adaggra attributed growth to large enterprises investing in hybrid and multi-cloud infrastructure and connectivity, and said Equinix serves about 60% of the Fortune 500. Chief executive Adaire Fox-Martin described “broad-based demand.” If that holds, bookings should be less tied to a few AI tenants, although the headline figure does not separate AI from other workloads.
Staying Under the Threshold Does Not Remove Indirect Exposure
BMI’s 53% figure measures where restrictions could arrive, not where they have been enacted. BMI also said any impact would likely become more apparent from 2028, because most projects planned for 2027 are already committed. So Equinix’s size argument may not face a real test for some time, and neither may the broader fears about the pipeline.
Russell Investments flagged a second-order effect: restrictions could raise costs across the AI infrastructure supply chain, while slower US construction could reduce demand for chips and other equipment, including from suppliers across Asia. A company outside a rule’s direct reach still buys from that supply chain. How those forces would net out for a smaller operator is an open question, since lower equipment demand and higher costs pull in different directions.
There is also a possible upside that remains unproven. If large projects are delayed, some customers might look to smaller sites for capacity. Nothing Equinix has reported shows that shift happening, and its stated growth drivers are enterprise connectivity and multi-cloud needs, not displaced hyperscale demand.
Asia Offers Growth, but Not an Escape From Constraints
Adaggra’s remarks came alongside growing attention to Asia. Sjoerd Leenart, CEO of JPMorgan’s banking business in Asia Pacific, said most data center activity after the US is expected to happen there, as hyperscalers look for land, electricity and grid capacity. Mirae Asset Securities said South Korea could become a key market, favoring operators that secure power, land and long-term customer contracts.
The same constraints travel with the investment. Risk adviser WTW cited power availability, grid connectivity and equipment delays as key challenges in Asia. Leenart also said authorities in Thailand and Australia had paused some projects to assess the effect of data center energy demand on other industries and consumers. That concern is about total load, which raises a fair question for any size-based argument: a regulator focused on aggregate demand may care less about how large each individual building is.
Background
Equinix, based in Redwood City, California, and listed on Nasdaq as EQIX, operates data centers and interconnection services used by enterprises, networks and cloud providers. It expects 2026 revenue of $10.325 billion to $10.365 billion, and it has raised full-year guidance for three consecutive quarters.
US data center development has come under closer scrutiny as the AI buildout expands, and the midterm election results have heightened concern that new restrictions could delay projects and raise construction costs. Much of the attention falls on very large projects, which is the context for Equinix’s argument about the size of its facilities. Source: Equinix points to smaller facilities as scrutiny of U.S. data center construction grows (CNBC) and Equinix bets on data center demand with $7 Billion capex push amid AI spending jitters (CNBC), on Equinix’s site sizes, capital spending and exposure to potential US restrictions. Primary sources: Equinix 10-Q filed 2026-10-28; Equinix 8-K filed 2026-10-28; Equinix Reports Third-Quarter Results and Raises 2026 Guidance (Exhibit 99.1 to 8-K filed 2026-10-28).Sources