Denver paused new data centers inside city limits until May 2027, and it was always a network market.

Why Denver is a market at all

Denver is a route market before it is a compute market. The transcontinental fiber that links Chicago and Dallas to the West Coast crosses the Front Range here, and for practical purposes it crosses inside one building. 910 15th Street, the Denver Gas and Electric Building, opened in 1910 and is now the primary interconnection point for the Rocky Mountain region. CoreSite bought it in April 2025 and described it as the number one network-dense carrier hotel serving Denver and the Rocky Mountain region.

That single asset sets the shape of the whole metro. Four buildings inside Denver city limits carry the clear majority of the metro's network connections, and one of them carries close to half on its own. A concentration like that is ordinary in Ashburn or Chicago. In a market this small it is the defining fact, and it tells you what Denver actually sells: reach, not raw capacity.

The buyers who have paid for that reach are regional. Energy and mining companies, aerospace and defense along the Front Range, federal agencies, telecom, and the Denver offices of national firms. Lumen grew out of Level 3, founded in Broomfield. Demand here has historically been enterprise scale, measured in cabinets and hundreds of kilowatts rather than in halls.

AI has changed less about that than the headlines suggest. CBRE's H1 2025 profile of the market made the point flatly: only two colocation facilities in Denver offer large amounts of power and space. One genuinely large campus is being built, in Aurora, and it is aimed at single large tenants, not at a buyer taking twelve cabinets.

Who is actually plugged in across Denver

A building existing and a building being connected are different facts, and only one of them decides whether your traffic can get where it needs to go cheaply. These figures come from PeeringDB, which the network operators themselves maintain, so every row below links to the building's own public entry.

The most connected buildings here

From PeeringDB, which is maintained by the network operators present in these buildings rather than by the landlords, which is what makes it worth more than a directory. Two limits. Entries are voluntary, so a building showing no networks may be single tenant or may simply never have registered. And a network being present says nothing about whether the building has space, power or a price for you: that is still a call.

What has been announced in Denver

Announcements are not filings. Nobody has to build what they announce, the figure is the one the developer chose to publish, and the date moves. They are here because megawatts are what a requirement is measured in, and because the size of what is coming changes who will take your call. Every figure below was read on the page it links to.

Announced capacity is collected by hand from trade reporting and company statements, and each figure links to the report it came from. It is kept apart from the permit record above and never added to it: one is a document a jurisdiction issued, the other is a plan a company published. Neither is availability.

Announced capacity is collected by hand from trade reporting and company statements, and every figure links to the report it came from and was read in that report before it was written down. It is never added to the interconnection count above: one is a plan a company published, the other is a building a network registered itself into. Neither is availability.

Two utilities, one unfinished tariff, and thin air

The first question to ask about any Denver building is which utility serves it, because the metro is split and the two halves are not variations on a theme. Xcel Energy serves Denver, Aurora, Boulder, Lakewood, Thornton and most of the population. CORE Electric Cooperative, a member-owned co-op, serves much of Douglas County including Parker and Castle Rock. Different cost structures, different queues, different regulators. Flexential's newest Denver-area building is on CORE, not Xcel, and that was not an accident.

On the Xcel side the rules for large loads are being written right now and are not finished. The Colorado PUC adopted guiding principles in Decision C25-0747 in November 2025. Xcel filed the resulting large load tariff on 2 April 2026 in Proceeding 26AL-0137E. Its Schedule TL would apply to customers with 50 MW or more of new load, with loads between 20 and 50 MW potentially caught as well. The terms are heavy: minimum 15 year contracts, a payment floor of 80 percent of contracted power, roughly $600,000 in deposits and study costs before construction starts, and an exit fee equal to the sum of the remaining minimum monthly bills. As of August 2026 the commission had not ruled, so the price of large-load power in Denver is not yet a knowable number.

The queue behind that tariff is far less solid than headline totals suggest. The Colorado Sun reported that in one six month period across 2024 and 2025, seven potential data center customers representing 4,000 MW withdrew their service requests from Xcel, while more than a dozen new prospects with 3,500 MW filed. Load in a utility queue is an option, not a commitment. Apply that churn rate to any capacity a broker describes as coming.

Then there is the altitude, which cuts both ways and is almost never priced into a proposal. Denver sits at roughly 5,280 feet, where air density is about 14 percent below sea level. The same fan moving the same volume of air carries meaningfully less mass, so it removes less heat, and air-cooled equipment has to be specified for the elevation rather than off a sea level datasheet. Diesel gensets lose output too, on the order of 3.5 percent per 1,000 feet on naturally aspirated sets. None of this is a reason to avoid Denver. It is a reason to ask whether a quoted kW per cabinet figure was derated for the site, and whether the N+1 generator math still holds at 5,280 feet.

Where the market actually is, submarket by submarket

When Denver is the right answer, and when it is not

Denver earns its place when reach matters more than megawatts. A requirement of a few cabinets up to a couple of megawatts that needs many carriers, a cloud on-ramp and a position roughly midway between the coasts is well served here, and served better than the market's size would imply. It is also a sound second site for a primary in Dallas, Phoenix or the Bay Area, because it shares almost no correlated risk with any of them. Setting aside hail and wildfire smoke, the hazard profile is mild: no hurricanes, no meaningful seismic exposure, and most of the metro sits outside a serious flood plain.

It is the wrong market if you need scale on a date. The state has exactly one facility above 100 MW, it is still under construction, and its operator will not publish a capacity figure. Inside Denver proper, no new project can even begin the zoning process until May 2027. CBRE counted two colocation facilities in the market offering large amounts of power and space, which means a 5 MW requirement produces a shortlist of two or three rather than ten. If you need competitive tension to get a good price, you will not find much of it here.

It is also the wrong market if the business case leans on incentives. Colorado has no data center sales and use tax exemption. HB 26-1030 would have created a 100 percent exemption for up to 20 years and it failed in the 2026 session, alongside SB 26-102, which would have imposed renewable energy matching and water reporting requirements. Both died. Neighboring states that do abate sales tax will beat Colorado on total cost for an equipment-heavy deployment, and no amount of pressure on a colocation provider closes a gap that sits in the tax code.

One geographic warning that catches people out. Boulder and Louisville have real research, aerospace and life sciences demand but almost no multi-tenant colocation of their own. Buyers up there generally end up backhauling to downtown Denver or the Tech Center, so budget for the circuit and the latency, not just the cabinet.