Overnight storms over Phoenix disturbed the power feed to a data center called PHX-01 on August 13, and its chillers dropped out. Three providers traced their outage to that building: Namecheap, Liquid Web with its Nexcess brand, and phoenixNAP. A fourth, hosting.com, reported the same cause at its own Phoenix site and dated the onset to the same 7:30 a.m. hour, telling customers the fault “eventually caused a shutdown of all equipment in the building.”

The company that runs PHX-01, RadiusDC, operates no public status page and posted nothing about the incident on its own site. Its one detailed account reached the public only because a tenant republished it.

Key facts (checked August 17, 2026; times are Phoenix local)

  • Cause: RadiusDC’s customer letter cites “elevated white space temperatures caused by multiple utility bump events attributable to overnight storms in the area.” Liquid Web told its own customers only of “a major chiller equipment failure at the facility.”
  • The building: RadiusDC PHX-01, listed by RadiusDC at 151,000 square feet and 10.5 MW of utility capacity. RadiusDC agreed in March 2026 to buy it from phoenixNAP, which kept about 80 percent of its global business and stayed on as a tenant.
  • Recovery: phoenixNAP put its services back at 6:23 p.m. on August 13 and closed its incident at 11:10 p.m. Liquid Web’s Cloud Sites platform stayed down until about 7 p.m. on August 14, roughly 37 hours after Liquid Web opened its incident.
  • Still open: a phoenixNAP object storage incident in Phoenix, opened August 14, was unresolved at the time of writing. No reason-for-outage report has been published by any party.
  • A fourth provider: hosting.com logged the same cause at its own Phoenix site the same morning, but names neither the facility nor its operator, so the link to PHX-01 is our reading.

A Storm, Then the Chillers

How the morning of August 13 ran, in Phoenix local time:

  • 3:28 a.m. phoenixNAP posts about “higher ambient temperature” and adds that “neither one service is critical but some alarms may occur during this.” The incident would run nearly 20 hours and end up flagged critical.
  • 5:53 a.m. Liquid Web opens its incident citing only “network issues affecting servers in our Phoenix Data Center.”
  • 6:21 a.m. phoenixNAP’s own support portal goes down, so customers cannot open tickets. A phone number appears in its updates two hours later.
  • 7:30 a.m. RadiusDC writes to its own customers, four hours after phoenixNAP first noticed the heat.
  • 7:45 a.m. Chiller D is back in operation.
  • 7:51 a.m. Liquid Web names “a major chiller equipment failure,” more than four hours after phoenixNAP’s first post.
  • 9:19 a.m. Liquid Web’s engineers begin “gracefully shutting down impacted systems where necessary to help protect customer data.”
  • 9:58 a.m. phoenixNAP republishes RadiusDC’s letter in full.

That letter is the most precise account of the failure anyone has published. It names the cause as utility bumps from overnight storms, reports Chiller D restored, and lists what was still to come, including temporary floor cooling and an additional chiller and backup generator later in the day. It promises updates every 20 minutes and is signed “Sincerely, RadiusDC.” It exists in public only because phoenixNAP republished it on its own status page.

Namecheap reached the shutdown decision from the other side of the contract. In an August 14 blog post, chief executive Hillan Klein named the cause as “a cooling system failure at the RadiusDC: Phoenix data center,” adding that “RadiusDC instructed us to take services offline.”

Hosting, EasyWP, Private Email and some billing operations went down, while DNS management failed even though existing zones kept resolving. The company says incoming mail was not “automatically lost” because sending servers retry, which is a hedge rather than a guarantee. Its support helpdesk could not assist customers through live chat or email either, after phoenixNAP’s own support portal had gone down hours earlier.

An AWS Region Has Been Offline for Months, and the Cloud’s Address Is No Longer a Technicality

The Chain Between the Customer and the Chiller

What the coverage has largely missed is how many companies sat between the customer and the failed equipment. Namecheap names RadiusDC as its provider. Liquid Web names no operator anywhere in its updates, but pointed customers to phoenixNAP’s incident page, calling it “the data center provider’s status page.” phoenixNAP in turn calls RadiusDC “our Phoenix node site provider” and places its own affected services in “the RadiusDC PHX-01 (Phoenix, AZ) Data Center.”

That arrangement is new, and the seam shows in Namecheap’s own account. Klein referred during the outage to “our PhoenixNAP data center” before the company settled on RadiusDC, and later appended a correction naming RadiusDC as its operator. The correction tracks a change of landlord.

RadiusDC announced on March 12 that it had agreed to acquire phoenixNAP’s Phoenix data center and colocation business. The seller kept its Bare Metal Cloud and network platforms and stayed on as a tenant in the building it used to run. Neither company has announced that the deal formally closed.

Five months on, a hosting customer with a fault in the cooling plant can be three parties away from anyone able to fix it. The party at the end of that chain talks to tenants rather than to the public.

The Facility Recovered First

The timings show where the recovery time actually went. phoenixNAP posted at 7:02 p.m. on August 13 that its services were back, and half an hour later put the restoration at 6:23 p.m.; it closed the incident at 11:10 p.m. Liquid Web was still restoring through August 14 and did not declare all services restored until the afternoon of August 15.

Its Cloud Sites platform came back about 7 p.m. on August 14, roughly 37 hours after Liquid Web opened its incidents that morning. Cloud Sites was tracked on a separate status site on a different domain. It never appeared among the affected components on Liquid Web’s main incident page, and was named in the update text only on the afternoon of August 14.

Restoring cooling, in other words, was the short part. Bringing servers, networks, storage and databases back in the right order, checking what survived and confirming stability took considerably longer, and that is the part customers experienced. Namecheap documents the same tail: after EasyWP’s site and dashboard were back, some customer sites still returned database connection errors.

One of the Largest Planned Data Centers Died on a Notice Technicality

No Postmortem, No Numbers, No Credits

Four days on, the record still has holes in it. RadiusDC has published nothing on its own channels, and its only account of the failure is the letter its tenant reproduced. phoenixNAP has said it will circulate a reason-for-outage report once it receives a formal one from RadiusDC, and none has appeared. Liquid Web has published no postmortem and never quantified the impact.

Namecheap has explained the most. Klein promised a thorough post-mortem covering where the company’s safeguards and protocols failed, and the blog says Namecheap will “implement even more redundancy” across data centers that it notes also include Europe and Asia.

The questions a hosting customer would want settled are all still open. No party has said whether any hardware was lost or any data destroyed, beyond repeated references to the risk. None has published a full accounting of affected sites or accounts. None has addressed service credits.

About the Data

The timeline, the cause, RadiusDC’s letter and the tenancy relationship come from phoenixNAP’s incident record, read through its status API and rechecked on August 17, 2026. Liquid Web’s timings and the Cloud Sites split come from its own two status pages. hosting.com’s building-wide shutdown comes from its status page, which names neither the facility nor its operator. All time conversions to Phoenix local are ours. Namecheap’s account comes from its August 14 blog post, and Klein’s remarks during the outage through TechRadar and Engadget. Facility figures come from RadiusDC’s Phoenix location page, and the acquisition terms from its March 12 announcement, which describes an agreement rather than a completed sale. We found no RadiusDC statement on its own channels.