Valtrus Innovations v. Data-Center Operators: US 6,854,287 and 6,868,682

From scattered complaints to one courtroom

Valtrus is a patent-licensing entity holding a large portfolio acquired from Hewlett Packard Enterprise. Since April 2026 it has sued colocation, interconnection and network operators in nine federal districts, asserting a hardware patent and a control-software patent that, read together, are directed at how modern data halls meter and manage cooling. Several of the same complaints also assert other HPE-origin data-center patents, and the JPML transfer order refers to a third patent, ’277, in some related actions.

The first group, five complaints in the Northern District of Illinois against CoreSite, Netrality Properties, Lumen Technologies, NTT Global Data Centers Americas and Prime Data Centers, was reported on 9 April 2026. Further waves followed in New Jersey, Louisiana, California, Texas, Pennsylvania and Massachusetts. Valtrus and Key Patent Innovations themselves moved to centralise on 28 April, and on 7 August the JPML created MDL No. 3190.

The structure is uniform: single-defendant or single-family complaints, filed in waves, with the same two patents asserted against the same category of equipment. The 365 Data Centers and Cogent complaints each name several affiliates. The same short exhibit list recurs, the patent and an infringement claim chart for each asserted patent, which is consistent with a templated filing model.

The sample is 12 dockets across six districts: E.D. Tex., N.D. Tex., S.D. Tex., D. Mass., W.D. Pa. and E.D. Va. Caption, court and docket status were confirmed for 7 of the 12 (00669, 00670, 02755, 02758, 02760, 13811 and 00382). The other five (00667, 00678, 01734, 02641 and 06818) are placed by sequence or listing without a verified caption.

Home-district filings, a Marshall endpoint

The campaign has not settled on one forum. April went to the Northern District of Illinois and the District of New Jersey, with one in the Eastern District of Texas (STACK Infrastructure). The May complaint against Lumen went to the Western District of Louisiana, and the August complaints went to the Northern District of Texas, the Western District of Pennsylvania, the District of Massachusetts and elsewhere. Much of that is defendant-adjacent venue, the operators’ home or facility districts, but the endpoint is Judge Gilstrap’s courtroom in Marshall. Cases filed elsewhere have moved in by conditional transfer order since mid-August.

Most sampled dockets are at the pleading or transfer stage. No dispositive ruling on the ’287 or ’682 patents was identified at the latest docket checks. The pressure on Valtrus is coming from elsewhere.

Matching cooling effort to measured demand

Both patents address one problem. A data center’s cooling plant usually runs at or near worst-case output whatever the actual heat load, and conventional units sense temperature at the air-conditioning unit rather than at the racks. The ’682 specification puts numbers on it: cooling can draw roughly half the power the servers themselves use, and compressors alone as much as 30 percent. The patents claim a way to match cooling effort to measured, location-specific demand. The ’287 does it with hardware that meters cooling fluid to individual units; the ’682 does it with software that decides, level by level, how much to deliver.

Valtrus asserts both because a modern enterprise cooling deployment can present both layers at once: physical variable-flow cooling equipment, which is the target of the ’287 theory, and automated cooling-control and monitoring software, the target of the ’682 theory. Each patent has its own independent claims, so each defendant faces two separate sets of infringement and validity questions, and liability could attach under one patent and not the other. Because the complaints rest on the operator’s use of third-party equipment, the evidence largely sits with the supplier, which is why Vertiv, rather than any operator, has been the most active respondent.

The operator is asked to answer for the equipment that moves the cooling fluid and for the logic that tells it how much to move.

One hardware layer, one control layer

Before these patents, the usual approach was central air-conditioning units whose compressors ran near maximum power regardless of load, sized so cooling fluid reached the equipment at around 100 percent of the estimated requirement. What distinguishes the portfolio is where the decision is made. Rather than fixing cooling output when the room is designed, the system senses temperature at distributed locations and adjusts at run time, by metering fluid to individual units (’287) and by letting a hierarchy of agents decide when to adjust locally and when to escalate (’682).

’287: individually metered cooling-fluid flow. Several heat-exchanger units are supplied with cooling fluid from an air-conditioning unit. Each unit takes in room air, cools it by heat transfer to the fluid and returns it through a fan. In claim 1, the controlling step individually manipulates the mass flow rate of cooling fluid supplied to each unit. Dependent claims meter that flow with valves or pumps on separate fluid lines, and claim 10 recites a controller operating a plurality of pumps to set each unit’s flow independently. A rack drawing little power receives little fluid, and the plant’s own output can drop to match.

’682: hierarchical agent-based control. A rack agent compares its sensor readings with its objectives and opens or closes vent tiles to change coolant flow; if it cannot meet its objective it asks the row agent for help. The row agent aggregates rack data, adjusts flow across its row, shunts cooling from neighbouring rows where possible and, failing that, calls on the cooling-unit agent, which adjusts fan speed and then compressor capacity and can ask other units for help. In claim language, a first agent adjusts the delivery rate of cooling fluid and requests a second agent when it cannot keep the temperature within range, and a third agent varies cooling output when the first two cannot.

Cooling and energy control in neighbouring fields

The review covers the twelve months before each patent’s earliest priority date, using a supplied set of 24 records. Seven are Hewlett-Packard cooling filings that share ownership with, or are family members of, the asserted patents, and are treated as part of the portfolio rather than outside references. Fifteen fall inside the ’287 window, four (the Robertshaw family) also fall inside the ’682 window, and two fall outside both. Only a handful of in-window records touch thermal or energy control at all; the rest concern vehicles, battery chargers and portable devices, and the review rests on supplied abstracts and claims.

The in-window references attack control of cooling and energy in adjacent fields, at different levels: the processor, the vehicle coolant loop, the on-board load network and the utility-to-device link. None addresses a room of computer systems cooled by multiple units whose individual fluid flows are metered, and none arranges agents in a rack-row-CRAC hierarchy that escalates when local control fails.

The five Aker Wade battery-charger records in the window are unrelated to thermal control of computing equipment and are not discussed. What the references show is that sensor-driven control of cooling and energy was active in neighbouring fields at the time, and that the contested ground lies in the specifics: individually metered fluid flow to multiple room-level units (’287) and an escalating agent hierarchy over rack, row and cooling unit (’682).

Where the claim fights will be

For the ’287, the central phrase is individually manipulating a mass flow rate of cooling fluid supplied to each of the plurality of heat-exchanger units. Vertiv’s declaratory complaint treats that phrase as the heart of its non-infringement position, asking in substance what evidence shows individual, per-unit flow control rather than collective or plant-level adjustment.

For the ’682, the likely contest is whether an accused control system arranges rack-, row- and plant-level logic in a hierarchy in which a lower agent requests help from a higher one when it cannot hold its range, or simply runs one centralised loop.

It also explains why Valtrus can assert two patents against one deployment. They address different layers of the same stack, flow hardware and control logic, so each can be pointed at a different feature of the same cooling installation, with separate non-infringement and validity questions for each.

A bounded exercise with a closing window

This is a mature, template-driven effort. A first group of complaints in April, another wave on 18 August and a move to centralise before a single judge point to coordination and staying power. Its shape is also bounded: both patents have expired, so the exercise is one of damages for past infringement, and Valtrus has resolved or dropped earlier targets (Equinix, Digital Realty).

Time works against the plaintiff in two ways. The six-year look-back in 35 U.S.C. § 286 reaches back only to roughly April 2020 for the April 2026 filings. If the ’682 expired on 16 January 2023, as Google Patents indicates, that leaves a recoverable window of about 33 months for that patent, and the ’287 window is similarly bounded by its own expiry. An operator that began using a given system after expiry has no exposure on that system, so the window differs by patent and by defendant and is worth quantifying with counsel.

Second, because the accused equipment is supplied by third parties, the events most likely to move the campaign are the supplier-led actions, Vertiv’s declaratory suits and the status of its PTAB petitions, and the first claim-construction rulings in MDL No. 3190 on the individual-flow and agent-hierarchy limitations. Any operator, equipment supplier, component vendor, investor or acquirer in or near this market should treat active or threatened assertion of this portfolio as a live exposure, and should expect further waves for as long as the look-back remains valuable.