If You Own the Cluster, You Own the Patent Risk
Why AI Infrastructure Operators Are the New End-User Defendants
Highlights
- Artificial intelligence (AI) infrastructure operators increasingly face patent infringement claims because they assemble and operate systems that combine hardware, networking, cooling and software technologies.
- Supplier indemnities may not cover operator-configured systems, making strong defense, cooperation, technical support and continued-use provisions important in supplier agreements.
- Operators can reduce patent risk by establishing a centralized patent assertion process, coordinating supplier response and reviewing public technical materials that could support infringement claims.
Companies that operate artificial intelligence (AI) infrastructure face patent infringement claims. Many operators have assumed that the chip, server, cooling or networking supplier would be sued first. That assumption no longer holds. Plaintiffs now often name the companies that own and run the clusters instead of the component suppliers. The logic is simple: An operator buys accelerators, servers, cooling systems, networking equipment and software tools, installs them in its own data halls, configures the scheduling and routing and sells capacity by the hour. To a plaintiff, the operator is the user of the deployed system. It is also the party with the revenue worth pursuing.
This Holland & Knight alert explains why operators have become targets as end users, as well as describes the three main risks and the steps operators can take now to reduce them.
Why Operators Make Attractive Targets
Patent liability does not stop with the company that builds a product. Anyone who makes, uses, sells, offers to sell or imports a patented invention without authorization can be liable. An operator does not have to manufacture a single component to be sued.
For operators, the key word is "use." Many patents claim a complete system: processors, memory, networking, cooling or power controls and software that schedules and manages work. The company that assembles that system and runs it in production is a natural defendant.
Damages follow the same logic. A patent owner will base its claim on the operator's use of the accused technology, and cluster revenue gives it a large number to point to. The plaintiff is not entitled to all of that revenue. When a system has many components, damages must reflect the value of the patented feature. Even so, the total figure creates settlement pressure and draws the board's attention early.
Notice also matters. A complaint counts as notice, and so can a pre-suit letter if it gives enough detail, and it may start important deadlines. A patent email sent to a sales inbox, business development alias, technical contact or shared mailbox can create real urgency long before anyone files suit.
Risk 1: You Are the Party Practicing the Claim
A plaintiff will usually find the operator attractive for two reasons. The first is the infringement story. A supplier may sell one component, while the operator combines chips, memory, cooling, networking, software and facility infrastructure into one running system. If the asserted claim reads on the combined system or on the way the system operates, the operator is the party the plaintiff will say practices the invention.
The second reason is money. The supplier's relevant revenue may be tied to a component sale. The operator's relevant business may be tied to capacity, uptime, utilization and compute hours. Plaintiffs know that distinction, and they will use it to make the case look bigger.
Operators should not concede the plaintiff's framing. The first response should separate system claims from method claims, identify each asserted limitation, determine who performs each step or controls each component, and test whether the accused configuration actually existed in production during the relevant period. A plaintiff's theory may depend on marketing language, architecture diagrams or assumptions about how the system runs. Those assumptions should be tested immediately.
Mitigation Steps
Start by treating every patent communication as a potential deadline. Do not ignore it, but do not assume every letter is legally sufficient notice either. Appoint one owner in legal responsible for patent assertion intake. Require each communication to be classified by patent, technology layer, accused product, claim type, supplier involvement, requested response date and preservation risk.
Require a short privileged claim-mapping memo. The memo should identify missing claim elements, noninfringement positions, invalidity themes, potential patent-exhaustion defenses, possible indemnity and evidence that may be overwritten if not preserved quickly.
Next, send every patent communication to a single legal-controlled intake address and audit that process quarterly against registered agent mail, sales channels, investor-relations contacts, procurement contacts and public-facing technical aliases. Demand letters that place the organization on notice can arrive through various channels, some of which are not monitored.
Finally, form a patent risk committee with members from legal, procurement, engineering, facilities and finance departments. It should meet monthly and maintain notes regarding decisions to design around, tender to suppliers, seek a license, challenge validity, preserve evidence, respond or take no action.
Risk 2: Your Configuration Choices Sit Outside the Supplier Indemnity Unless You Control the Supplier Response
Supplier indemnities usually cover the product as delivered. They often do not cover the system you build by combining accelerators from one vendor, fabric from another, cooling equipment from another and your own scheduler, topology and offload settings. The claims most dangerous to operators frequently read on that combination, which means the dispute can fall between contracts unless the operator has already built a strategy for controlling the supplier response.
Do not treat supplier indemnity as a passive reimbursement right; treat it as an operational control issue. The goal is not only to ask whether a supplier must pay. The goal is to determine, before a complaint arrives, whether the supplier must defend, as well as who 1) controls the defense, 2) selects counsel, 3) can settle, 4) must preserve technical evidence, 5) must provide witnesses, 6) must support invalidity challenges and 7) must keep the deployed system running while the dispute is pending.
5 Supplier-Control Problems
- Combination and Modification Exclusions. The supplier declines to defend because you used its product with another vendor's product or configured it yourself. That describes nearly everything an AI infrastructure operator does.
- Reimbursement Instead of Defense. You choose and pay for counsel, then argue later about repayment. That leaves the operator controlling the emergency but carrying the cash burden.
- No Technical Cooperation or Obligation. The supplier has the engineers, design documents, firmware history, prior art knowledge and product road map information, but the contract does not require meaningful litigation support.
- Settlement Rights That Threaten Operations. A supplier may want to settle by removing, replacing or limiting use of a product that the operator needs to keep clusters online.
- Missed Notice and Preservation Obligations. Notice windows are short, and technical evidence may sit with the supplier rather than the operator. If the supplier is not engaged immediately, critical evidence can disappear before the defense is organized.
Mitigation Steps
- Build a Supplier Control Strategy Before the First Demand Letter Arrives. Create a supplier control map for each major hardware, cooling, power, networking, optics and software supplier. The map should identify what is covered, what is excluded, who controls the defense, whether the supplier has a duty to defend or only reimburse, what settlement consent rights exist, whether the supplier must support technical discovery, whether continued use is guaranteed, and whether the supplier must provide an equivalent replacement if the accused product becomes unavailable.
- Make Supplier Cooperation a Provision in Every Purchase Contract. The contract should require prompt defense, technical support, document preservation, engineer access, litigation cooperation, support for invalidity or noninfringement positions and a continued-use remedy that protects deployed infrastructure.
- Prepare a Supplier Escalation Packet in Advance. When a letter or complaint arrives, the legal department should be able to send the supplier a package that includes the assertion, accused products, purchase documents, indemnity provisions, preservation instructions and the operator's proposed defense path. The faster the supplier is pulled into the merits, the harder it is for the supplier to stay on the sidelines.
- Treat Supplier Strategy as Part of Your Defense. The supplier may have the best noninfringement evidence, invalidity evidence, design-around option and commercial incentive to stop a campaign before it spreads to other customers. The operator's job is to turn that supplier knowledge into a coordinated defense, not just a reimbursement demand.
- Do Not Stop at Indemnity. Look for instances where a license covers downstream uses. Ask whether the component was sold under a license, cross-license or settlement that may support exhaustion, implied license or other pass-through defenses. Record who was authorized to sell, as well as whether downstream combinations were contemplated, any sale was outside the license scope and the accused component substantially embodies the asserted technology.
Risk 3: You Wrote the Plaintiff's Evidence
To sue an operator, a plaintiff must show that the accused system runs in production. Your own public material often helps them. Marketing pages, architecture white papers, investor presentations, conference talks, engineering blogs and online videos may describe how your systems work. Engineering blogs and videos can be especially useful to a plaintiff because they are specific, dated and attributable to the company.
Patent counsel, not only communications, should review public technical material twice a year. The goal is not to strip useful technical marketing. The goal is to avoid publishing the plaintiff's claim chart and damages story.
Finally, keep a ready-to-send engineering hold notice that the legal department can issue within 24 hours of a credible assertion. It should cover configuration management records, deployment manifests, firmware versions, coolant temperature logs and utilization telemetry. Telemetry is often overwritten on short cycles. By the time outside counsel is retained, records showing what actually ran may already be gone.
Before the Letter Arrives
Here are four steps to take in the next 90 days to mitigate patent risk.
- Name one owner for patent assertion intake.
- Publish the triage deadlines.
- Complete the indemnity and supplier control map for your top five compute, cooling, power, networking and software suppliers.
- Build the supplier escalation packet before you need it.
If you have any questions about patent risks related to AI infrastructure, please contact the author.
Information contained in this alert is for the general education and knowledge of our readers. It is not designed to be, and should not be used as, the sole source of information when analyzing and resolving a legal problem, and it should not be substituted for legal advice, which relies on a specific factual analysis. Moreover, the laws of each jurisdiction are different and are constantly changing. This information is not intended to create, and receipt of it does not constitute, an attorney-client relationship. If you have specific questions regarding a particular fact situation, we urge you to consult the authors of this publication, your Holland & Knight representative or other competent legal counsel.