Where to spend a quantum patent budget: what the USPTO and the EPO actually grant
The USPTO and the EPO grant quantum applications at different rates and for different reasons. What the examination records of 16,896 quantum applications mean for where a limited patent budget should go.
When a quantum computing company files the same invention at the United States Patent and Trademark Office (USPTO) and the European Patent Office (EPO), usually from one specification, the two offices can reach different results, at different rates and for different reasons. For a company with a limited patent budget, that difference decides where the money does the most good: which inventions to file in Europe, which to keep to the United States, and which to protect another way. The examination records of 11,306 USPTO and 5,590 EPO quantum applications show where the two offices part ways.
The headline gap
In the core quantum computing class, the USPTO has granted 4,100 of the 4,514 applications that have reached an outcome, about 91%. The EPO has granted 898 of 1,219, about 74%. The gap is not the same across the field. For superconducting quantum devices the two offices are closer: 89% at the USPTO and 75% at the EPO. The widest gap is in algorithms.
Segment by segment
| Segment | USPTO grant rate | EPO grant rate |
|---|---|---|
| Error correction | 94% | 88% |
| Physical qubit hardware | 93% | 80% |
| Compiling and control | 91% | 74% |
| Architectures | 90% | 79% |
| Quantum key distribution | 90% | 83% |
| Superconducting devices | 89% | 75% |
| Algorithms | 86% | 62% |
Grant rate here means grants as a share of applications that have been granted, refused, withdrawn or abandoned. Pending applications are excluded. The quantum key distribution row counts applications whose claims or titles are about QKD; the patent classification group for quantum cryptography also contains quantum networking, post-quantum cryptography and unrelated filings, which are left out.
Algorithms show the widest gap, 24 points. Error correction shows the narrowest, and the EPO Register records no refusal of an error-correction application in this class. For budgeting, the table is a rough guide to return on a European filing. Error-correction and hardware filings have a good chance of becoming EPO patents. Algorithm filings succeed far less often, and the refusals we reviewed turned on whether the claims were built around the hardware.
Why the offices differ
The USPTO examines quantum software mainly on prior art. If the claimed method is new and not obvious over the references, the application is usually allowed. Eligibility objections arise, and a few examiners raise them repeatedly.
The EPO asks a different question first: whether the quantum hardware shaped the method, or whether the method would run the same way on an ideal machine. A method that would run the same anywhere is treated as mathematics, and only the technical features count toward inventive step. One EPO examining division put the test in a sentence when it refused a Google application: the claimed constraints concerned “what is computed not how it is computed.”
The same application in both offices
The clearest evidence comes from applications filed in both places. Of the 25 applications the EPO refused that also had a USPTO counterpart, the USPTO granted 19. In 19 of those 25 files the USPTO examiner never raised the question the EPO had decided.
A few files show the difference directly. Google’s method for reducing the measurements a variational algorithm needs was granted twice at the USPTO after the claims stated which processor performed each step. The EPO refused it, accepting that fewer measurements resulted but holding that the method was not adapted to the processor. Northrop Grumman’s application to simulate a superconducting logic circuit was allowed at the USPTO after it added that the simulation serves “to determine an RQL circuit to be fabricated.” The EPO refused it, because an intention to build something later is not a technical effect under its rules.
The offices do agree on hardware read against a drawing. Intel’s qubit coupler application was refused at the EPO and abandoned at the USPTO, in both cases on the same figure in an IBM paper.
How each office says no
At the USPTO, a quantum application that fails usually ends in abandonment after a final rejection. Appeals are rare but effective. Three Microsoft appeals against one examiner’s theory that quantum software cannot be enabled until fault-tolerant machines exist were all won on appeal, and each became a patent within eight months. Across the 34 quantum appellate decisions of the USPTO’s Patent Trial and Appeal Board, every reversal of the examiner became a patent and 15 of the 18 decisions upholding the examiner ended in abandonment.
At the EPO, formal refusals are uncommon. In the core quantum computing class, 15 applications have been refused and 306 withdrawn, 249 of them before any examination report. Withdrawals after examination often follow a summons to a hearing, and in ten of the lapsed files we reviewed the examining division had already indicated an allowable claim. None of the 31 substantive quantum refusals has been carried through an appeal to an appellate decision, so the examining divisions’ practice is the EPO’s working law on quantum software. Each of these outcomes has a budget lesson. A European application that is withdrawn after examination has used most of its cost already. One that lapses with an allowable claim on the table has used it for nothing. At the USPTO, an appeal can be the better spend than a fresh application or an abandonment.
Time and effort
An application drafted for the EPO’s question can go through quickly: 83% of Google’s EPO grants came after one examination report or none. When it is not drafted that way, the process is long: the refused algorithm applications typically went through a hearing, with a median of about 41 months in examination. A hearing adds preparation, attendance and often new claim sets, so the same refusal costs more in Europe than an early decision not to file there would have.
One specification, two offices
A USPTO grant on a quantum algorithm says little about the EPO outcome, because the USPTO usually did not ask the EPO’s question. A specification written for both offices from the start can say which hardware limit the method is built around, what the output is used to do, and which claimed steps run on the quantum processor. Those details cost little at drafting and are hard to add once the application is pending. For the budget as a whole, the same records point to a simple allocation: European filings focused on hardware, error correction, and claims tied to the processor; pure algorithm filings at the USPTO unless European claims can be built around the hardware; and contract and trade-secret protection for material neither office will cover, such as simulation outputs.
Source: CipherLaw; based on the USPTO’s records for 11,306 quantum applications, the EPO’s for 5,590, the 34 quantum appellate decisions of the USPTO’s Patent Trial and Appeal Board, and the 31 EPO refusals with the USPTO file of each.