What each U.S. maintenance window costs in 2026, a six-question rubric to score every patent in an hour, and why to sell before you let one lapse.

At the first two U.S. maintenance windows you are buying four more years of a patent, and at the third you are buying the rest of its term, so the question is what you plan to do with those years. Our view: run every patent through a short rubric before the window opens, keep the ones that cover a product, read on a competitor’s product, or can be licensed, and put the rest up for sale a year before you would let them lapse. Abandonment should be the last option, not the default.
The USPTO charges three maintenance fees on a utility patent, due at 3.5, 7.5, and 11.5 years after it issues. At undiscounted rates on the current USPTO fee schedule they are $2,150, $4,040, and $8,280, so $14,470 over the life of one patent. Small entities pay 40 percent of that ($860, $1,616, and $3,312) and micro entities pay 20 percent ($430, $808, and $1,656).
Each fee can be paid without surcharge during a six-month window, years 3 to 3.5, 7 to 7.5, and 11 to 11.5 after issue, per the USPTO’s maintenance page. Miss the window and you get a six-month grace period at the same fee plus a $540 surcharge ($216 small, $108 micro). Miss the grace period and the patent expires. It can be revived by petition if the entire delay was unintentional, with a petition fee of $2,260 in the first year and $3,000 after that, and MPEP 2590 says a petition filed more than two years after expiration needs an additional explanation of the circumstances. Talk to your patent attorney before relying on any of that. A lapse is cheap to avoid and expensive to undo.

One entity-status note, because it bites here. The USPTO’s entity status page says to re-check status every time a fee is paid, and small entity status is gone once you have licensed or assigned the patent to a company that does not qualify. Redbrick IP’s 2026 guide describes the penalty for paying the small fee after you stop qualifying as no less than three times the underpayment; ask your attorney to confirm the current rule. The licensing deal that closed last year can change the fee you owe this year.
Dennis Crouch at Patently-O pulled the USPTO’s maintenance records in March 2026 and found that only about 40 percent of patentees now pay all three fees, down from a peak of roughly 52 percent around 2012 and 2013, so roughly 60 percent of patents lapse before full term. He also notes the 11.5-year fee has gone from $3,100 in 2001 to $8,280, and asks whether that is what pushes owners out.
Our read is that the fee is only the trigger. What we see in portfolios is a decision made by whoever receives the docketing reminder, with one question sent to the business (“do we still use this?”) and nobody checking whether a competitor does. That produces both failure modes at once. Companies keep paying on discontinued products, which our portfolio management page calls the most common and most fixable leak in an unmanaged portfolio, and they drop patents a competitor is quietly practicing because nobody in the room knew.
A high lapse rate can be healthy for a company whose business moved on. What is unhealthy is a lapse rate nobody chose.
LexisNexis IP estimated in October 2025 that companies will spend more than $177 billion on renewal fees over the next two decades, nearly $10 billion of it in 2026, and put the average pruning rate across 133 large companies near 20 percent, from 13 percent in aerospace and defense to 28 percent in automotive. Those are vendor estimates about very large portfolios, but the direction matches what we see: everyone prunes, and the well-run portfolios prune on purpose.
An IPWatchdog panel on AI-assisted pruning in September 2026 listed legal status, projected maintenance cost, business alignment, claim detectability, competitive overlap, and infringement signals. Good list. Ours puts the business questions first, because that is where the answers are usually missing. For each patent, ask:
Weak patents cost nearly as much to keep as strong ones, as our quality versus quantity page argues; the rubric puts a number on that.
Score each question 0, 1, or 2, where 2 is a clear yes (or, for cost, a low number). The total is out of 12.
Two overrides. A 2 on product coverage keeps the patent regardless of total. A 0 on product coverage, competitor use, and buyer interest together sends it to the sale list regardless of remaining life. The rubric is deliberately coarse. In the sessions we run, six people can score forty families in an hour, agree on the top and the bottom, and spend the argument on the middle, which is where it belongs.
Vendors now sell software for this: Patlytics scores up to 250 patents against competitor products for read likelihood, and LexisNexis scores on citations and market coverage. On several hundred families, a screen like that is worth running first. Under fifty, the six questions in a spreadsheet are faster, and the people scoring remember why each patent was filed, which no citation count knows.
The fence patents are the ones worth ten more minutes. Run each of them through our free Patent Valuation Calculator for a directional number, and if it comes back well above the fee, that patent belongs on the sale list rather than the lapse list.
Here is the step most companies skip. A patent that fails the rubric for you can still pass it for someone else: a competitor who practices the claims, a company whose product line sits in the patent’s field, or an aggregator building coverage there. On our page on what patents sell for, most sales close in the five to six figures, with outliers far higher, and buyers pay a premium for clean title with maintenance fees current. An expired patent has nothing left to sell, and a patent inside its grace period is a distressed asset whose fee date the buyer can read off the USPTO record.
That sets the calendar. Published accounts on the same page put a sale process at six months to two years, so start it at the previous window, or at least a year before the fee date. Pull the assignment chain and the file wrapper before you approach anyone, because a buyer’s counsel will. If the deal is real and slow, pay the fee and keep talking; we tell clients that buyers who pass often come back six months to two years later, when their own product plans change. We covered the buyer’s side in How to sell a patent (and what buyers actually pay).
For the fence and sale lists, the tool we use is a monetization assessment, around $5,000 on the same page: it tells you whether there is a buyer, and who, before you spend a year finding out.
A maintenance fee buys years of a right. Decide what you will do with them before you pay for them.
At 3.5 years the fee is small and most of the term is ahead, so the default is to keep unless the technology is dead or the family is redundant. The real decision here is usually about continuations: if one is still pending, this is the moment to decide whether the claims need reshaping toward the product that actually shipped, and that is a patent attorney’s call. We talked through continuations and the other post-grant decisions on Invent Anything episode 71: So You Have Your First Patent! Part 2.
At 7.5 years the fee nearly doubles and the picture is honest. The product shipped or it did not, the competitors are known, and with a 20-year term measured from filing and a few years of pendency before grant, there are roughly ten years of term left, which is what a buyer wants. This is the window where selling makes the most sense and where the rubric earns its keep.
At 11.5 years you are paying $8,280 at undiscounted rates for the last stretch of term. Keep it if it covers revenue or sits in an active licensing program, and let the rubric say no otherwise. A short-life patent with strong evidence of use can still be worth asserting, but that is a decision to make with counsel and a budget in hand.
One reason to keep a patent that the rubric misses: option value. We have advised founders whose market was not ready to hold a patent dormant through a window or two, because the fee is a fraction of what it costs to keep a company running while the market matures. Score it honestly, then override on purpose. Schell IP put the other half well in June: there is no shame in letting a patent lapse on purpose when it no longer protects anything valuable. The operative words in both cases are on purpose.
Outside the U.S. the decision is annual. A European application owes a renewal fee every year from the third year after filing, rising from 725 euros to 1,865 euros by year ten and flat after that on the EPO schedule in force since April 1, 2026, which would total about 29,000 euros if those EPO rates ran for a full twenty years. In practice, once the grant publishes, the EPO Guidelines hand the renewal bill to the national office of each country you validated in, each on its own schedule, so the real total depends on where you validated. Broad international coverage runs 10 to 15 times a U.S.-only budget over a patent’s life.
So the rubric abroad gets one more question: do we sell or make there, or does a competitor? A country that answers no to both is the first place to prune, and pruning one family across five countries usually frees more budget than any single U.S. decision.
Pull every maintenance fee and annuity due in the next two years into one list, with entity status checked on each. Score that list with the rubric in one meeting, with sales and engineering in the room, and give every fence patent an owner and a date. Then start sale conversations on everything that scored 0 to 4 now, while there is still term to sell.
If the list is longer than an afternoon can handle, or the fence pile is most of it, Talk with our team about a portfolio review before the next window opens.
Work with ipCapital Group
From invention to monetization, our team has guided 2,000+ engagements across the full IP lifecycle. Start with a free 30-minute discovery call.
Written by
Seth Cronin