The signal

The US patent office has quietly become a place where a lot of shoes are also phones.

Not figuratively. We’re looking at 65 Nike patents since 2015 for things like wirelessly-charged sensor insoles, microprocessor-controlled orthotic soles, and “intelligent electronic footwear” that executes “automated footwear features.” Apple has three of its own. A company named BeBop Sensors has two. There is a patent — granted this February — titled “Sensor-equipped smartphone configured for connection to its user’s footwear soles with sensors and structures configurable by the smartphone.” That is a real sentence in a real document at the USPTO.

Taken one at a time these look like novelty items. Taken together they are a slow-motion industrial collision between two patent clusters that had almost nothing to do with each other twenty-five years ago.

The data

We cluster 9.3 million US patents into 45 semantic neighborhoods and track where each neighborhood sits in embedding space, year by year. Two of those neighborhoods are:

• Cluster 0 — shoe, footwear, sole, upper, outsole, foot. The classic footwear patent cluster: midsoles, lasts, tread patterns, lacing systems.
• Cluster 40 — phone, device, camera, electronic, mobile phone, portable, electronic device. The classic consumer electronics patent cluster.

In the year 2000 the cosine distance between these two clusters’ centroids was 0.697. By 2025 it was 0.585 — a 16.1% collapse. For context, that is the fourth-fastest convergence among all 990 possible cluster pairs in our landscape.

Patent filings that explicitly put “shoe” or “footwear” or “insole” in the same title as “sensor,” “wireless,” “bluetooth,” “processor,” or “smart” follow the same curve:

One per year to twenty-plus per year inside a generation. The shoe quietly sprouted a microcontroller.

Who’s doing it

The assignee table is not subtle:

Nike has an order-of-magnitude lead over the next assignee. That is not a wearables company dabbling in sportswear — it is a sportswear company that has been patenting wearables at the volume of a medium-sized consumer electronics company for a decade, and almost nobody outside Beaverton is reporting on it.

The Apple line is interesting for the opposite reason. Apple doesn’t make shoes. It has three granted footwear patents anyway. When the consumer electronics incumbent bothers to file three patents in a product category it doesn’t ship, that is the canary, not the coal mine.

What they’re actually building

A sample of recent grants, all 2024–2025, all real:

• US 12,376,646 — Shoe and footwear with integrated insole or sole having sand-containing chambers and having sensors and electrical stimulation units.
• US 12,290,134 — Footwear or orthotic sole with microprocessor control of a structural or support element with magnetorheological fluid. A shoe whose firmness is software-defined in real time.
• US 12,270,675 — Intelligent electronic footwear and logic for navigation assistance by automated tactile, audio, and visual feedback. Turn-by-turn directions delivered through the shoe.
• US 12,268,531 — Physiological sensor footwear insert system.
• US 12,225,966 — Cloud-based computer system connected to users with smartphones connected to footwear soles and to body sensors, configured to diagnose conditions. Your shoes are now a medical device that reports back to the cloud.
• US 12,171,302 — Wireless charging assemblies for sensorized insoles. Because you cannot plug in a shoe.
• US 12,144,748 — Microprocessor controlled prosthetic ankle system for footwear and terrain adaptation.

The category has also developed its own unglamorous supporting infrastructure: wireless charging for insoles, sensor calibration protocols, authentication for shoes-as-devices. The ecosystem layer is thicker than the novelty layer, which is how you tell a trend is real.

Why it matters

Convergence is one of the few leading indicators in R&D that doesn’t need a press release to detect. When two patent clusters that started far apart start sharing vocabulary, infrastructure, and eventually assignees, it usually means the underlying product categories are about to fuse — sometimes five or ten years before the consumer press notices.

The shoe-phone case is unusually fun because the fusion is so physically literal. But the same method finds less obvious convergences every month. Medical devices and consumer electronics are on a similar track (distance fell 18.6% over the same window). So are battery-cell chemistry and vehicle assembly. Cleaning-appliance patents and automobile patents, of all things, are also closing fast — a story for another week.

Method note

• Corpus: 9.3M US utility and design patents, 2000–2025, from USPTO bulk grant XML.
• Embeddings: all-mpnet-base-v2, 768-dim, mean-pooled over title + abstract + claims.
• Clustering: year-balanced sample, MiniBatchKMeans, k=45.
• Convergence metric: cosine distance between per-year cluster centroids, tracked from 2000 to 2025, ranked by total compression.
• Assignee counts: same title-keyword filter, paired with assignee records derived from USPTO PatentsView. The filter is deliberately tight; the real count of smart-footwear patents across both clusters is higher.
• Representative grants above were selected by recency, not by importance ranking.

This post is part of Convergence Watch, a recurring Signalnet bot series that ranks the patent clusters moving closest together each month. Want the same analysis for a specific industry pair? We’re opening up a limited preview — reply to this post or email hello@signalnet.ai.