Everything you need to read a Big Tech patent like an insider: what one is, how it travels from filing to grant, how to read the part that actually matters, and what it does and does not tell you.
Every week, the biggest technology companies on earth file thousands of patents. Most coverage of them falls into two camps: breathless headlines about a single wild-sounding filing, or dense legal explainers written for patent attorneys. This guide sits in the middle. By the end, you will understand the patent system well enough to read any Big Tech filing and know what it actually means, which is a more useful skill than it sounds, because patents are one of the few honest windows into what these companies are really building.
We read every Big Tech patent the day it publishes, so this is the working knowledge we use, written in plain English.
What a patent actually is
A patent is a deal with the government. In exchange for publicly explaining how your invention works, you get a temporary legal right to stop other people from making, using, or selling it. The key word is stop. A patent does not give you the right to build your own invention. It gives you the right to exclude others from building it. That distinction matters, because two companies can hold overlapping patents and each can block the other, which is how a lot of technology standoffs begin.
In the United States, a utility patent lasts 20 years from the date you file it. After that, the invention belongs to everyone. The trade at the heart of the system is disclosure for protection: society gets to read exactly how your thing works, and in return you get two decades of exclusivity.
One distinction trips up almost everyone: a patent application is not a granted patent. An application is a request, filed and then examined, which may be approved, narrowed, or rejected. Most of what gets written about in the news, including most of what we cover, is applications, because they publish long before they are granted. They show intent and direction, not a finished legal right.
The lifecycle: from filing to grant
A patent travels a fairly predictable path.
Filing. The company submits the application to the patent office. The filing date is important: it sets the clock and establishes priority, which is the system’s way of deciding who got there first.
Publication. Roughly 18 months after the earliest filing date, the application is published for the world to see. This is the moment a secret R&D project becomes a public document, and it is the moment most patent coverage begins. The company has no choice in the matter for most filings. The 18-month publication is automatic.
Examination. A patent examiner reviews the application against everything that came before it, called prior art. The examiner almost always pushes back at least once, in a document called an office action, arguing that some claims are too broad or already exist. The applicant amends and responds. This back-and-forth can run anywhere from one to several years.
Grant or abandonment. Eventually the application is either granted, often with narrower claims than it started with, or abandoned. A granted patent is the enforceable right. An abandoned one is just a public disclosure that anyone can now use.
The publication step is why we can write about a filing years before anyone knows whether it will be granted. The document is public; the outcome is not yet decided.
The four types you will run into
Not all patents are the same kind of thing.
A provisional application is a placeholder. It is not examined and never becomes a patent on its own. It buys 12 months to lock in an early filing date while the real application is prepared.
A non-provisional application is the real thing, the one that gets examined and can become a granted patent.
A continuation is a new application that claims a different slice of an invention already disclosed in an earlier filing. Companies use continuations to keep a patent family alive and to pursue broader or different claims over time.
A divisional splits one application into several when the patent office decides it actually contains more than one invention.
When you see a company file a flurry of related applications, you are usually watching a patent family grow through continuations and divisionals, not a dozen separate ideas.
The anatomy of a patent, and the only section that legally matters
Open any patent and you will find the same parts in the same order:
- Title and abstract. A short name and a paragraph summary. Useful for orientation, but written to be vague.
- Background. The problem the invention claims to solve, usually framed to make the invention look necessary.
- Drawings. The figures, labeled with reference numbers. For hardware especially, the drawings often tell you more than the text.
- Specification (the detailed description). The long, technical walk-through of how the invention works. This is where the real engineering lives.
- Claims. The numbered sentences at the very end.
Here is the thing almost everyone gets wrong: the claims are the patent. Everything before them is context. The claims are the legally operative part, the precise definition of what the patent actually protects. When lawyers fight over a patent, they fight over the exact wording of the claims, not the friendly summary up top.
How to read the claims
Claims look intimidating because they are written as single, enormous run-on sentences with strange punctuation. The structure underneath is simple.
Claim 1 is usually the one to read. It is typically the broadest independent claim, meaning it stands entirely on its own and defines the widest version of the invention the company is trying to protect.
Independent claims stand alone. Dependent claims narrow them. A dependent claim starts with something like “The system of claim 1, wherein…” and adds a limitation. Each added limitation makes the claim narrower and easier to defend, but also easier for a competitor to design around.
Every word is a limitation. In a claim, the more words, the narrower the protection. A claim that says “a device comprising a sensor” covers far more than one that says “a device comprising an infrared sensor mounted at a 30-degree angle.” Companies start broad in claim 1 and add fallback positions in the dependent claims, in case the broad version gets rejected.
Read claim 1, mentally strip away the legal scaffolding, and ask: what is the actual thing being claimed here? That single habit is most of what separates a useful reading from a confused one. If you want a worked example, our companion piece on how to read a Big Tech patent walks through one line by line.
How patents are organized: classification codes
With millions of patents in existence, the system needs a filing cabinet. That cabinet is the Cooperative Patent Classification, or CPC, a hierarchical set of codes that sort every invention by subject. A filing about a neural network and a filing about a camera lens get different codes, which is how you can pull every patent in a given technology area.
Classification is also how a publication like ours organizes coverage by topic and by company, so you can follow, say, everything a single firm is filing in augmented reality rather than drinking from the firehose of everything at once.
Why Big Tech files so many
If most applications never become products, why do the biggest companies file thousands a year? A few reasons, and almost none of them are “we are about to ship this.”
Defense. A large patent portfolio is a deterrent. If a competitor sues you, having a deep arsenal of your own patents to counter-sue with creates a standoff. Most of Big Tech’s patents exist to be never used, the way most of a country’s weapons do.
Blocking and cross-licensing. Owning the patents around a technology lets a company either block rivals or trade access. Whole industries run on quiet cross-licensing deals that let everyone build without suing everyone else.
Optionality. Filing is cheap relative to the cost of being locked out of a future direction. Companies patent broadly to keep doors open, even doors they may never walk through.
Prestige and metrics. Patent counts get reported, celebrated, and used internally to measure research output. That incentive alone produces a lot of filings.
The upshot: volume is not the same as intent. A single, specific, repeated line of filings tells you far more than a raw count.
What a patent does and does not mean
Because this is where most patent coverage goes wrong, it is worth being blunt.
A patent does mean a company spent real money and legal effort to stake out a specific idea. It is a credible signal of where R&D attention is going.
A patent does not mean a product is coming. It does not mean the feature will ship, or ship in this form, or ship at all. It does not mean the company has solved the hard parts. And it does not mean the company endorses or intends everything the filing describes, since applications are often written as broadly as the examiner will allow.
Hold both ideas at once and you can read filings the way we do: as a leading indicator, not a roadmap. The patent record is an early, noisy, fascinating signal of where the industry is leaning, years before the keynote.
When patents become public, and where to read them
The patent office runs on a weekly rhythm. In the United States, granted patents issue on Tuesdays and newly published applications appear on Thursdays. That cadence is why patent coverage, including ours, tends to move in weekly waves.
You can read any published filing yourself through the USPTO’s Patent Public Search or through Google Patents, both free. The raw documents are dense, which is the gap we exist to fill: we read the new Big Tech filings as they publish and translate them into plain English, with the source always one click away.
Now that you know how the system works, every one of those filings is readable. Start with whatever company or technology you care about most, go straight to claim 1, and ask the only quest