According to the brochure titled "General Information Concerning Patents," any person "who invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent, subject to the conditions and requirements of the law." Corporations can get patents on inventions they hired employees to create, and the employment contracts of researchers generally contain a clause saying the employees sign their patent rights over to make sure there's no gray area.
An idea cannot be patented, it has to be an invention -- a device or method for actually doing something. You can't patent the suggestion of making a PC the size of a credit card, you have to actually invent a credit card-sized PC before you can patent it. The application must describe the patented invention specifically, and in detail. This doesn't mean your invention has to actually WORK; that's not required to get a patent. Plenty of stuff gets patented that has no actual useful purpose at all, and the patent office doesn't test functionality. The invention simply has to be unique and original; it's a new mousetrap, not necessarily a better one.
Patents may be licensed in about the same way as copyrights, a license being legal permission for someone other than the patent holder to use the patented invention. The owner of the patent does not need a license to use their own intellectual property, and actually selling or giving away the patent is a totally separate matter than simply licensing it.
Two companies in our portfolio depend on patents for their survival: our drug companies Pfizer (NYSE: PFE) and Schering-Plough (NYSE: SGP). Their job is to come up with patentable inventions (i.e., new drugs) that they can then sell at monopoly profit margins for the entire time the patent is in effect. Their business model is a race against time, developing new drugs before the patents on their current crop of cash cows expire.
Just because a patent expires doesn't mean the drug company has to stop selling their old drugs. It simply means they no longer have the legal authority to prevent competitors from manufacturing the same compound and selling it, usually for less money. Companies try to develop strong brand names (using trademark law) while their patents are in effect, such as "Tylenol" vs. the generic name "Acetaminophen," and thus can often maintain somewhat higher margins than competing generic brands even after their patents expire. But without the patent, their monopoly is no longer absolute, and thus they face competitive pricing pressures narrowing their margins below the limit of what the market will bear.
Some of our other companies use patents as part of an overall strategy, along with copyrights and brand names. Intel's (Nasdaq: INTC) main line of intellectual property defense against competitors is that its microprocessor designs are copyrighted, so other people can't copy Intel's processors directly but have to design their own circuitry to do the same thing. Ever since the Pentium, Intel has also been wielding trademarks like a chain saw, with "Intel Inside" stickers everywhere and television commercials piling on brand awareness, and even a mascot (the dancing guy in the multi-colored clean room "bunny suit").
On top of this, Intel regularly files for patents on specific inventions within its processor designs, such as a new cache architecture or new way of connecting the processor to the motherboard.
On the cutting edge of technology, patents get about as messy as trademarks. Unfortunately the messiness of trademarks is due to the involvement of judgment calls and common sense in determining the scope and value of brand identity, and the messiness of patents has to do with the patent system being broken. Patents aren't supposed to be granted on something that is obvious" to someone "skilled in the art," but the fact is that patent clerks and even examiners are more federal bureaucrats than scientists and engineers. Whatever credentials they may have, their job is filling out paperwork, not advancing the state of the art. It's hard to retain the best and brightest and keep their skills current under those conditions, but the state of the art is, by definition, what gets patented. Add in the fact that the patent office is positively swamped with a flood of patent applications, and they sometimes wind up relying on the filers of the patent to find and disclose any "prior art" a lot more than they should. Even when they make the time and find the resources to look for it themselves, they wind up missing a lot.
The end result is that in areas like software development, patents are almost universally considered to be evil. Since the PTO can't determine what is original and non-obvious in that area, they approve patents every day on things that are either incredibly obvious to most specialists in the field, or were in widespread use in the industry well before a patent was filed for. (Here's an old but good article from Wired on that: Patently Absurd.) Sometimes, the PTO doesn't even seem to look at patent applications before approving them, as in the classic patent on a "method for exercising a cat."