A Seed Factory is a new kind of production system which has two major differences from conventional factories: self-expansion and integration. It includes a starter set of equipment which can expand in three ways to a larger capacity, while also making useful products:

Replication - making copies of its own parts so as to eventually copy the entire set.
Diversification - making parts for new equipment, thus expanding the range of possible outputs.
Scaling - making parts for larger equipment than what is in the starter set.

The factory is designed as an integrated system. It brings together multiple production steps from raw materials and energy to finished items. Each part of the factory produces resources needed by the other parts to function, making it self-sustaining. Since the parts are all in one place, the whole production chain can take advantage of automation and robotics. Integrated processes can also take waste outputs from one step, and use them as inputs for another. The combination results in a highly efficient design.

As the factory expands, it produces a growing variety of products for end users, and an increasing percentage of the items for its own growth.

In an interview, he answers some questions about seed factories:

Why use the metaphor of a “seed” for self-constructing factories?

Biological seeds grow into larger organisms using local matter and energy, and eventually produce copies of the original seed. By analogy a seed factory grows from a small starter set to a larger factory using local matter and energy and high levels of automation. The factory is flexible and general purpose, and produces a variety of useful products. It's also intended to be self-replicating, producing more seed factories.

We talk in terms of a factory rather than a single machine because (1) a number of different production processes are required which are best carried out separately, and (2) for the size and quantity of products we want to make, the final set of equipment is closer to commercial building size than garage or desktop size.

Why hasn’t this been done before?

For any number of reasons, historical, social, political, and economic, we tend to divide production into separate factories—the oil refinery is separate from the steel mill, which is separate from the auto plant. This separation requires tremendous cooperation and coordination among the various manufacturers to bring all of the raw materials and fabricated parts together to assemble the complex *things* of our modern lives. In short, the old manufacturing model makes 21st century optimization impossible.

What's the biggest obstacle to Seed Factories?

There is no simple answer, but it's easy to illustrate the problem. In 1908, Ford revolutionized manufacturing and the automobile market with the introduction of the Model T. There was a popular joke at the time that you could get one in any color, as long as it was black. These days, black may be considered the height of fashion, but only because everything comes in such a wide variety of colors, sizes, shapes, flavors, and so on ad nauseam.

The point being, manufacturing is a much more complex process 100 years later. Our old-style factories have been abandoned, because it seemed impossible to retrofit and update them (though labor issues and profit margins played a large part in this). This has left much of the West with an industrial vacuum; an enormous decrease in manufacturing activity, because we've innovated ourselves out of the business.

Science fiction fans have been anticipating this development since the idea was explained and fully visualized by sf great Philip K. Dick, in his 1953 short story Autofac:

The cylinder had split. At first he couldn't tell if it had been the impact or deliberate internal mechanisms at work. From the rent, an ooze of metal bits was sliding. Squatting down, O'Neill examined them.

The bits were in motion. Microscopic machinery, smaller than ants, smaller than pins, working energetically, purposefully - constructing something that looked like a tiny rectangle of steel.

"They're building," O'Neill said, awed. He got up and prowled on. Off to the side, at the far edge of the gully, he came across a downed pellet far advanced on its construction. Apparently it had been released some time ago.

This one had made great enough progress to be identified. Minute as it was, the structure was familiar. The machinery was building a miniature replica of the demolished factory.
(Read more about Philp K. Dick's autofac)