After actually reading the paper, I see that you are correct, and they did indeed do a lot of work to paste all the 1000 bp fragments into a 1M bp sequence. It is certainly very cool.And these machines are incapable of producing anything beyond 1000 bp, as the strand grow longer on the plate they tend to cover each other and reduce the accuracy, by 1000 bps there a virtually no strands of perfect desired sequence. To achieve a 1,000,000 required pasting together many many smaller strands and repeated testing and filtering for defective sequences, newer micro scale (microchip labs) have allowed for improvement in fidelity and automated filtering out of defective strands. This achievement was a brute force undertaking that cost millions, now that the technology is developed the next ones will be a fold or two cheaper.
But like I said, they didn't "create new life from nothing," or even design a new genome from scratch. They just made a synthetic copy of an already-sequenced genome (with a few watermarks etc) and stuck it in a cell, and then showed that the cell worked with the new DNA. Cool stuff, but not exactly what some of the posters here seem to be imagining.