Can a sequencing method be 100 times faster than ABI but more expensive?

The short answer is YES.
The 2005 Nature paper from the 454 group demonstrates a 580kbp genome at $5000.
That is $9 per kbp while ABI is about $7 per kbp (at better accuracy too).

The longer answer is below:
100 ABI machines in a box could be called "a machine which is 100-times faster" than a current Applied Biosystems Instrument 3730xl, but it would not be cheaper. Cost analyses include the cost of lab space and equipment amortization.
ABI vs 454 vs Polony Cost-Error Comparison & Comparison spreadsheet
Polony cost details

Below is a related correspondence/correction for a recent NBT news item.
"Sequencing thoroughbreds"

A news story by Jim Kling that appeared in the November issue is timely and well researched (Nat. Biotechnol 23, 1333-1335,2005); however two points merit clarification.

Kling suggests that the 'polony' sequencing method developed by my group at Harvard and collaborators at Washington University [ref 1] "avoids potentially costly PCR". In fact, our method does not completely "avoid" PCR, it merely attempts to use less of the costly enzymes by employing beads 20,000 times smaller in volume than those used in the method of the 454 group [ref 2].

The article also states that "454 estimates that it achieved a 100-fold decrease in cost compared with the Sanger method". This is quite different from "100-fold increase in throughput", which is what they said in their Nature paper. The 454 cost was $5000 for a 580 kbp genome = $9 per consensus kbp (at an error rate of 4E-5). Sanger/ABI is $7/kb at an error rate of 4E-6. Both methods can achieve tradeoffs between cost and accuracy, but published data does not yet indicate any point where 454 is less costly than Sanger for a given accuracy goal. We estimated polony sequencing at $0.8 per consensus kbp ($0.11 / raw-kbp [ref 1]) and an error rate of 3E-7. More detailed cost-accuracy tradeoff analyses are to be encouraged and a possible start point is available here:

--George Church, Jay Shendure and Greg Porreca
Harvard Medical School, Boston, MA 02115
1. Shendure, J. et al. Science 309, 1728-1732 (2005).
2. Margulies, M. et al. Nature 437, 376-380 (2005)

23-Nov-2005 (updated 21-Jan-2006) by GMC from Arep.