Manuel DeLanda's preeminent virtue as a scholar is the way in which he applies the ideas of complexity theory (emergence, feedback, etc.) to the historical record, and War in the Age of Intelligent Machines (1991) follows this template, looking at moments where technological developments (the conoidal bullet, wireless technology) spur military systems to evolve (a process which, in turn, triggers other armies to evolve in response).
If you accept this premise (fail to at your peril), it naturally suggests that the militaries of today will one day evolve even further. So in addition to sketching out historical instances of this sort of thing, DeLanda spends a lot of time drawing attention to contemporary developments in technology or military theory that might be putting us on the road to future phase shifts that might spell Bad News for soldiers and civilians alike. Artificial intelligence, RAND-style war game simulators, and predatory machines (of the sort outlined in DARPA's "Strategic Computing Initiative") all come in for an extended critique, although DeLanda seems more optimistic about technological systems that don't take human beings "out of the loop" (the book ends with an appreciation of humanist interface designer Doug Engelbart).
All in all, this book is pretty essential reading for anyone interested in the "machine" part of the war machine, although it could definitely benefit from a little revision and expansion: some of the Cold War anxiety undergirding the book has lost some of its edge in the intervening years, and I could stand to lose some of it in favor of having DeLanda as a guide through past two wars (although War was published in 1991, Desert Storm hardly ranks a mention, a little odd, given the use of Israeli-built Pioneer UAVs in that conflict).