His Dell rep is taking him for a ride. He needs to call HP and Lenovo, and get quotes, pick the lowest one, show it to the others, and ask for new quotes. Repeat a few times. Even if he still picks Dell in the end he should be able to get a better price and probably some perks. We get our Dell and HP reps to compete over 100 desktop orders, I'm sure they will be willing to negotiate for a 1000+ order. Building your own machine is asking for disaster. I love building my own machines, and wind up with a better machine for it, but I'm a geek and if my motherboard blows up tomorrow I have enough spare parts laying around to cobble together a new pc. When all the ram in all your pc's starts failing over a 2-3 month period, and the ram vendor is blaming the motherboard vendor and the motherboard vendor is blaming the ram vendor and neither will give you the time of day, you are going to be screwed. That's not a fictional scenario, that's a real life scenario from my personal experience, and it happened to about 300 pc's and was a nightmare, I wouldn't want to see it with 1000.
Your rep probably knows you are a "dell shop" and you want to stay that way (because mixing vendors is a pain in the ass for support) so he figures he will through out a high number and hope you will bite. If he won't budge, hang up the phone, look up the sales line on Dell's site, and start calling until you get someone willing to play ball. The real money is in servers anyways, so while you are negotiating feel free to threaten to take your desktop AND server business to another vendor. You have to fight for it but PC vendors will sell to you at almost any margin above break even for them, because they want their foot in the door. Building your own PC's is kind of like not having insurance. Sure, you save a little money, and for most people, nothing bad will happen. The few that do have something bad, it will be catastrophic. You buy insurance to spread the risk. The same with PC vendors. They do a lot more testing up front, and they are large enough to absorb particular models having high failure rates without it becoming a complete catastrophe.