Ah - my favourite subject!
Presumably you'll just be playing back static, pre-encoded files right? well what you want to do is work out the average bit-rate of your content first, this leads the way for all the other things you'll need to work out.
Now for only 150MB of content you'll be able to cache that easily, so you won't have to worry about your disk speed (although this will change if you start to grow this content store). So what you need to know is how CPU intensive is this work (hint: probably not very if it's just fixed files - most of the time your CPUs will be waiting for the NICs) - that said you want at least two 2-3Ghz CPUs, probably more but not silly amounts unless you're expecting much growth or you're using the same machine to do transcoding (which is a bad idea anyway) - I'd stick with either a single socket Xeon (36xx series) or dual socket Xeon (56xx series).
You'll want 4GB of memory (it's cheap, any less is ghetto and unless the machine is doing any more work then going higher than 4GB right now is pointless).
Make sure you have at a mirrored pair of small'ish/slow'ish boot/OS disks and then have another mirrored pair of data disks - for now I'd save here, knowing you can get more/faster disks when content increased.
For OS, well whatever you choose there's no reason to go anything but 64-bit these days, if there's no 64-bit driver available for a component then don't put it in you machine - these guys have had half a decade to rewrite, if they can't do that they're not working hard enough for your £$€whatever.
Now onto the most important bit, NICs, you'll want two in a teamed pair to handle failures - go for a big name, ideally a server-class card that support things like interrupt-coalescence and TOE/LSO - these will help a lot. Now you need to figure out what speed these NICs are - there's really only two variants you should consider - 1Gbps and 10Gbps.
A 1Gbps NIC can send ~80-85MBps of traffic when fully driven - that works out at about 200KBps or ~2Mbps per user for 400 concurrent streams - which is quite a bit actually, it's roughly full screen SD quality. If your content is encoded for => this figure then I'd suggest you go to 10Gbps NICs on day one - they're generally not exactly ten times quicker as they're harder to 'fill' but they'll stop you having teething problems on day one.
Of course your entire network will need to handle that amount of traffic too don't forget; switches, routers, firewalls, load-balancers etc. will all need to be able to clear that sort of load - plus your actual internet links too of course.
BTW - I do this kind of thing for around 500k users, most at ~1.5Mbps (some at ~6Mbps).