If I were doing this, there'd be a number of questions to ask yourself before you dumped a lot of money into it: 1. Is there a version of Windows for the RasPi

If I were doing this, there'd be a number of questions to ask yourself before you dumped a lot of money into it: 1. Is there a version of Windows for the RasPi (I think this is true, but it's been a while)? 2. Will Swarm and any other software you'd need for the cluster also run on the RasPi version of Windows? 3. How well does a single instance run on the RasPi? At that point, assuming #3 looks good - you could take it to the next level. Try two RasPi systems connected via a gigabit switch (cheap 5 port), and see how the performance scales vs a single instance. Then try four RasPi systems. At that point you should have enough data to determine how well a large render farm would run, compared to your desktop or whatever else you could come up with (would it be cheaper to spin up virtual servers as needed on AWS or similar?). Also realize that $3500 (approx) will only cover the RasPi systems; you'll need even more money for the power supply(s), cooling fans, switches, networking cables, power cabling, "enclosure" (even if just stacked using standoffs, you should still figure out how much all that will cost), etc. $4500 might be a more realistic figure. Even so - if it worked, and worked well - that dollar amount might not be that big a deal. Worst case, you could get a HN article out of it (whether as triumph or post-mortem - it would still be an interesting read).