You can't extract much energy from low-grade heat. Maximum efficiency = (T2-T1)/T1, where T2 is the temperature of the low-grade heat (exhaust from the computer system), and T1 is the ambient temperature of where you dump the heat. All temps are absolute (K). Typical chip surface temps are 60-80C, and 70C is about 343K. If your cold end is a cool day (say 15C), then T1 is 288K. The maximum energy you can recover from that heat is 19% of the input. That's a theoretical upper limit. Actual heat engines don't do that well.
Systems for absorption-Rankine low-grade heat recovery
have been proposed.. At 120C in, 25C out, it might be possible to get 8% efficiency from a practical system. So yes, you can get some energy out, but it's barely worth it.
We're pushing 180f today - 200 is likely next week. Our intent is not to make low grade heat it is to make high grade heat, primary heat, from computation. I really appreciate the comments but this has been vetted by folks with advanced degrees and we're partnered with a national lab, guys. Its not our intent to pull the wool over anyone's eyes~