The canonical proof-of-work system, hashcash, tries to solve the problems "free e-mail makes spamming profitable" and "users won't pay for e-mail" by imposing a

The canonical proof-of-work system, hashcash, tries to solve the problems "free e-mail makes spamming profitable" and "users won't pay for e-mail" by imposing a cost in CPU time on the sender: this is free[1] for a normal sender, but a spammer will only be able to send out a limited number of e-mails. Unfortunately, this idea doesn't actually work. If an old "dumb" phone can send e-mail to a couple of your friends, a cool gaming rig can send zillions[2]. (Limiting senders by memory bandwidth instead of CPU power helps, but I'm not sure if that's enough.) Spammers have been using botnets for a long time. I'm also not aware of any scheme that has considered attacks based on custom hardware. Mailing lists are also a problem. [1] Of course, the extra calculations do increase power use, but that's sufficiently invisible that people won't actually care. For instance, SETI@home and other BOINC projects may consume ~$7.40 per month in electricity, but people generally don't care (source: http://boinc.berkeley.edu/wiki/Heat_and_energy_considerations http://boinc.berkeley.edu/wiki/Heat_and_energy_consideration...). [2] [EDIT: added in response to andrewcooke.] Let's substitute a cheaper computer (because it's indeed more cost-effective than a gaming rig), and guess at some figures. Assume your 100x speedup (four-core 2GHz vs single-core 100MHz gets you to 80x; highly-optimized vs crappy code can probably get you far higher, but let's stick with 100x for now.) If the phone can send one message per 10 seconds, the PC would send out 100 x 24 x 60 x 60 / 10 = 864 000 messages per day (assuming it's only limited by the proof-of-work!) http://www.icsi.berkeley.edu/pubs/networking/2008-ccs-spamal.. http://www.icsi.berkeley.edu/pubs/networking/2008-ccs-spamal.... cites a ~0.00001% conversion rate on a ~$100 product, or a profit of $1/10 000 messages; that would allow a mind-blowing $86 of scalable profit per day. In reality, that message count sounds unreasonably high - I'd guess the spammer runs into bandwidth/antispam limitations before CPU limitations.