It's like replacing the radiators in your house with servers or bitcoin miners; you still pay for electricity, and it's still warming up your house, but you get compute cycles or bitcoins "for free". If the waste heat/light of a fusion device is of comparable efficiency as conventional heaters/lighting, then we're "breaking even"; if the device manages to perform any fusion, that liberated energy would be a "net positive".
Of course, a fusor doesn't seem like a particularly good idea for household heating or lighting, and that argument is completely inapplicable to things like power plants.
Actually, it's not unheard of for waste heat from power plants to be used as residential heating.
https://en.wikipedia.org/wiki/Cogeneration
Some of the nordic countries use local power plants to create municipal steam for heating homes, I believe.
If you put a bounding box around a Farnsworth Fusor, and put in 100W of energy, would you not get a little more than 100W output from the fusion reaction?
Obviously still not viable as self-sustaining, since 100W of electric energy in would be lower entropy than the 100W+ coming out, but it would be a method of producing some energy as long as it's fed from another source.
Obviously, using a heat pump would be a much better way of turning 100W of electricity in to >100W heat, but still, a Fusor isn't net energy negative (although it is terrible with conversion of entropy).
As to 100W vs. > 100W. 100.001W might as well just be 100W in just about any meaningful way. However, Jet is vastly cheaper per watt of liberated fusion energy. https://en.wikipedia.org/wiki/Joint_European_Torus
https://en.wikipedia.org/wiki/Fusion_energy_gain_factor is meaningful, but it's talking about inputs not outputs.
PS: Ok, heat is a poor description. 1,000 pars light, 99,000.9 units of heat, and 0.1 unit high energy particle radiation.