Lower limits for steady-state black holes Stellar mass (and larger) black holes emit Hawking radiation below the temperature of the cosmic microwave background; thus, they should always absorb more energy from space than they emit, and always grow (at a very slow rate).
A black hole somewhere around the mass of the moon, on the other hand, were such an object to exist, should radiate at a high-enough temperature to remain in a steady state, emitting as much as it absorbs.
Going further, a black hole of $\sim4\times10^{20}$ kg (a little more than 1/200th the mass of the moon) should radiate at around a toasty 30 degrees Celsius. On its own, such a black hole would slowly evaporate, but, if we were to find one, it would not be terribly difficult to maintain it in a steady state by shining bright lights at it, or dropping in a slow stream of hydrogen atoms. At that point it is effectively acting as a very low-power but high efficiency matter-to-energy converter, turning hydrogen into thermal photons.
As a black hole becomes smaller and smaller, it seems that there are two basic factors which would make it increasingly difficult to arrange a steady-state situation. The first is increasing radiation pressure, blowing potential feed material away from the event horizon. The second is that the steadily decreasing size of the event horizon will make it more difficult to feed matter back in to the black hole.
The first issue is contemplated in the question Could micro black holes obey the Eddington limit?. Some back-of-the-envelope calculating suggests that the second issue may, however, present a more significant bottleneck. For reference, a black hole massing $4\times10%{15}$ kg should radiate about 22 watts of power, which is really not all that bright, but is only 2.97 picometers in radius - about 1/18th the size of a hydrogen atom. It seems to me that the black hole would have to be immersed in a fairly dense medium in order to make the rate of particle capture high enough to offset that 22 watts. How dense, I am uncertain; would air suffice? Or water? Or solid iron? Or electron degenerate matter, or neutronium? The issue of just how quickly you can force material into a sub-atomic sized blackhole is partially addressed by the paper Fermion absorption cross section of a Schwarzschild black hole, but I'm not quite sure how to get from that to calculating the necessary material density.
So, the question is: just how small can a black hole get, such that it can be kept in a steady state by some means?
(A necessary corollary to that is "what does the steady-state environment look like?"- i.e., does it have to be embedded in neutronium, or whatever. I realize it is preferable to ask one question at a time, but I'm not really sure how these can be separated cleanly; I welcome suggestions for improvement.)
 A: Take the example of a black hole of mass $M$, with a Hawking radiation output. Let's immerse it in a gas of hydrogen at temperature $T$ and density $\rho$ with no bulk motion or turbulence.
Ignoring radiation pressure, the mass-energy accreted at the Bondi-Hoyle accretion rate could be equated with the Hawking luminosity. 
$$ \frac{4\pi\rho (GM)^2 c^2}{c_s^3} \simeq \frac{\hbar c^6}{15360 \pi (GM)^2}$$
Assuming the sound speed $c_s = (5kT/3m_u)^{1/2}$
$$ \rho T^{-3/2} = \left(\frac{5k}{3m_u}\right)^{3/2} \left( \frac{\hbar c^4}{61440 \pi^2 (GM)^4}\right)$$ 
$$ \rho = 1.1\times 10^7 \left(\frac{T}{10^4\ K}\right)^{3/2} \left(\frac{M}{10^{10}\ kg}\right)^{-4}\ kg\ m^{-3}$$
It therefore seems (and a big caveat is someone should check my numerical calculation!) accumulating enough mass inside the Bondi-Hoyle radius is really not a problem for a black hole of mass $\sim 10^{15}--10^{20}$ kg.
At $M=4\times 10^{15}$ kg, and $L=22$ Watts, you only need to accrete $1.5 \times 10^{11}$ atoms per second, from a Bondi-Hoyle radius of $R = R_{Sch} (c/c_s)^2$. For a gas at $10^{4}$ K this means you only need to keep the gas flowing into a sphere of radius $\sim 10^9$ times that of the Schwarzschild radius. This is roughly the radius at which the gravitational potential energy becomes larger than the thermal energy of the gas particles. Note though that I'm happier with this answer at $10^{20}$ kg, than $10^{15}$ kg, since I don't have a quantum theory of GR for when the Schwarzschild radius is smaller than an atom!
A much bigger issue is how would you then dump enough angular momentum to allow the material to get within a few Schwarzschild radii (after which, accretion is inevitable)? I don't know enough about the scalability of the Physics to say much about the formation of mini accretion discs.
