Here is an answer which is more rough, but easy to recall and useful as a first approximation.
Only the case of a NPN bipolar junction transistor is dealt here; things are similar for PNP bipolar junction transistors.
The basic assumption is that the B-E current is negligible with respect to the current through the collector, so, the collector current is approximately equal to the base current:
IE=IC=I.
If this assumption does not hold, then the transistor is probably misused or subject to a catastrophic failure.
Now, the power dissipated by the transistor is of course
P=VCEI.
To obtain an upper bound that is useful in the general case, we model the problem by considering that the collector is connected to
VCC through a resistor
R3, and that the base is connected to the ground through a resistor
R4 (this includes the load etc.). This is exactly the case in the OP problem. We have:
VCE=VCC−R3I−R4I=VCC−(R3+R4)I,
hence
P=(VCC−(R3+R4)I)I.
Using infinitesimal calculus, you find this expression of P is maximal whenever
I=VCC/2(R3+R4),
and equal to
P∗=V2CC/4(R3+R4).
This is the desired upper bound for the dissipated power whenever
R3 and
R4 are known. It means that:
Theorem: the power dissipated by the transistor is not larger than 14 of the power that would be dissipated by the two resistors R3 and R4 if they were directly connected.
In the OP problem, R3 is furthermore allowed to vary between 0 and 10kOhm, so, it is obvious that the expression of P∗ will be maximal for R3=0. This gives the upper bound
P∗∗=V2CC/4R4=100mW,
larger than, but not so far from, Olin Lathrop's bound.