There is some debate about the theoretical size of a multi-stage fusion device. Some say it is unlimited, while others debate this.
The basic action of the Teller-Ulam design is to use the energy from one explosion to set off another. There is an important requirement however, that energy must be deposited very rapidly so the second reaction has largely completed before the mechanical effects of the original explosion reach the next stage. In the case of a two-stage device, a small plutonium bomb (typically) is used as the trigger and it's x-rays are the energy source for the next stage. The x-rays fill the inside of the case at the speed of light, while the explosion is travelling orders of magnitude slower. This gives the secondary time to heat, compress and explode before the debris from the first stage reach it even across the space of the bomb casing.
Three-stage designs are more difficult because you can't simply place the third stage next to the second, or the x-rays from the primary will hit it too. You need to arrange the third stage in a way that the initial explosion is not visible to it, or use some other energy from the secondary. It is not clear which the Tsar Bomba used, and there is speculation that there were several third stages, not one large one.
It remains debated whether a truly unlimited size bomb is possible. It is known that the US was working on the theoretical design of a 1 GT weapon and even a 10 GT one. However, such weapons would be so large as to be almost fixed. I do know a 150 MT bomber-carried design was seriously considered, but disappeared during the move to ICBMs.
Even if there is no theoretical limit, there is a very practical limit, however, which has resulted in the reduction in yield over time. Since the energy of the bomb is distributed spherically, as the bomb yield grows more and more of its energy is wasted travelling up into the atmosphere. Four small bombs falling in a square pattern will cause more damage than one large one of four times the energy. Although the single bomb is likely to have a better power-to-weight ratio than four small ones (due to both physics and mundane issues like minimum case thickness) the geometrical improvement far offsets this. So in the early 1960s ICBMs carried a single large ~5 MT warhead, whereas by the 1970s they would instead carry multiple smaller ones around 250 to 500 kT.