I think it depends on two things: the LED spectrum and the photocatalyst absorption spectrum. But in general, considering the power for photocatalytic applications, I guess it could be about two 10W LED lamps for one 300W xenon lamp.
Ivan Moreno is fully right to state that the match between the LED spectrum and the photocatalyst spectrum mainly determines the efficiency. Furthermore, the means to concentrate light onto a certain spot, or produce short very intense pulse of light can also influence the choice (think of a production line where discrete spots need treatment). Another highly important aspect is the avoidance of irradiation at unproductive wavelengths, as this creates a heat load to the irradiated system. There are many cases where processes have a tolerance window for heating. In biomedical and organic chemistry applications, e.g. damage arises upon excessive heating. In general, if LEDs are available at useful wavelengths, they tend to be more efficient to produce the desired effect with less collateral damage by heating. And LEDs offer more possibilities to "overpulse" (i.e. short bursts of light at LED currents far higher than the continuous wave current can be).
I agree with both Ivan Morento and Fokko Pieter Wieringa. The match between the LED emission spectra with the photocatalyst absorption spectra can provide a far better process efficiency than the use of a broad spectrum Xe lamp. Also, the use of multiple LEDs (multiple point sources instead of an single lamp) can allow better irradiation geometries (better sample coverage) that will also be useful to increase the efficiency. The photocalytic process can be a lengthy one and source stability over time is a key issue. Depending on the LED technology used (again - depends on your required emission spectra), they tend to be more stable and durable than the Xe lamps yielding more reproduceable results over time.