The density of CdTe is 5.8 g/cc. This gives 3.08 g/cc of Te in CdTe.
The efficiency of FSLR's modules is ~11%. At a solar irradiance of ~1100 W/m^2, their cells should have a power density of ~115 W/m^2.
Assuming the thickness of their CdTe films is 3 µm, and at a 2.7 GW target production in 2012, they should be using 71 m^3 of tellurium per year in the cells alone.
This would mean 218 metric tonnes of Tellurium in the cells per year. Global production is currently at 200 and NREL estimates it can't go higher than 1,600.
The link below to a 2014 article that was largely referencing 2010 papers confirms the 3 µm Matt stated in the other answer in 2010. Here in 2019, the thickness has probably not reduced because there’s a fundamental limit due to light being able to pass through such thin films, especially since they have increased efficiency. Efficiency has risen to 17% from 11%, so combining this with Matt’s numbers gives 52 metric tons per GW. The 2014 article cited a 2010 article that gave 91 tons/GW whereas Matt calculated 80 tons/GW. So the 50 to 60 tons/GW should be accurate unless they substantially changed the thickness. Matt’s 200 tons/yr production seems pretty low, but the USGS was not providing that number that year. I had estimated about 400 tons/yr in 2008. USGS estimates about 500 ton/yr for 2019. Thanks to Trump’s tariffs, FirstSolar’s production is rising from 2.7 GW in 2017 (maybe 15% of world total) to estimating 5.4 GW in 2019. They are estimating 7.6 GW capacity at end of 2020 thanks to their new modules. This would require 395 tons which is uncomfortably close to word wide Te production. The Te spot price is currently about $80 per kg which is 0.42 cents per Watt, which is 2% of their series 6 manufacturing costs ($0.20/W).