Box weight is 500g+25g, so 10 boxes weigh 5000g+250g. Carton weight is 2000g+100g. Full carton weight=7000g+350g. The standard deviations have simply been added together to give the worst case deviation.
The z-score corresponding to 90% in normal distribution is 1.28, so 1.28*350=448g deviation from the mean of 7000g. Therefore, x=7448g, or 7.448kg is the weight below which 90% of the full cartons can be expected to weigh.
Perhaps a better way of combining the standard deviations would be to go for the RMS value based on the percentage of the mean of the standard deviation; so 250/5000=5%=100/2000. The RMS value is (sqrt(2*0.05^2))/2=0.025sqrt(2)=0.3536. Converted to weight this is 7000*0.03536=247.5g. In this case, 1.28*247.5=316.78g, so 90% of full cartons can be expected to weigh below 7247.5g approx.