Does the power test tell me I need way high N than if I use prop.test?

power.R

> prop.test(x = c(300,250), n = c(50000,50000), conf.level = 0.05)

	2-sample test for equality of proportions with continuity
	correction

data:  c(300, 250) out of c(50000, 50000)
X-squared = 4.3896, df = 1, p-value = 0.03616
alternative hypothesis: two.sided
5 percent confidence interval:
 0.0009506696 0.0010493304
sample estimates:
prop 1 prop 2 
 0.006  0.005 

> pwr.p.test(h = 0.001, power = .8, sig.level = .01)

     proportion power calculation for binomial distribution (arcsine transformation) 

              h = 0.001
              n = 11678968
      sig.level = 0.01
          power = 0.8
    alternative = two.sided