Abstract: The circulant embedding method (CEM) is introduced to fast generate fractional Gaussian noise signals exactly. Fractional Brownian motion can be synthesized through summation of fractional Gaussian noise. Hence it can be applied to the pricing of financial derivatives via Monte Carlo simulation. Empirical result of pricing a warrant (Guodian CWB1) in the Chinese stock market for 100 trading days shows that the described method outperforms more accuracy than that based on standard Brownian motion.