We show that exogenous technology shocks that increase the speed of trading from microseconds to nanoseconds dramatically increase the order cancellation/execution ratio from 26:1 to 32:1 but do not have any detectable impact on liquidity, price efficiency or trading volume. We find evidence consistent with 'quote stuffing,' which involves submitting an abnormally large number of orders followed immediately by a cancellation to generate order congestion. The stock data are handled by six randomly grouped channels in NASDAQ, and the message flow of a stock can slow down the trading of stocks in the same channel but not stocks in a different channel. We detect an abnormally high level of co-movement of message flow for stocks in the same channel using factor regression and a discontinuity test. Our results suggest that an arms race in speed at the sub-millisecond level is a positional game, where a trader's pay-off depends on her speed relative to other traders. Private benefit then leads to offsetting investments on speed by different high-frequency traders even if there is no social benefit.