UNAIDS has embraced an ambitious global target for the implementation of treatment for people living with HIV. This 90-90-90 target would mean that, by 2020, 90% of all those living with HIV should know their status, 90% of these would be on treatment and 90% of these would have fully suppressed plasma viral loads. To reach this target in the next five years presents a major logistical challenge. However, the prevalence of HIV varies greatly by risk groups, age, gender, geography and social conditions. For reasons of effectiveness and impact the focus must first be on those who are most likely to be infected with HIV and therefore most likely to infect others. In Kenya the prevalence of HIV in adults varies by two orders of magnitude among the counties. The effective implementation of 90-90-90 will depend on first providing ART where the prevalence of infection is greatest, then to those that are most easily reached in large numbers and finally to the whole population. Here we use routine data from ante-natal clinics and national survey data to assess the variation of the prevalence of HIV among counties in Kenya; we suggest reasons for this variation, and estimate the effectiveness of targeting the role out of ART. The highest prevalence occurs in some of the counties bordering Lake Victoria and these are most in need of ART. These districts in Nyanza Province, account for 31% of all cases in Kenya but make up 10% of the population and cover 1.8% of the land-area. The highest concentrations of HIV cases are in Nairobi and Mombasa which account for a further 18% of all cases in Kenya but make up 12% of the population and cover 0.1% of the land-area. Providing ART in these two cities will be relatively straightforward given their small geographical area.