The concentrations of utilisable energy per unit of price is important information to allow choices to be made between feedstuffs for practical dietary formulations. There has been a long lasting debate as to whether net energy (NE) or metabolisable energy (ME) is preferable. Only one large data set of empirically determined net energy concentrations (NEp) of feedstuffs for poultry has been published. Although the ME system for describing the utilisable energy concentrations of feedstuffs is now widely used, different metabolised nutrients (protein, fat and carbohydrate) are utilised with different relative efficiencies for energy deposition. More recently, methods of predicting the NE of feeds from equations have been suggested. These include information on nutrient digestibility with or without determined ME values with corrections according to the efficiency by which the nutrients are utilised. The objectives of the present study were, firstly, to evaluate the relationship between determined apparent metabolisable energy (AME) and determined NEp concentrations for a range of poultry feedstuffs and, secondly, the relationship between four different NE prediction equations and determined NEp concentrations was evaluated using the same data set. There was a positive linear regression (p < 0.001) between the determined AME and NEp concentrations in a data set of 62 feedstuffs with a range of AME from 0.1 to 20.5 MJ/kg. This regression equation accounted for 92.8% of the variance in NEp. Each 1.0 MJ increase in AME gave a 0.69 MJ increase in NEp. However, this data set had a large proportion of low energy feedstuffs that are unlikely to be used in modern proprietary poultry diets. A subset of these data that included 40 feedstuffs with an AME range of 8.0–18.0 MJ/kg was also examined. The regression between determined AME and NEp concentrations in this smaller data set only accounted for 77.7% of the variance in NEp. There was evidence (p < 0.01) that AME overestimated the NEp concentrations of high protein feeds of animal origin compared with cereals, cereal byproducts and high protein vegetable feeds. Three NE prediction equations that require knowledge of the digestibility coefficients for protein, ether extract and carbohydrates all improved the precision of the NEp prediction. However, a fourth prediction equation that requires only AME and crude protein determinations gave an equally precise estimate of NEp.