ASIC hash engines are specifically optimized for parallel computations of cryptographic hashes and thus a natural environment for mounting brute-force attacks on hash functions. Two fundamental advantages of ASICs over general purpose computers are the area advantage and the energy efficiency. The memory-hard functions approach the problem by reducing the area advantage of ASICs compared to general-purpose computers. Traditionally, memory-hard functions have been analyzed in the (parallel) random oracle model. However, as the memory-hard security game is multi-stage, indifferentiability does not apply and instantiating the random oracle becomes a non-trivial problem. Chen and Tessaro (CRYPTO 2019) considered this issue and showed how random oracles should be instantiated in the context of memory-hard functions. The Bandwidth-Hard functions, introduced by Ren and Devadas (TCC 2017), aim to provide ASIC resistance by reducing the energy advantage of ASICs. In particular, bandwidth-hard functions provide ASIC resistance by guaranteeing high run time energy cost if the available cache is not large enough. Previously, bandwidth-hard functions have been analyzed in the parallel random oracle model. In this work, we show how those random oracles can be instantiated using random permutations in the context of bandwidth-hard functions. Our results are generic and valid for any hard-to-pebble graphs.