I've always felt the calculations about Bitcoin's energy consumption ignore the main cause of consuming any energy: the revenue Bitcoin miners earn. Many of the comparisons mention the hashrate per unit of energy, or improved energy efficiency, but none of that really matters. What matters, is how much money miners have to spend. That money goes to:

• hardware
• electricity
• profit

If hardware gets more efficient, miners will buy more of it. If hardware gets more expensive, it's likely the hardware manufacturer uses more energy to produce it. If they make more profit, they (or other miners) buy more hardware to get a bigger piece of the pie. In the end, the main thing deciding energy consumption is how much money they earn. Bitcoin halvings are a major driving force to reduce miner's income, and thus miner's energy consumption. Until now, Bitcoin's value went up faster than the block rewards went down, but long-term, the halvings will lead to less money to spend. By that time, Bitcoin's energy consumption will depend on transaction fees.
And that got me thinking: long-term, converting to dollars doesn't really work because of inflation. So how about comparing Bitcoin miner's revenue to the oil price? I know it's not directly related, and I know miners mainly choose places with low electricity prices, but the oil price is one thing that's more or less the same worldwide.

Would anyone be interested in a graph, since the beginning of Bitcoin, showing the number of oil barrels earned per day from mining Bitcoin?