ASICs are inevitable and a good thing. ASIC-resistance is nonsense. GPU-mining does not favour the small player. Did you try buying a GPU back in 2017? They were non-existent. People chartered Boeing 747s to fly around the world to scoop up GPUs. Anyone close to the supply chain has the advantage.
GPUs are built by two major manufacturers. SHA256d is simple and patent free. Greater opportunties for more manufacturers to enter the game. Asicboost was unfortunate, but manageable. There are less ways to cut corners and look for efficiencies in a simple mechanism. Greater chances to improve the supply chain.
GPU-mining eventually favours secret ASIC optimisations. Grin is a GPU-coin that had some sort of an ASIC roadmap. At least they knew GPUs were only good for the bootstrapping phase. Grin GPU PoW favoured a particular ‘class’ of GPU. Only some cards with a specific spec could mine it initially. GPU coins favour whoever can get closest to the supply chain. GPU coins are also vulnerable to liquid hash markets.
Secret ASIC optimisations on GPU-algorithms are also heavily CAPEX-focused. There is an incentive to keep them secret. Further, once they eventually hit the market, who can afford them? This is all bad for decentralisation.
How about memory-hard or other more complicated algorithms? Power consumption is relative. Once CAPEX costs come down (on the inevitabile ASICs).. electricity will eventually become the dominating metric for profitability again.
Electricity is in abundance worldwide, much of it clean and stranded. Geography is not an issue for bitcoin. Simple/cheap PoW ASICs (not encumbered by patents) offer the best model for acheiving high levels of security and decentralisation. The ecosystem around SHA256d and its supply chain has taken years to build and mature.