I'll clarify a few things on the mining side here.
There are several factors that determine profitability. Bitcoin's price, block rewards/fees, electricity costs, miner efficiency, and bitcoin's difficulty level.
So the dynamics that play out are this:
Bitcoin is designed to try and keep the average block time at around 1 block every 10 minutes. So if blocks are being "discovered" every 9 minutes, that means there are more miners contributing hashing power to the network that results is blocks being found quicker. Bitcoin adjusts the difficulty approximately every 2 weeks (after every 2016 blocks). So less blocks being found as a result of less hashrate on the network means a difficulty adjustment downware to make it easier to find blocks and vice versa.
All things being equal (bitcoin price, electricity costs, miner efficiency, etc) a higher difficulty level means less profitability since the task at hand is becoming more difficult for the equipment you have. You're going to be solving less blocks and so therefore your rewards are getting diminished.
To try and achieve the same profitability you had before the difficulty adjustment, you'd either need to find cheaper electricity or acquire miners that are more efficient and provide more hashrate per unit of electricity consumed. If you don't make any changes to the work you're doing, eventually as the amount of computing power expands (either due to more hashrate being contributed from more efficient miners or more people bringing miners online) your share of the total hashrate diminishes which means you make less profit. It is also important to note that larger mining operations don't "out compete" smaller ones necessarily. There are very slight economies of scale with initial capital costs of mining machines. You can achieve a slight discount from manufacturers if you buy in bulk. You might also be able to find cheaper electricity prices through contract negotiation and location. But in general, miners are computing against the current difficulty level. So if I am profitable at a given electricity price with a given miner machine model, then it doesn't matter if I am using 1000 of those units or just 1, my profit percentage will be the same. So the idea that a small-time miner can't be profitable isn't really true. You can set up a single mining unit and as long as you have access to some cheap electricity, you can still mine bitcoin at a profit.
Price also plays a big factor into things. The higher the price goes, the more profits all miners are receiving. This is what really drives the increase in energy usage by the network. Because right now, a vast majority of the rewards a miner receives per block found comes from the block subsidy. Every block found gives the miners that found it 6.25 bitcoin. Fees that bitcoin users pay to have their transactions confirmed by miners are also contributed as part of the rewards for finding a block, but at the moment those only account for about 2% of the total rewards. So with the price of bitcoin going up, all miners receive higher profits which can result in expansion of their operations and overall expansion of the industry. But every 4 years, the block reward gets cut in half. So instead of 6.25 bitcoins being rewarded for finding a block, only 3.125 bitcoin will be given out. If the price doesn't follow suit, then there is an immediate and impactful cut in profits for miners across the industry. This generally results in some miners that were border line being profitable shutting down [some of] their operations. This reduction in hashrate thus then leads to a lowering of the difficulty level and a new equilibrium is found in the mining industry for the miners that remain.
This decrease in the block rewards is also a shock to the markets since there is less incoming supply of bitcoin and thus a lowering of liquidity. So price does typically respond upwards after these halvings, but it isn't always immediate. Since a vast majority of the rewards comes from this block subsidy, mining bitcoin is essentially minting new bitcoins into the market and is why this phase of mining is the capital production phase. A majority of the energy being spent on mining is being spent to generate new bitcoins rather than confirming transactions (fees).
Eventually though, since the price can't go exponential forever, these halvings will result in a higher pressure on miner profitability unless transaction fees can begin to fill in some of that void. If transaction fees don't take up some of that slack, bitcoin mining will continue to use less energy until a new equilibrium is met at which point, bitcoin's energy usage will be a direct correlation to how much the bitcoin network is being used economically. The amount spent on fees will go almost entirely to the cost of electricity to confirm it with very little overhead. The market will pay for how much security it needs. If it is a massive market and bitcoin is processing a large number of transactions, then fees will rise and this the security of the network is increased. Miner profits go up which means more energy is being spent to secure the network and thus the network is more secure against attacks. If not many people are using the network, the fees are lower and thus electricity usage goes down and bitcoin's security budget doesn't need to be as high since the network won't face as many attacks. So it theoretically all balances out.
As far as mining hardware lifespan goes, now that the mining and ASIC chip industry has matured, efficiency gains for mining machines has tampered off quite a bit and is more inline with efficiency gains in chipsets for other general computing needs. The lifespan of profitable mining machines is about 4 years at the moment. This is a good website (search for sha-256 to filter for bitcoin miners) that gives a breakdown of miner profitability based on the models available. As you can see there are many miners that were released in 2018 that are still profitable today and there are many 2020 models that are still some of the top profitable machines available.
https://www.asicminervalue.com/If you look year to year, there really is only modest incremental improvements in efficiency at this point. A miner in 2020 that gave you 85Th/s @ 3200W might give you 95Th/s @ 3200W in 2021's next release model. Furthermore, there is quite a large aftermarket now for mining machines. Since a business cares about operational costs and turning a profit, they need to make sure their mining machines are turning as much profit as possible. An individual though has more leeway. An individual who is looking to purchase bitcoin for themselves from any exchange anyway, minus fees paid, could opt instead to mine some bitcoin at a slight cost to themselves as a means of "purchasing" bitcoin for themselves. So many miners in the list that might not be profitable for a business to run actually have a lot of secondary life for individuals that are willing to mine bitcoin at a slight cost as a means of acquiring some. So you can find a lot of those mining machines for really cheap on the secondary markets like eBay. So while they might have a lifespan of about 4 years in the business world, they could see any additional year of life in the aftermarket.
As far as energy usage for each ASIC goes, it really depends on the model, but generally there will be different tiers. You'll have some models that can operate on a 120V circuit and thus can plug into the typical home outlet and run at ~10 amps for a wattage of about 1200. Then there are models that require a 240V circuit that draw much more power at around 3000W. You don't want to have to always upgrade or change the circuit you're operating on when you swap out mining machines, so you'll generally upgrade from unit to unit with similar energy demands.
It seems like Moore's law might be slowing down a bit and the supply chain issues also seemed to slow down chip gains, this leads to an additional ceiling for the overall energy consumption of the network. It just takes time to produce miners and bring them into production and there just isn't a way to exponentially expand a now mature industry.
Also, as I mentioned in one of my earlier posts, the mining industry is much more decentralized now than it has ever been. There are now many more mining pools available and the mining entities within those pools has grown significantly over the years. Hashrate will certainly always be dominated by the large corporate miners, but there are simply more corporate miners mining bitcoin these days than there was in the past.
EDIT: I see maizefolk responded with some of the same points I made albeit much more concise. :)
I also just wanted to add that the amount of computation power required to confirmed blocks is uncorrelated to the number of transactions being processed. A block could contain 1 transaction or 1000 and it will required the same amount of electricity to "find" either block. This is why energy calculations on a "per transaction" basis don't make much sense here.