Ethereum operates under a permissionless proof-of-stake model, it has not been mineable since the Merge in 2022.
There are obviously large staking pools (including decentralized pools), but there is also a substantial community of people who stake from home on a validating node of their own. A good example is the EthStaker community on Reddit [1].
The compute and bandwidth requirements are very permissible (quad-core CPU, 16-32 GB memory, 2-4 TB SSD, 20/20 Mb/s), and the theoretical size of minimum stake has gone down substantially over time due to Rocket Pool [2].
The minimum staking amount is very close to zero. Of course, directly participating in consensus by having the opportunity of building blocks requires 32 ETH (≈$75k), or 16/8 ETH ($37k/$18k) if you run a pool for other people through e.g. the trustless Rocket Pool system.
The minimum cost to acquire an ASIC so that you can even _contribute_ to the consensus is very high (thousands), and to be profitable you have to have close to free electricity. The ironic thing is, that in PoW the "rich get richer" is abstracted away into a combination of economies of scale, and access to free/very cheap power. Plus all the e-waste and wasted power, whereas PoS runs on e-waste people run in their closet, pulling as little as sub-10 watts.
"Actual transaction costs are estimated around 0.001 cent/tx (at high transaction rates, amortized over billions of transactions, excluding migration costs). Note that this is an early estimate -- and could thus easily be 10x higher." [1]
Transactions costing singular cents [2], and very soon fractions of a cent [3], are already doable on public rollups providing actual control of your own funds, with escape hatches in case the system sequencer falls over [4]. These are, of course, still under heavy development. If privacy is a requirement, there's a zk-rollup for that (or, just use Monero). Not to mention support for social recovery.
This is a very PoS-Ethereum-centric (and L2-scaling-centric) view. There are even cheaper chains that trade L1 decentralization for faster blocktimes, instant finality, and/or much cheaper fees.
Ethereum's clients (execution and consensus) are only really concerned about a sufficiently fast SSD, more specifically in terms of random IO. Any mid-tier SSD should work though: stick it to a Raspberry Pi (or a Rock 5B) and you'll be validating on a low-power arm board. Or grab a used micro-PC/NUC, as most people running home nodes do.
I do overnight oats; in the morning, all you have to do is pop the lid.
* Wholegrain oats (1.5 to 2 dl)
* Greek yogurt (1.5 to 2 dl)
* Milk, skimmed/whole (3 dl)
Then, throw in whatever. A sliced banana or berries, honey, a pinch of vanilla extract. Prep it in the evening, chuck it into the fridge, and it's done in the morning.
Mev-boost, a piece of software that consensus clients delegate their block building to, is currently being used on mainnet. The difference is that there can be multiple block builder _relays_, such as Flashbots, bloXroute, Manifold Finance, and Blocknative.
These relays effectively accept bundles from MEV searchers, which are then sent to the validator running mev-boost as they propose their block.
APY depends on the current validator count, decreasing as more validators come online. Currently, the base APY is around 4 %.
On top of that, you have to factor in luck: block proposals are randomly assigned to validators, and they include a random amount of tips depending on network activity. You could propose a block that nets you 0.01 ETH in tips, or a block with 50 ETH in tips.
On average, I believe the post-Merge APY has been estimated to be ≈6 % (non-cumulative).
Hardware can be anything. Even a RasPi 4 with 8 GB of memory can be made to work. Any old x64 machine with two or more cores works. 16 GB of memory is recommended, as is a 2 TB SSD to minimize down-time. Storage requirements are likely to go down in the medium-term, as e.g. state expiry get implemented.
I’d recommend an Intel NUC. They use laptop CPUs, so electricity consumption is around 5–20 watts. For networking, you’ll need around 5 Mb/s up and down after the initial sync.
Because stable currencies do not exist in the world of cryptos? Dai [1], Rai [2], and USDC [3] are some examples that are all stable, and all dramatically different from each other.
Fees are currently as low as $0.03 on Ethereum layer-2 solutions (!= side-chain) [4]. The fees will further fall as the technology improves, and once L1 gets (data-/dank-)sharding and other L2-scaling oriented upgrades.
You can stake on a Raspberry Pi and an external SSD, if you hate yourself enough. Most home stakers run on micro-PCs like Intel NUCs or Optiplex Micros.
Attestations occur for every validator once every epoch, i.e. once every 6.4 minutes. It is not related to network activity in any way.
The validator that proposes the block with the slashing in it gets a small reward, while the rest is burnt. There are dedicated slashers on the network that observe attestations and block proposals for violations. They can either selfishly keep those observations until they can be included in a block, or broadcast them to the network so whoever is up next for proposal can slash the misbehaving validator. The whistleblowers themselves (altruistic slashers) get no reward, only the validator including it in a block does.
The current estimate for the upgrade can be found at [1]. It's not a regular upgrade that happens at a specific block height, hence the variance. Instead, the consensus is flipped over from PoW to PoS at a specific total difficulty, also known as mining difficulty. One of the reasons is to avoid miner attacks around the merge event.
Because not attesting is against the network's fork-choice rules. If that censoring majority chooses to avoid said block, because it contains transactions that should be censored, they will perform an illegal re-org around the block. Sure, it won't be "illegal" as a majority of the network follows said re-org.
It's an extremely nuanced topic, and it was extensively discussed in the last core-developer call on Thursday. Every single operator has been warned that going against the fork-choice rules could, and ultimately will, result in social mitigations against spec-deviating behavior.
This could be set of socially executed slashings of validators that don't respect the fork-choice rules, or an honest-minority executed hard-fork that effectively causes misbehaving validators to bleed until they start respecting the protocol's rules.
I could be wrong on multiple counts, especially regarding re-orgs, but this is how I've understood the issue.
You are wrong about the majority of validators having to approve a certain transaction for them to be included. Even if 80 % of the network were censoring, those transactions would, on average, make it into every fifth block.
There are obviously large staking pools (including decentralized pools), but there is also a substantial community of people who stake from home on a validating node of their own. A good example is the EthStaker community on Reddit [1].
The compute and bandwidth requirements are very permissible (quad-core CPU, 16-32 GB memory, 2-4 TB SSD, 20/20 Mb/s), and the theoretical size of minimum stake has gone down substantially over time due to Rocket Pool [2].
[1] https://reddit.com/r/ethstaker/
[2] https://rocketpool.net/node-staking/rocket-pool-vs-solo-stak...