Conversation

An Ultimate Vision for Serum

2) Well, it depends on the blockchain, uses, etc. Ok; let's say that your goal is to build the most ambitious plausible DeFi ecosystem: one that could end up with a substantial fraction of the world's economy on it.

projectserum.medium.com

It’s easy to get lost in the weeds. Everything has strengths and weaknesses; what matters isn’t rattling them off or proving you know of…

3) Well, the largest companies tend to use ~100k-10m TPS or so. They make something like $1-100B/year of net profit.

4) Ok, so where does that leave us? Well, that's about $10,000 in annual net income per TPS, or about $0.0003 per transaction.

5) So, if your blockchain wants to be theoretically able to house a top company without being a huge drag on its business model, it should maybe be targeting something like $0.0001 per transaction in gas costs.

6) How doable is that? Well, right now, fees on Solana are ~0.000005 SOL, which is roughly..... $0.000125.

7) How will that change over time? I don't know! Here are some thoughts, though.

8) First, what if demand outstrips supply for transactions by a massive amount? Well, then we just get auctions, like what Ethereum has, and prices could diverge upwards. And if supply outstrips demand?

7:21 PM · Jun 22, 2021Twitter Web App

9) Over time, nodes will increase in TPS, and maybe cost too. Which increases faster? TPS, probably. As computers get more efficient, so do Solana nodes. So the cost per tx should go down over time if demand doesn't keep up.

Replying to

10) On the other hand, more validators --> greater compute cost per tx. Right now, Solana has ~50k max TPS and ~600 validators. If it's going to grow to ~10m TPS and ~10k (?) validators, that would be: --500x growth in TPS --20x growth in # validators --??? change in cost/node

11) So theoretically cost per transaction might go down by a factor of ~25, though possibly some of that will be counteracted by other factors. Which is to say: idk, it kinda seems like Solana is on path to be cost efficient for huge companies over the next ~5 years?