Every transaction costs money. The question is how much.

Most payment infrastructure teams guess. They spin up a node, pay the cloud bill, and divide by transaction count. That is not economics. That is accountancy. Real economics requires understanding fixed costs, variable costs, and scale.

You need to know these things. At what transaction volume does running your own infrastructure beat using Infura? At what volume do you add a second region? At what volume do you optimize the chain selection to shift traffic to cheaper routes? These questions require numbers. Build your model.

Running blockchain payment infrastructure has three cost categories. Fixed infrastructure costs do not change if you process one transaction or one million. A Geth node on AWS runs on c5.2xlarge or c6i.3xlarge, approximately 600-900 USD per month. Storage is 2-4TB of NVMe at approximately 50-100 USD per month. Bandwidth is harder to pin down but assume 20-50GB egress daily at 0.02 USD per GB, so 400-1000 USD monthly. Total for a single Geth node, approximately 1500-2000 USD per month. A Solana validator is cheaper to run (smaller state) but more demanding on network (higher bandwidth requirement). Approximate cost is 800-1500 USD per month. A bridge validator is lighter, so approximately 300-600 USD per month.

Variable costs per transaction include RPC calls to third-party services, bridge fees, gas fees on other chains, and data storage growth. If you do not run your own Ethereum node and use Infura or Alchemy, you pay per call. Infura free tier is 100k requests per day. Paid is 0.003-0.01 USD per request depending on call type. A transaction typically requires 3-5 RPC calls, so approximately 0.01-0.05 USD per transaction. If you run your own node, RPC calls are free but you pay the fixed node cost.

Bridge fees vary wildly. Stargate charges 0.05% of transaction value. LayerZero charges 0.04-0.1% depending on route. A $10,000 transaction pays 50-100 USD in bridge fees. A $100 transaction pays 50 cents to 1 USD.

Gas fees vary by chain and network congestion. Ethereum gas can be 5-50 USD per transaction depending on the hour. Polygon is typically 0.10-0.50 USD. Solana is effectively free.

Self-Hosted vs Third-Party Breakeven

Assume you settle $1 million per month for your customer base.

Option A involves using Infura. RPC cost is 0.01 USD per transaction. If the average transaction value is 10,000 USD, then you are processing 100 transactions for $1M. Cost is 100 multiplied by 0.01 equals 1 USD. But realistically, you process much higher volume of small transactions. If the average is 1,000 USD, you process 1,000 transactions. Cost is 1,000 multiplied by 0.01 equals 10 USD. Cost per dollar settled equals 0.00001 USD (for 1,000 tx/month).

Option B involves running your own node. Fixed cost is 2,000 USD per month. Same transaction volume of 1,000 per month. Cost per dollar settled equals 0.002 USD (for 1,000 tx/month).

This looks like Infura wins by 200x. It does not because you are not accounting for other costs.

Now assume your volume grows to 100,000 transactions per month.

Option A involves Infura remaining 0.01 USD per transaction. Cost is 1,000 USD. Cost per dollar settled equals 0.001 USD.

Option B involves your own node staying at 2,000 USD fixed. Cost per dollar settled equals 0.00002 USD.

At 100x volume, the math flips. Your own node is 50x cheaper.

The breakeven is at approximately 5,000-10,000 transactions per month. Below that, use Infura. Above that, run your own nodes.

But this analysis is missing multi-chain complexity. You probably do not run just Ethereum. You might run Ethereum, Polygon, Arbitrum, Optimism, Solana, and Avalanche. Different costs per chain.

Ethereum. Yes, run your own. Volume is high enough to justify the 2,000 USD per month fixed cost.

Polygon. Maybe. It depends on your volume. If you settle 100M per month on Polygon, run your own. If you settle 10M, use Infura.

Solana. Probably use Infura (or Helius, which specializes in Solana). The fixed cost is lower than Ethereum but the volume is often not there to justify it.

Avalanche. Use Infura unless you have very high volume.

This is a decision tree. Calculate it for your actual volume.

Hidden Costs and Break-Even Model

Storage growth compounds. A Geth node starts at 1TB. After one year, it is 2TB. After two years, 3TB. Your instance is now too small. You upgrade. That is a 30% cost increase. Then 50% increase after the third year.

Bandwidth is often free within a region but expensive across regions. If you have customers in Europe and Asia, you pay egress from one region to another. Your 500 USD monthly bandwidth cost in a single region becomes 2,000 USD across three regions. This is a 4x multiplier that sneaks into the budget.

DevOps labor scales with complexity. When you run one chain on one region, you need 0.2 engineers on-call. When you run six chains across three regions with cross-chain bridges, you need 1 engineer on-call full-time, plus support from senior ops. That is 150k-250k USD per year.

Use a spreadsheet. Create columns for each chain you run. For each chain, you need to track fixed cost per month, variable cost per transaction, transaction volume forecast, total monthly cost, and cost per transaction.

Iterate. Change the forecast up or down by 50%. See where your economics change. If volume goes down 50%, do you still run your own node or switch to Infura?

Create a second section for multi-region. If you add a second region, what is the incremental cost? Fixed cost doubles for each chain. RPC call costs might decrease. Bridge fees might increase.

Create a third section for R&D and operations. How many engineers per region? How much engineering time per new customer?

Calculate total cost of goods sold per dollar of transaction volume. At 100M per month in volume, what is your COGS? At 1B per month, what is it? This is the line that matters to your P&L.

Compare against your price. If you charge customers 0.1% of transaction value and your COGS is 0.08%, your margin is 0.02%. That is thin. If it is 0.005% COGS, your margin is 0.095%. That is healthy.

Your goal is to reach a point where COGS per transaction is low enough that you can compete on price and maintain 70-80% gross margin. For most payment infrastructure, breakeven is at approximately 500M-1B per month in settled volume. Before that, you operate at loss. After it, you are profitable.