AI money math · from Indonesia

Four assumptions the standard advice rests on:

The audience pays with US cards. Half of mine doesn't. The card networks and payment processors that handle Indonesian cards have different fees, different decline rates, different fraud profiles.

$10 is a "no decision" amount. It is in San Francisco. In IDR — roughly 150-160k at current rates — it's not a no-decision amount. It's a real recurring expense.

Stripe just works. It does in 40-something countries. Indonesia isn't one of them for direct accounts. Indonesian solo founders end up routing through Stripe Atlas, US LLCs, or alternative providers.

Infra cost is dollar-stable. It isn't. Every dollar billed by Vercel, Anthropic, Supabase, or OpenAI is one dollar in IDR with a 1-3% card markup on top. Cost as a percentage of revenue is higher when your revenue is in mixed currencies.

The cumulative effect: a strategy that pencils out comfortably for a Bay Area creator can be marginal for an equivalent operator in Indonesia. Not impossible. Just tighter.

The actual options, ranked by setup cost vs. ongoing friction:

Stripe Atlas — register a US LLC, get a Stripe account, run everything through it. ~$500 one-time + ~$120/year. Cleanest for serious operators. Tax filing in two jurisdictions is the ongoing cost.

Wise / Payoneer with local rails — receive payments via Wise Business or Payoneer, convert to IDR. Higher fees per transaction. Easier setup.

Xendit / Midtrans — Indonesian payment gateways. Lower fees on IDR transactions. Higher friction if your audience expects card forms in English.

Lemonsqueezy — handles VAT / sales tax / merchant-of-record for you. Higher transaction fee. Lower compliance burden. The "no-decision" option for most digital products under $10k/month MRR.

For most Indonesian solo founders selling globally, Lemonsqueezy or a similar merchant-of-record is the realistic answer. The Stripe Atlas path has more upside; the merchant-of-record path has less setup friction.

Purchasing power parity is the thing US-based pricing advice systematically ignores. A $10/month subscription costs an American buyer roughly 0.02% of median income. The same $10 costs an Indonesian buyer roughly 0.5% of median income. The buyer's decision math is 25x different even though the dollar amount is identical.

The two ways to handle this:

Localized pricing tiers. Detect country, offer regional pricing. The risk: arbitrage — people use VPNs to get the cheaper tier. The control: tighten the geo check, accept some arbitrage as cost of doing business.

Anchor to one currency, let buyers self-select. Charge in USD, let the international audience that values the product pay the full price. The audience self-selects to "willing to pay USD-priced product."

I'm in the second camp for most things, with one exception: anything aimed at an Indonesian-specific audience gets priced in IDR. The friction of paying in unfamiliar currency outweighs the upside of higher absolute pricing.

A back-of-envelope for a typical AI-product solo-founder stack at minimal scale:

• Vercel Pro: $20/mo (USD) → ~300k IDR after card markup
• Anthropic API for ~1M tokens/day usage: $30-150/mo → 450k–2.2M IDR
• Supabase: $25/mo → ~375k IDR
• Domain + email: ~$15/mo → ~225k IDR
• Stripe / payment processor fees: 2.9% + $0.30 per transaction

The infra-only cost before any tool subscriptions: ~$90-210/mo USD = ~1.3M-3.2M IDR/mo at current rates.

That's a meaningful number for an Indonesian solo developer running an experimental product. If the product takes three months to find any paying users — a normal timeline — you've burned 4-10M IDR on infrastructure with zero revenue. The math is survivable but it's not the "trivial cost" the US-perspective advice implies.

The implication: take the cost-cutting parts of the standard advice seriously. The free tiers, the deferred payment processor signup, the not-pre-buying-the-full-Vercel-pro-until-you-need-it discipline. They matter more from Indonesia than from California.

Not everyone. This is calibrated for:

• Solo builders in SEA building for a global (or partly global) audience
• Working engineers with stable other income who can afford a 6-12 month runway on experiments
• People comfortable with mixed-currency accounting and basic international tax considerations

It doesn't apply to:

• B2B enterprise in Indonesia — that has its own rules, totally different rails
• Pure local consumer apps targeting Indonesian users only — Xendit / Midtrans and IDR-only pricing is correct
• Anyone doing this without an existing income stream to fund the 12-month "find product-market fit" runway

If you're outside these conditions, the advice doesn't fit cleanly. Particularly the price in USD for a global audience recommendation — that only works if you're prepared to spend a quarter or more building before the first paying user shows up.

Three concrete things I wish I'd known earlier:

1. Skip Stripe Atlas at the start. Use Lemonsqueezy or equivalent merchant-of-record. The legal + tax burden of running a US LLC from Indonesia isn't worth it until you're seeing $5k+ MRR consistently.
2. Price the psychology, not the math. $7 USD/month sounds cheap; 100k IDR/month sounds expensive. Same money. Choose the framing that fits the buyer.
3. Track everything in IDR. Even if billing is in USD, your runway is in IDR — that's the currency your rent, food, and groceries are denominated in. Don't let dollar accounting hide the rupiah reality of your runway.

The macro version of the advice: don't import US monetization wholesale. Translate it. The rupiah math is what determines whether you can keep building.