I remember the first invoice I sent to a client when I started reselling LLM API access. I'd quoted them $0.012 per 1K tokens, the upstream cost was $0.010, and I thought I'd nailed it. After payment processing, support time, and a couple of failed charges, my actual margin worked out to about 12%. That's brutal. After fees, taxes, and the occasional refund, I was netting closer to 8%. I was essentially working as a very expensive billing department for someone else's infrastructure.

That was two years ago. Today, my blended gross margin across the reseller book sits at 35%, and on the top-performing accounts it's pushing 42%. The jump didn't come from finding a cheaper upstream provider or some clever arbitrage trick. It came from rebuilding how I price, package, and contract. This article is the playbook I wish someone had handed me in month one.

Key Takeaways

• Most LLM API resellers launch at 10–15% gross margin because they price per-unit and absorb every cost category in silence.
• Tiered packaging (starter, growth, scale) plus negotiated volume commitments can move margins from 12% to 30–35% within one quarter.
• Contract structure matters as much as headline rate — auto-recharge penalties, support SLAs, and minimum commitments are where the real margin is hidden or recovered.
• Pairing a reseller business with an affiliate program (15% first-order, 8% recurring) creates a parallel revenue stream that subsidizes customer acquisition while you learn the market.

Why Most Resellers Start at 12% (and Why That's a Trap)

The default path for new resellers is the same almost everywhere. You sign up for a provider's partner program, you get access to a markup-friendly rate card, you add a small percentage on top, and you start selling. Simple. The problem is that this approach treats your reseller business like a passthrough utility instead of a real business.

Here's what's actually eating your margin in that model:

• Payment processing — 2.9% + 30¢ on every Stripe charge, 3.5% on PayPal, more on international cards. On a $200 invoice, that's $6 to $7 gone before you've paid your upstream cost.
• Failed payments and dunning — Industry data suggests 4–9% of recurring card charges fail on first attempt. Every retry costs you. Every refund costs you more.
• Support time — A single confused client on a Sunday afternoon can burn 90 minutes of your evening. At even a modest $50/hour value of your time, that's $75 against a thin markup.
• Currency conversion — If you're billing in USD and paying in USD you're fine. If clients pay in EUR, GBP, or BRL, the FX spread silently chops 1–3% off the top.
• Idle capacity — Pre-purchased credits that don't get used in a billing window are pure margin loss.

Add it all up and that 20% headline markup you thought you were making is closer to 10–12% by the time it lands in your bank. That's why so many reseller operations churn out in the first six months. The unit economics simply don't work at low scale.

Token-Tier Pricing: The First Major Lever

The single biggest mistake I see is selling one flat rate per unit. Whether your underlying cost is low, medium, or high per call, the market doesn't actually value a unit of LLM output the same way across use cases. A startup running a customer support summarizer will burn 50 million units a month and treat each one as nearly disposable. A law firm using the API to draft contracts treats every unit as high-stakes output worth paying a premium for.

Token-tier pricing solves this by mapping price to value rather than to cost.

How I Structure the Three Tiers

• Starter Tier — Marketed to indie developers, hobbyists, and small experiments. Price is set just above upstream cost with a thin margin (8–12%), but the goal here isn't profit — it's volume and learning. These customers become your feedback loop and your case studies.
• Growth Tier — The workhorse tier for funded startups and product teams running real workloads. Margin here is 22–28%, justified by usage analytics dashboards, priority queueing, and human-on-call support during business hours.
• Scale Tier — Custom contracts for high-volume customers. Margin is 30–38%, but you trade headline rate for committed volume, longer terms (12 or 24 months), and annual prepayment discounts.

The result: instead of averaging 12% across every customer, you blend a low-margin acquisition tier with high-margin retention tiers. My own book now averages around 35% gross margin because the Growth and Scale tiers carry the unit economics, while Starter customers fill the funnel at acceptable cost.

Usage Tiers and Commitment Ladders

Pricing per unit is a starting point, not a ceiling. The real money in reselling is in usage tiers — pricing bands that reward customers for committing to more.

Here's the framework I use, with real numbers from my own contracts:

• 0–1M units/month: Published rate, no discount. This is your catalog price.
• 1M–10M units/month: 12% off published rate, billed monthly.
• 10M–50M units/month: 22% off published rate, quarterly billing with net-30 terms.
• 50M+ units/month: Custom pricing, typically 30–35% off published, with annual prepay option that adds another 5% off in exchange for cash upfront.

The magic of usage tiers is that the discounts look generous to the customer, but your cost basis also drops at higher volumes because your upstream provider gives you better rates too. You're not absorbing the discount — you're sharing the savings you negotiated.

One important note: don't advertise your usage tiers as a discount ladder. Frame them as commitment rewards. Customers psychologically accept paying more when they feel they're climbing a status ladder, and they tolerate bigger invoices when they feel they're unlocking a privilege rather than haggling for a markdown.

Contract Negotiation Tactics That Move Margin 5–10 Points

Once you have prospects at the table, the contract is where the rest of the margin lives or dies. These are the tactics I've used successfully, with specific phrasing where it helps.

1. Annual Prepay Discount

Offer 8–12% off the monthly rate in exchange for a single annual payment. You get cash upfront, you eliminate collections risk, and the discount is well below the 15–20% your time and risk would otherwise cost. On a $50,000 annual contract, a 10% prepay discount is a $5,000 concession that buys you $50,000 in day-one cash and zero dunning overhead.

2. Auto-Recharge Minimums

For usage-based accounts, require an auto-recharge threshold. If a customer's wallet drops below $200, it tops up to $500 automatically. This eliminates the "I forgot to fund my account" churn problem and keeps your utilization rate high.

3. Overage Caps With Buffer

Hard caps at 110% of committed volume, with overage billed at 1.5x the contracted rate. This protects you from runaway usage on a flat-rate deal and creates a natural conversation point when customers hit the cap ("would you like to move into the next tier?").

4. Support Tiering

Free email support with a 24-hour response window. Paid support at 5% of monthly spend gets you a 4-hour response and a named contact. Almost no one buys it — but the existence of the paid tier makes the free tier feel generous and filters out the customers who would otherwise consume 80% of your support time.

5. Multi-Year Lock-Ins With Rate Escalator

For your largest accounts, a 24-month commitment with a capped annual rate increase (3–5%) is often more valuable than a higher headline rate. Predictability is worth real money, both to you and to the customer's finance team.

Income Calculation: A Realistic Monthly Example

Numbers without a worked example are just theory, so here's what a healthy month looks like at modest scale for a solo operator or small team.

Assume the following book of business after about 9–12 months of operation:

• 8 Starter customers averaging $180/month in revenue (8% margin = $14.40/customer)
• 12 Growth customers averaging $1,400/month in revenue (25% margin = $350/customer)
• 3 Scale customers averaging $6,500/month in revenue (33% margin = $2,145/customer)

Monthly revenue: 8 × $180 + 12 × $1,400 + 3 × $6,500 = $37,440

Monthly gross profit: (8 × $14.40) + (12 × $350) + (3 × $2,145) = $10,651

That's a blended gross margin of about 28.4% on a relatively modest customer count. Deduct infrastructure tooling (~$300), payment processing (~$1,100), and a part-time contractor for support (~$1,800), and you're at roughly $7,450 in net operating profit before tax on a solo operation.

Push the customer mix toward more Scale accounts and that same operation easily clears $15K/month. Add a second product line, a usage analytics layer that customers pay extra for, and a marketplace for prompt templates, and you're looking at a real business, not a side hustle.

And here's the part most resellers overlook: recurring revenue compounds. If you net even five new customers per month and your monthly churn is 3%, your book grows by roughly 30–40 customers per year without any improvement in close rate. After three years, the same operation is doing six figures monthly with the same overhead structure.

The Affiliate Layer: A Parallel Income Stream

Most resellers focus entirely on their own customer relationships and miss an obvious adjacent opportunity — recommending providers to people who aren't ready to become customers yet, but are a perfect fit to be referred.

This is where affiliate programs become powerful. A well-structured program pays you for sending customers to a platform, and the right programs pay recurring commissions, not just one-time bounties. That means a single referral can pay you for months or years, often as long as the customer stays active.

For example, Global API runs an affiliate program structured around exactly this model. Standard referrals earn 15% on the first order and 8% recurring on subsequent usage. Premium tier partners, who can white-label the platform and resell under their own brand, earn 10% recurring on top of the markup they set themselves. With access to 150+ AI models through a single integration, the platform handles the messy multi-provider orchestration problem that would otherwise eat weeks of engineering time.

For a solo developer or small agency, the math on a modest affiliate effort is genuinely attractive. Refer just 10 active customers spending an average of $400/month, and the recurring 8% on that book is $320/month passive. Double it, and you're at $640/month for a few hours of content and community work. It's not retirement money, but it directly subsidizes the customer acquisition cost of your own reseller business, which is the real unlock.

Common Margin Killers I Watch For

After running this kind of operation for a while, you start recognizing the patterns that destroy margin. A few of the worst offenders:

• Over-customizing for one client. If a single account is generating 25% of your revenue and asking for custom features, you don't have a customer — you have an employer. Charge accordingly or diversify.
• Letting customers drift into Starter-tier pricing on Growth-tier workloads. Audit your accounts quarterly. Anyone using 3x their tier's intended range should be migrated up, not silently subsidized.
• Ignoring FX and payment geography. A few percentage points of hidden spread on international cards can quietly halve your effective margin.
• Bundling free support into every plan. Support should be a line item, even if the line is "free with 24-hour response