CompuX Revenue Model: How We Make Money (Transparently)

We believe transparency builds trust. When you commit capital, compute budgets, or idle GPU capacity to a platform, you deserve to know exactly how that platform earns its revenue. This page explains every fee, every margin, and every incentive that drives the business. No fine print. If something changes, we update this page first.

Key Takeaways:

• Origination + spread model --- Revenue comes from origination fees (1--3%), annual servicing fees (0.5--1%), and a spread on bulk-purchased compute credits. There are no hidden charges.
• Bulk arbitrage funds the multiplier --- Providers offer 15--30% volume discounts. Most of that savings flows to borrowers as bonus credits; a controlled spread stays with the platform.
• Every side wins --- Capital partners earn 9--12% yield. Startups get 25--50% more compute per dollar. Providers fill idle capacity. The platform earns on volume, not on squeezing any single party.
• Network effects compound value --- More providers mean better bulk rates, which attract more startups, which generate more origination, which attract more providers. The flywheel is the moat.

Where Does the Multiplier Come From?

The headline promise---turn a financing facility into 25-50% more compute---sounds too good to be an honest margin. So let us explain the math.

Cloud compute pricing has a well-known gap between on-demand rates and committed-use or bulk-purchase rates. AWS Reserved Instances, for example, can be 40--60% cheaper than on-demand. GCP Committed Use Discounts range from 20--57%. Independent GPU providers such as Lambda, CoreWeave, and Voltage Park offer even steeper discounts for long-term or high-volume commitments, often in the 15--30% range depending on contract length and capacity tier.

The platform aggregates demand from dozens of startups and negotiates bulk purchasing agreements with providers across the network. Because the aggregated volume exceeds what any single startup could commit to, the negotiated rates are substantially lower than what an individual buyer would get.

Here is where the multiplier materializes. Suppose a provider offers a 25% discount on a $1M bulk purchase. That capital now buys $1.333M worth of compute at on-demand equivalent pricing. After the platform retains its spread (discussed below), the borrower receives somewhere between $1.25M and $1.30M in usable credits. The exact figure depends on provider mix, contract terms, and facility size. You can model your own scenario using the interactive calculator at calc.compux.net.

This is not financial alchemy. It is purchasing arbitrage---the same mechanism that wholesale clubs, group purchasing organizations, and corporate travel agencies have used for decades. The difference is that it is applied to AI compute, where the gap between retail and bulk pricing is unusually wide because providers have high fixed costs and strong incentives to fill idle capacity.

Fee Structure

Three revenue streams, each disclosed upfront before any agreement is signed.

Origination Fee: 1--3% of Facility Size

Charged once when a financing facility is established. Covers underwriting, provider negotiation, credit structuring, and legal documentation. The percentage depends on facility complexity: a straightforward single-provider deal is closer to 1%; a multi-provider, multi-currency facility with custom collateral terms is closer to 3%.

For a facility of $1M at 2% origination, that is $20,000. This fee is deducted from the facility at closing, not billed separately.

Servicing Fee: 0.5--1% Annual

Covers ongoing portfolio management: monitoring credit utilization, managing provider relationships, handling credit rebalancing when a startup shifts workloads between providers, and reporting to capital partners. Charged annually on the outstanding facility balance.

For a typical mid-size facility at 0.75% servicing, that is $7,500 per year on each million deployed.

Spread on Compute Credits

This is the margin between the bulk rate negotiated with providers and the credit price offered to borrowers. It varies by provider, volume tier, and contract length. The spread is not published as a fixed percentage because it fluctuates with market conditions, but it is always visible to the borrower in the credit pricing schedule attached to their facility agreement.

To put a rough boundary on it: if a provider offers a 25% bulk discount and the borrower receives a 20% effective discount (the credit multiplier), the platform retains a 5-percentage-point spread. On a standard facility, that translates to approximately $50,000 per million deployed over the credit consumption period.

What We Do Not Charge

No monthly platform fees. No per-API-call surcharges. No markup on compute usage beyond the disclosed spread. No exit fees. If a startup repays early, there is no prepayment penalty.

Why the Math Works for Everyone

A three-sided marketplace only sustains if every participant captures enough value to stay. Here is the breakdown.

Capital partners deploy capital into blockable credit facilities backed by tangible compute commitments. The yield target is 9--12% annually, driven by origination fees, servicing fees, and a share of the credit spread. Because the underlying collateral is compute credits with deterministic value (a GPU-hour is a GPU-hour), default risk is structurally lower than unsecured venture lending.

Startups access compute at 25--50% below retail pricing without diluting equity. For a seed-stage company burning $50K/month on inference, a 30% discount saves $180K per year---roughly the salary of a senior engineer. The financing is non-dilutive, and the credits are blockable, meaning they can be frozen as collateral without interrupting active workloads.

Providers fill idle capacity through guaranteed bulk commitments. GPU utilization rates industry-wide average 30--50%. A provider running at 40% utilization who fills an additional 20% through bulk deals sees dramatic margin improvement because the marginal cost of serving committed capacity is near zero. The provider does not need to discount to retail customers; they discount to a wholesale aggregator, preserving their retail pricing integrity.

The platform earns on volume. Origination fees scale linearly with facility count. The credit spread scales with total compute throughput. Neither revenue stream requires squeezing any single party harder. The incentive is to grow the network, not to extract more from existing participants.

How the Business Scales

Network effects in a three-sided marketplace create a compounding flywheel.

Stage 1: Provider onboarding. Each new provider added to the network increases the diversity of available compute (different GPU types, regions, pricing tiers). This makes the platform more attractive to startups with varied workload requirements.

Stage 2: Startup acquisition. More startups mean higher aggregated demand, which strengthens the negotiating position with providers. Better bulk rates follow.

Stage 3: Capital attraction. A growing portfolio of performing facilities with low default rates attracts more capital partners. More capital means larger and more frequent facilities can be originated.

Stage 4: Rate improvement. As volume grows, provider discounts deepen. A provider offering 15% at $500K volume might offer 25% at $5M volume. Those savings flow partially to borrowers (better multiplier) and partially to platform margin (wider spread on larger deals).

The result is that unit economics improve with scale rather than degrading. Customer acquisition cost per dollar of originated facility decreases. Provider negotiation power increases. Capital partner yield stabilizes as the portfolio diversifies.

This flywheel effect is documented in more detail in the token operator analysis, which maps how Layer 5 operators in the AI value chain capture margin.

Competitive Position: Why Financing Is a Moat

Several companies operate compute marketplaces. Several offer multi-provider API access. A handful provide GPU financing. No competitor currently combines all three: multi-provider access, credit financing with a multiplier, and blockable collateral.

This matters because each capability reinforces the others. Multi-provider access enables better bulk rates. Better bulk rates make the financing multiplier possible. The multiplier attracts startups. Blockable collateral de-risks capital deployment. De-risked capital attracts more lenders. More lenders fund more facilities. More facilities mean more provider volume.

A competitor who offers only marketplace access (like OpenRouter or Together AI) cannot provide the financing multiplier. A competitor who offers only financing (like traditional venture debt) cannot provide multi-provider compute access or blockable credit collateral. Replicating the full stack requires building relationships on all three sides simultaneously, which is a slow and capital-intensive process.

The window to establish this position is finite. As the AI compute market matures, the bulk pricing arbitrage will narrow (though it will not disappear---wholesale-retail gaps persist in every commodity market). The defensible asset is the network: the provider relationships, the capital partner trust, and the startup base. That network is what compounds.

Frequently Asked Questions