{
  "checked_on": "2026-09-24",
  "publisher": "AI Infra Credit",
  "scope": "Selected public offer descriptions and billing documentation, not executed customer contracts, invoices, realized revenue or provider-wide cash-flow evidence.",
  "sources": [
    {
      "id": "CW-FLEX",
      "url": "https://coreweave.com/blog/how-coreweave-spot-and-flex-reservations-work-and-when-to-use-each",
      "type": "company_product_description",
      "published_on": "2026-03-10",
      "updated_on": null,
      "locator": "Flex Reservations / How Flex works",
      "qualification": "The dated announcement calls Flex a preview; current general availability is not established here."
    },
    {
      "id": "CW-PLANS",
      "url": "https://coreweave.com/coreweave-capacity-plans",
      "type": "company_product_page",
      "published_on": null,
      "updated_on": null,
      "locator": "Capacity-plan comparison and Flex FAQ"
    },
    {
      "id": "CW-TOKENS",
      "url": "https://coreweave.com/products/serverless-inference",
      "type": "company_product_page",
      "published_on": null,
      "updated_on": null,
      "locator": "FAQ: How is Serverless Inference billed?"
    },
    {
      "id": "LAMBDA-BILLING",
      "url": "https://docs.lambda.ai/public-cloud/billing/",
      "type": "provider_operational_documentation",
      "published_on": null,
      "updated_on": null,
      "locator": "ODC instances; 1-Click Clusters; Filesystems"
    },
    {
      "id": "LAMBDA-LIFECYCLE",
      "url": "https://docs.lambda.ai/public-cloud/on-demand/creating-managing-instances/",
      "type": "provider_operational_documentation",
      "published_on": null,
      "updated_on": null,
      "locator": "Terminating / suspending an instance"
    },
    {
      "id": "RP-SERVERLESS",
      "url": "https://docs.runpod.io/serverless/pricing",
      "type": "provider_operational_documentation",
      "published_on": null,
      "updated_on": "2026-07-20",
      "locator": "Compute cost breakdown; worker types"
    },
    {
      "id": "RP-PODS",
      "url": "https://docs.runpod.io/pods/pricing",
      "type": "provider_operational_documentation",
      "published_on": null,
      "updated_on": "2026-07-20",
      "locator": "Savings plans; storage pricing"
    },
    {
      "id": "RP-PREPAID",
      "url": "https://docs.runpod.io/accounts-billing/billing",
      "type": "provider_operational_documentation",
      "published_on": null,
      "updated_on": "2026-07-27",
      "locator": "Credits and balance; low balance behavior; refunds and credits"
    },
    {
      "id": "RP-POSTPAID",
      "url": "https://docs.runpod.io/accounts-billing/post-paid-billing",
      "type": "provider_operational_documentation",
      "published_on": null,
      "updated_on": "2026-09-23",
      "locator": "How it works; standard prepaid behaviors do not apply; invoicing",
      "qualification": "Enterprise accounts only; navigation labels this feature Beta. Updated date is not a verified launch date."
    }
  ],
  "records": [
    {
      "id": "coreweave-flex",
      "provider": "CoreWeave",
      "record_type": "billing_offer",
      "source_ids": [
        "CW-FLEX",
        "CW-PLANS"
      ],
      "meter": "reserved_capacity_time_plus_running_instance_time",
      "facts": "A continuous capacity-holding component is described separately from an incremental usage component.",
      "invoice_rule": null,
      "payment_due_days": null,
      "cash_regime": "not_established",
      "observed_rate": null,
      "model_action": "Add a fixed reservation component; do not multiply all charges by physical GPU activity.",
      "energy_pass_through": null
    },
    {
      "id": "coreweave-serverless-token",
      "provider": "CoreWeave",
      "record_type": "billing_offer",
      "source_ids": [
        "CW-TOKENS"
      ],
      "meter": "tokens",
      "facts": "This Serverless Inference offer is token-priced rather than GPU-hour-priced.",
      "invoice_rule": null,
      "payment_due_days": null,
      "cash_regime": "not_established",
      "observed_rate": null,
      "model_action": "Use actual token categories and prices; throughput and demand need their own bridge to hardware cost.",
      "energy_pass_through": null
    },
    {
      "id": "lambda-odc",
      "provider": "Lambda",
      "record_type": "billing_offer",
      "source_ids": [
        "LAMBDA-BILLING",
        "LAMBDA-LIFECYCLE"
      ],
      "meter": "running_instance_time",
      "facts": "Minute-granularity charges span successful launch health checks to termination, including idle running time; invoices cover prior-week use.",
      "invoice_rule": "weekly_in_arrears",
      "payment_due_days": null,
      "cash_regime": "invoice_timing_disclosed_settlement_not_observed",
      "observed_rate": null,
      "model_action": "Separate resource lifecycle from kernels executing and from invoice settlement.",
      "energy_pass_through": null
    },
    {
      "id": "lambda-1cc",
      "provider": "Lambda",
      "record_type": "billing_offer",
      "source_ids": [
        "LAMBDA-BILLING"
      ],
      "meter": "reserved_gpu_time",
      "facts": "Weekly reservation increments; invoice on approval, payable within ten days of receipt.",
      "invoice_rule": "on_reservation_approval",
      "payment_due_days": 10,
      "cash_regime": "relative_to_service_start_not_established",
      "observed_rate": null,
      "model_action": "Use the reservation start and actual invoice/payment dates; do not presume all are pre-service cash.",
      "energy_pass_through": null
    },
    {
      "id": "runpod-serverless",
      "provider": "Runpod",
      "record_type": "billing_offer",
      "source_ids": [
        "RP-SERVERLESS"
      ],
      "meter": "worker_lifetime_seconds",
      "facts": "Charges run from worker start through full stop, including startup, execution and idle timeout; seconds round upward.",
      "invoice_rule": null,
      "payment_due_days": null,
      "cash_regime": "depends_on_account_regime",
      "observed_rate": null,
      "model_action": "Meter non-overlapping worker lifetimes, not the sum of overlapping request times or GPU-busy samples.",
      "energy_pass_through": null
    },
    {
      "id": "runpod-pod-savings",
      "provider": "Runpod",
      "record_type": "billing_offer",
      "source_ids": [
        "RP-PODS"
      ],
      "meter": "gpu_compute_term_entitlement",
      "facts": "Three- or six-month upfront compute plans have fixed expiry; storage is separate and can continue on stopped Pods.",
      "invoice_rule": "upfront_term_purchase",
      "payment_due_days": null,
      "cash_regime": "prepaid_term_plan",
      "observed_rate": null,
      "model_action": "Separate compute entitlement, cash payment and surviving storage obligations.",
      "energy_pass_through": null
    },
    {
      "id": "runpod-prepaid-account",
      "provider": "Runpod",
      "record_type": "account_payment_regime",
      "source_ids": [
        "RP-PREPAID"
      ],
      "meter": null,
      "facts": "Customer-funded credits are consumed by resource charges; the public policy describes deposits as non-withdrawable.",
      "invoice_rule": null,
      "payment_due_days": null,
      "cash_regime": "prepaid_credit_wallet",
      "observed_rate": null,
      "model_action": "Track top-ups, credit consumption and service obligations separately; usage debit is not another cash inflow.",
      "energy_pass_through": null
    },
    {
      "id": "runpod-enterprise-postpaid",
      "provider": "Runpod",
      "record_type": "account_payment_regime",
      "source_ids": [
        "RP-POSTPAID"
      ],
      "meter": "contract_specific_resource_meter",
      "facts": "Enterprise usage is invoiced after the period at contracted rates; payment follows contract terms, outside the prepaid balance workflow.",
      "invoice_rule": "in_arrears",
      "payment_due_days": null,
      "cash_regime": "enterprise_contract_postpaid",
      "observed_rate": null,
      "model_action": "Do not apply prepaid zero-balance stopping or top-up logic; actual credit and payment terms remain required.",
      "energy_pass_through": null
    }
  ],
  "cross_checks": [
    "Runpod's Serverless no-upfront-cost wording is a product-pricing description, not proof that an account never prefunds credits.",
    "The enterprise postpaid documentation explicitly distinguishes its accounts from the general prepaid overview.",
    "Lambda filesystem charges can continue without an attached instance; compute shutdown is not a universal cost stop."
  ],
  "missing_evidence": [
    "Negotiated unit rates and provider product mix",
    "Executed cancellation, service-credit and minimum-commitment provisions",
    "Actual customer payment and payment-processor settlement dates",
    "A verified electricity-cost pass-through formula for any selected offer",
    "Issuer balance-sheet treatment, unrestricted cash and actual funding volumes"
  ],
  "model_status": "The prior technology-cashflow model remains a single-rate post-billing-collection experiment. The companion billing-mechanics module illustrates additional billing/cash identities without silently changing its frozen results."
}