The economic coordination layer for AI inference.

Keld is the neutral marketplace and trusted clearinghouse matching enterprise demand with global model capacity. Maximize buyer ROI, capture seller yield, and run production AI on an institutional-grade financial foundation.

Register your interest → Explore the platform
scroll
40%
Average AI cost reduction
100+
Model providers on the exchange
<5ms
Routing overhead
Real-time
Spend visibility & alerts
Enterprises

Absolute ROI.

Lower your effective cost-per-token by routing to the most cost-efficient, high-performance nodes in real time — without sacrificing latency or data security.

  • Effective cost-per-token ↓ optimised
  • Named model guarantees enforced
  • Routing latency < 2ms
Keld
Model Providers

Maximum Yield.

Turn idle capacity into recurring revenue. List your inference supply on a neutral exchange that dynamically clears volume based on live market demand.

  • Idle capacity utilisation ↑ maximised
  • Demand clearing real-time
  • Revenue stream recurring
Observability & Finance Classification

Every token and dollar,
fully accounted for.

Stop treating inference like a black box. Track exactly how tokens are being consumed across your organization down to the specific team, department, and individual. Automatically classify AI spend into clear Capex or Opex buckets for institutional-grade financial reporting.

  • Capex / Opex classification applied automatically at point of inference
  • Granular attribution by team, project, department, model, and job type
  • Real-time budget guardrails with threshold alerts to finance and engineering
  • One-click export to your finance, ERP, and BI stack
AI Observability ● Live
Capex
$18,400
R&D Infrastructure · 67%
Opex
$9,000
Operations · 33%
Engineering
$12,400
Product
$8,100
Data
$4,700
Sales AI
$2,200

Product team nearing monthly budget limit
Model Exchange 100+ providers
Model$/1M tokensLatency
GPT-4o
$5.00 320ms
Claude Sonnet
$3.00 280ms
Llama 3.1 70B
$0.54 190ms
Gemini Flash
$0.10 140ms
Best value for your workload Llama 3.1 70B
Model Exchange

The neutral inference
Marketplace

A completely transparent, conflict-free order-matching engine. Keld executes orders programmatically between buyers and sellers based strictly on time, price, performance, and volume, driving bottom-line efficiency for enterprises and unlocking net-new revenue for infrastructure providers

  • 100+ vetted providers across every major model family
  • Live price and latency comparison, side by side
  • Switch providers with a single config change
  • No lock-in — run multiple providers simultaneously
Deadline

Give Keld a task, a budget, and a time window — it executes as a durable function with automatic retries and model selection, delivering results on completion with no polling required.

Supply Order Management

Turn spare capacity into
a revenue stream.

Model providers gain direct access to Keld's institutional order book. Dynamically manage your inference supply — adjusting price, capacity, and availability in real time as your compute utilisation shifts. Keld's OMS handles clearing, settlement, and demand matching so you can focus on running great infrastructure.

  • Real-time order book with live demand visibility by model and tier
  • Dynamic pricing controls — respond to utilisation changes in seconds
  • Automated clearing and settlement with full transparent audit trail
  • Demand forecasting to plan compute capacity ahead of market spikes
Order Management ● Exchange Live
REVENUE · MTD
$41,280 / $48,900 fcst
DEMAND TREND
↑ +18% vs last month
Model tierDemand$/1MRev/day
Llama 3.1 70B
$0.52 $8,100
GPT-4 tier
$5.10 $2,400
Gemini Pro
$1.20 $3,200
Custom 8B
$0.18 low demand
Compliance Enterprise
Cloud Hosted
Multi-AZ redundancy
99.99% SLA
Auto-scaling
Managed upgrades
Self-Hosted
Air-gapped deployment
Data residency control
Your infrastructure
Full ZDR support
ZDR SOC 2 GDPR SSO / SAML Audit Trail
Security & Compliance

Zero data retention.
Sovereign privacy.

As a neutral clearinghouse, Keld operates under a strict, audited Zero Data Retention (ZDR) architecture. We secure your economic transactions without ever compromising your proprietary intelligence. Your data remains entirely yours, protected by enterprise-grade compliance at every step of execution.

  • Zero Data Retention architecture — audited and independently certified
  • Your prompts and completions are never stored, never logged
  • Data residency and sovereignty controls for every jurisdiction
  • Every cleared prompt with the exact response quality your production applications demand
Teams & Budgets

The control center
for your AI operations.

Provision teams, set budget envelopes, and control who has access to what — all in one place. Provider credentials and API keys are managed centrally by Keld and abstracted away entirely, so your engineers just use the endpoint they're given and never touch a key again.

  • Per-team budget envelopes with real-time utilisation tracking
  • Role-based access control for users, teams, and projects
  • API keys and provider credentials fully abstracted — no spreadsheets, no rotations to manage
  • Audit trail for every request, every change, every team
Teams & Budgets 5 teams · 32 users
ORG BUDGET · MAY
$30,682 / $40,000
76% used · 6 days left in cycle · alert at 70%
PACING
On track — under budget by $3,180
TeamBudget · MTDStatus
E
Engineering
Sara Patel
$12,400 / $15,000
83% used
P
Product
Marco Rossi
$8,100 / $10,000
81% used
M
Marketing
Iris Chen
$4,700 / $5,000
nearing cap
D
Data
Jonas Lim
$3,680 / $6,000
healthy
S
Sales AI
Ana Ortiz
$1,802 / $4,000
healthy
“

Keld. The financial era of AI infrastructure is here.

Keld

Register your interest

Tell us a bit about your team and what you're trying to solve. We'll get back to you within 24 hours.