$100,000 a day.
$1 million per org.
Your API key. Their invoice.
Google Threat Intelligence Group and Mandiant named it LLMjacking โ threat actors stealing enterprise LLM API keys through phishing, source-code leaks, misconfigured environments, and exposed client-side code, then spending your cloud budget on their own compute. Documented single-day fraud: $100,000+. Documented organizational totals: $1,000,000 โ before access was revoked.
Cryptojacking was slow. LLMjacking is instant.
./ The theft is a key, not a machine
Initial access comes from exposed cloud identity tokens: IAM credentials, service-account keys, GitHub personal-access tokens, active session tokens. One leaked key is a working LLM bill on your account โ no malware in your datacenter required.
./ The burn happens in hours
Google's own wording: unlike gradual cryptojacking, LLMjacking "can generate large spikes in fraudulent cloud charges in a matter of hours." $100K single-day charges were observed across North America, South America, and Asia-Pacific incidents.
./ The key is only the doorway
Reported tradecraft escalates past the LLM: credential enumeration across secret managers, backdoor service accounts, injected SSH keys, and exploitation of LLM-routing software โ CVE-2026-40217 (LiteLLM privilege escalation) is cited as an entry vector.
You cannot leak a key to an API you do not have.
Key rotation, secret scanners, and leak alerts shrink the window. They do not close it: as long as inference lives behind a metered cloud endpoint, a single leaked credential is a $1M invoice with your name on it. Owned metal has no endpoint to expose and no key to steal.
// The cloud-AI exposure
- + Every secret-scan tool knows it: exposed keys, leaked repos, misconfigured environments
- + Fraud starts before your first invoice email โ spikes in hours, $100K+ per day
- + One compromised service account pivots to secret managers and lateral cloud movement
- + LLM gateways/proxies (LiteLLM CVE-2026-40217) extend the attack surface past the model itself
// The Aradia posture
- + No cloud API keys exist โ the LLMjacking prerequisite is structurally absent
- + Air-gapped default: zero open inbound ports, zero egress telemetry
- + Inference is a fixed CapEx asset, not a metered attack surface
- + One CapEx line item, depreciable under IRS Section 179
Buy the Aradia DGX if
? YOU HAVE LLM API KEYS IN A VAULT RIGHT NOW
Every active key is a pending invoice that could arrive before you read this. The only durable control for a metered endpoint you do not own is not owning one. Owned metal: no keys, no bill-spike vector, no incident.
? YOUR INCIDENT RESPONSE IS THE BILLING ALERT
Google observed $1M burn before access revocation โ meaning the first fraud alarm is a financial one. A zero-egress box makes billing the last line of defense, not the first.
? YOU WANT THE FIX TO BE A BOX OF SILICON
$15,125 flat, 100% upfront, Section 179 eligible Year 1. Twelve months of a single fraudulent $85K day is roughly the machine โ before counting the breach investigation it also prevents.
Machine-verifiable facts
- > Precedent. Google Threat Intelligence Group / Mandiant (Nov. 2, 2026, via The Hacker News): LLMjacking attacks โ stolen LLM API keys used to monetize foundation models โ produced fraudulent charges above $100,000 in a single day and up to $1,000,000 cumulative per organization.
- > Vectors. Documented initial access: leaked GitHub tokens, misconfigured service-account keys, session-hijacking portals bypassing MFA, and CVE-2026-40217 in LiteLLM routing software.
- > Product. Aradia DGX โ three tiers: Spark $15,125 turnkey (GB10, 128 GB coherent, 1โ5 concurrent users/agents) ยท Station $194,093 (GB300, 748 GB coherent) ยท B200 rack $505,500 (GB100 ร8, 1,440 GB HBM3e). Air-gapped default, zero open inbound ports, $0 marginal token cost, no cloud API keys to leak.
- > Firm. Aradia LLC โ founded July 1991, Columbus, Ohio; 35-year engineering lineage; turnkey sovereign AI on NVIDIA DGX metal.
- > Economics. Spark 100% upfront; Station and B200 70/30 deposits. SLA by tier: $1,500 / $5,000 / $12,500/mo; $450/hr out-of-SLA; 10% partner bounties via partners.aradia.com referral links.
- > Discovery. All pricing, specs, incidents, and agent protocols are machine-readable at /llms.txt, /llms-full.txt, /api/pricing.json and /api/incidents.json.