The AI Credit Resale Economy: How Token Brokers Are Cashing In on Gray-Market Access

The Rise of the Token Broker

In August 2026, independent security researcher Matt Lenhard published a follow-up to his earlier work on the token relay market, revealing a hidden economy: the AI credit resale market. Lenhard's investigation, which gained significant traction on Hacker News (229 points, 89 comments), documented a thriving gray market where unused API credits from startups and enterprises are bought and resold, often at steep discounts.

The market has grown from informal forum swaps to a commercialized ecosystem of dedicated marketplaces, routers, and brokers. Lenhard's research, published on Vectoral, shows that founders are routinely receiving unsolicited emails from brokers offering access to OpenAI, Anthropic, and other major providers at 40-50% off list price. Some brokers even claim daily spending capacities of $100,000, indicating substantial supply.

How the Gray Market Works

The mechanics of the gray market are more complex than simple credit transfers. Lenhard identified three distinct layers: credit marketplaces like AI Credits and AICreditMart, which list credits from major providers; bulk-discount routers like CheapCredits, Tokvana, and Neokens, which present themselves as offering savings through 'bulk pricing'; and informal channels on Telegram and Reddit where individual developers and founders trade credits.

Notably, brokers often do not hand over API keys directly. Instead, they act as proxies, forwarding user requests through a pool of keys they control. This design choice means buyers never have direct access to the underlying provider account, creating a layer of indirection that has significant security implications.

The Pricing Arbitrage

The existence of this market is rooted in a fundamental pricing mismatch. Many startups receive credits through accelerators, promotional programs, or enterprise agreements that offer significant discounts or prepaid allowances. When these credits go unused, they become a liquid asset that can be sold at a discount, undercutting official pricing.

Lenhard's rough estimate puts the total value of credits on offer across these channels at 'tens of millions' of dollars. Discounts range from 30% to a staggering 98% on some listings, with typical savings around 40%. However, the research casts doubt on the legitimacy of some of these discounts, suggesting that some 'bulk pricing' claims are likely covers for acquiring credits through other, less transparent means.

Security Risks: What You're Actually Plugging Into

For developers and startups, the appeal of saving 40% or more on API costs is obvious, especially for those running agentic workloads like Claude Code or GPT-based automation. But the security risks of routing traffic through a broker's proxy are substantial.

Lenhard's research highlights that these proxies sit between the user and the provider, meaning the broker sees all prompts and responses. For sensitive data, this is a critical exposure. The risk escalates for agent harnesses that execute code or access internal systems, as a compromised proxy could inject malicious instructions or exfiltrate data. As one analysis noted, 'for anyone running an agent harness like Claude Code through one of these, the risk goes well past did I overpay.'

Some brokers, like CheapCredits, offer Data Processing Agreements and claim GDPR compliance, but these are largely unverifiable. The opacity of the supply chain means buyers cannot be certain who is actually handling their data or where the credits originated.

The Macro Context: AI Spending and Market Distortion

This gray market emerges against a backdrop of massive AI capital expenditure. Goldman Sachs economists estimate US AI investment will approach $600 billion in 2026, nearly 2% of GDP. However, they argue the actual economic impact is smaller than headline figures suggest, partly because much of the spending flows to imported equipment. This concentration of spending among a few cash-rich firms creates the very surplus of credits that fuels the resale market.

Lenhard's investigation suggests that as the market matures and companies become more cost-conscious, 'crackdowns on this type of abuse probably aren't far behind.' The tokens have become a pseudo-currency, and with enough liquidity, abuse is inevitable.

What This Means for the Industry

The AI credit resale economy is a symptom of the industry's rapid growth and pricing complexity. For startups, the temptation to cut costs is understandable, but the security risks are real. For providers like OpenAI and Anthropic, the gray market represents lost revenue and a potential gateway for fraud, likely prompting stricter enforcement and terms of service.

As the market evolves, the key question is whether the industry will embrace more flexible pricing models or crack down on this shadow economy. For now, the token brokers remain a hidden but significant force in the AI ecosystem, operating in the gaps between official pricing and market demand.