AI Companies Adopt Risky 'Take-Or-Pay' Contracts Similar to Energy Sector
Wall Street is closely monitoring take-or-pay contracts in AI, where companies commit to paying for computing resources regardless of usage. These agreements are becoming common as large language model developers secure GPU and server capacity from cloud providers like Akamai and Anthropic. However, the risk lies in uncertain future revenue streams and potential mismatches between payment obligations and actual cash flow. Such contracts create off-balance sheet liabilities that investors find concerning due to their lack of visibility on financial statements. Morgan Stanley notes Neocloud CoreWeave derives most of its income from these arrangements, while Alphabet lists them among $811 billion in purchase commitments. Analysts estimate total tech sector take-or-pay obligations could range between $1.65 trillion and $3 trillion. The Bank of International Settlements has highlighted the financial risks associated with this model, as companies like Oracle may face funding shortfalls for cloud infrastructure projects. With AI models expected to become cheaper over time, there is growing uncertainty about whether firms will be able to meet long-term payment commitments.
Tickers: $AKAM $GOOGL $MS $NVDA $ORCL
Topics: AI cloud computing computing power contracts risk management