AI Token Price Index Drops 20% From Peak; Crusoe in Talks for $3B Raise

Two signals from the AI monetization and infrastructure markets: token prices are showing a pricing power test as Crusoe seeks $3 billion at three times its last valuation.

📊 AI Token Price Index Falls 20% From May Peak, Raising Pricing Power Questions

Decoded: The Silicon Data LLM Token Expenditure Index — the clearest real-time gauge of what users pay for AI tokens — is down almost 20% from a high in May after nearly doubling since its inception in December 2025, Bloomberg reported July 3. The index decline is fueling a debate about whether AI companies are losing pricing power with cost-sensitive enterprise buyers, or simply reflecting a demand shift toward cheaper open-weight models. Louis Navellier, a veteran investor, cited "increasing reports that users of AI solutions, priced in tokens, are having to restrain unlimited use due to high costs." Separately, OpenAI is reportedly leaning toward pushing its IPO to 2027, which the market reads as a sign profitability remains a challenge. Allianz Research calculates there is a 46% growth gap between AI investment and actual AI-related sales — worse than the 32% divergence recorded during the 2001 telecom bust. The index's downward trend has paused over the past week; Silicon Data, which built the gauge, warns against reading it solely as a price tag, calling it "a proxy for marginal willingness to pay." (Bloomberg, Business Times, July 3, 2026)

Why it matters: The Silicon Data index is one of the few real-time signals that connects the $700 billion-plus annual AI capex buildout to actual revenue realization. A 20% drop from the May high, coming as Allianz Research documents a 46% gap between AI investment and sales, moves the conversation from "when does AI revenue arrive" to "at what price." For investors in Nvidia (NVDA), memory producers, and data center operators, the bull case rests on continued growth in total AI spend — which has held even as per-token prices have compressed more than 90% since 2023. But if the index reflects genuine softening in enterprise willingness to pay rather than just cheaper model substitution, it introduces margin pressure across AI service providers before they have generated returns on the current capex cycle. OpenAI's reported 2027 IPO delay is the most visible symptom: the company cannot price a public offering until revenue growth and profitability timelines are credible.

💰 Crusoe in Talks to Raise $3B in Round That May Triple Valuation to $30B

Decoded: Crusoe, an AI data center startup with contracts to supply GPU compute to Meta Platforms and other large AI customers, is in active talks to raise approximately $3 billion in a funding round that could triple the company's valuation to around $30 billion, Bloomberg reported July 2, citing people familiar with the situation. Crusoe was founded to monetize stranded energy by co-locating data centers with natural gas flaring sites, but has since expanded to build conventional AI data center campuses. The company's customer base includes hyperscalers and AI labs that require dedicated GPU cluster capacity outside of the major public cloud providers. Details of the round's investors and timeline were not disclosed. (Bloomberg, Reuters, July 2, 2026)

Why it matters: A $3 billion raise at a $30 billion valuation for Crusoe — which remains private and pre-IPO — illustrates the premium the private market is assigning to dedicated AI compute capacity not routed through AWS, Azure, or Google Cloud. Hyperscalers have sold-out GPU allocations and long lead times; specialized neocloud operators like Crusoe, CoreWeave, and Nebius fill the gap for AI labs and enterprises that need guaranteed compute access on their own terms. Meta's contracts with Crusoe are notable in context: the same company that Bloomberg reported July 1 is developing its own cloud business to sell excess AI compute is simultaneously paying a third-party neocloud for dedicated capacity — signaling that even a $145 billion capex program does not generate enough internal supply for all of Meta's AI workloads. At $30 billion, Crusoe would rank among the most valuable private AI infrastructure companies globally, behind only CoreWeave in the neocloud segment.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team