The article examines the lack of transparency in AI infrastructure deals, which have recently reached over $500 billion in commitments. Unlike mature infrastructure markets, where contract terms and pricing mechanisms are well defined and verifiable, the AI sector operates in a murky environment. Major deals, such as Nvidia-OpenAI's $100 billion collaboration, are often based on vague press releases and lack enforceable agreements. This absence of standardization leads to a situation where the market struggles to accurately price these commitments. Key examples highlight the issue. Oracle's reported $300 billion figure lacks a named counterparty in its SEC filings, while AMD's $100 billion agreement has some definitive terms but remains missing critical details like payment schedules. The term "gigawatt" in AI can refer to various stages of project completion, from aspirational goals to actual utilization, complicating valuation further. Without clear definitions and observable economic terms, investors are left guessing, which invites speculation and mispricing based on optionality rather than firm commitments. The opacity isn't just a quirk; it's built into how AI deals are structured. Factors like fluid negotiations and sensitive pricing information prevent public disclosure of critical terms. The article argues that while these headline figures can signal intent and coordination among stakeholders, they should be viewed as high-risk options rather than guaranteed investments. This misinterpretation can lead to inflated market expectations, as the real economic implications often remain hidden behind layers of uncertainty.