The Anthropic Data Center Gambit: 80 LOIs, Zero Verification

Seventy to eighty letters of intent. That is the number circulating in infrastructure circles. The claim: Anthropic is locking down data center capacity at scale. The narrative: the AI startup is moving from cloud tenant to infrastructure owner, a strategic pivot to compete with OpenAI and Google. The crypto media, specifically Crypto Briefing, picked it up and ran with it. But anyone who has audited a whitepaper knows that letters of intent are not contracts. They are non-binding expressions of interest. They are the digital equivalent of a handshake before the due diligence begins. And in a market where capital is expensive and execution is everything, the gap between an LOI and a signed lease is where the real story lives. Let me be clear: I am not disputing that Anthropic needs more compute. The company’s flagship model, Claude, demands inference at scale, and its enterprise push requires low-latency, geographically distributed capacity. The data center industry is already benefiting from this demand. But the numbers—70 to 80 LOIs—are being presented as a signal of strength. In my experience, starting with the Neo whitepaper audit in 2017, I learned that numbers without technical verification are noise. A project can claim anything. The ledger does not forgive. Follow the coins, not the claims. So what do these LOIs actually mean? Let’s dissect. First, the context. Crypto Briefing is a publication that covers cryptocurrencies and emerging tech. Its sources are often unnamed, and its editorial bias leans toward hype. The article itself contains no technical details: no mention of total megawatt capacity, no chip supplier agreements, no geographical distribution. It is a single data point wrapped in a narrative. The hook is the number, and the hook is designed to attract attention. For a cold dissector, the lack of corroborating evidence is a red flag. I have seen this pattern before—in 2020, when Curve Finance’s stableswap invariant was celebrated for its elegance, but formal verification revealed rounding errors under volatility. The hype was ahead of the math. The same applies here. Second, the core analysis. Let’s assume the LOI count is accurate. What does it imply? If each LOI covers a typical 10-20 MW data center, the total capacity ranges from 700 to 1,600 MW. That is substantial—equivalent to one or two hyperscale campuses or a dozen medium facilities. This suggests Anthropic is planning a distributed inference network, not a single monolithic cluster. That is a rational architectural choice for latency-sensitive applications. But the key question is: how many of these LOIs will convert to signed contracts? Industry averages for LOI-to-lease conversion range from 30% to 50%. That means Anthropic might secure only 21 to 40 facilities. The rest are negotiation leverage or abandoned options. The market is pricing in 70 as a certainty, but the math says otherwise. Furthermore, the capital expenditure implied by this scale is enormous. A 1,000 MW data center build-out, including land, construction, power infrastructure, and GPU procurement, could cost $10 billion or more over five years. Anthropic is not yet profitable. Its revenue is growing, but its burn rate is accelerating. In 2022, I tracked LUNA’s supply dynamics and documented the precise sequence of its collapse. That taught me that complexity in financial engineering often masks fraud. Here, the complexity is in the capital structure. LOIs are not cash. They are promises that require future funding. If Anthropic’s next fundraising round fails to meet expectations, the LOIs become liabilities, not assets. The company could be forced to renegotiate or walk away, damaging its reputation with suppliers. Third, the contrarian angle. What did the bulls get right? The strategic shift is real. Anthropic is moving from renting capacity on AWS and GCP to securing long-term, dedicated infrastructure. This is a necessary evolution for any AI company that wants to control its own destiny. It reduces reliance on cloud providers, improves cost predictability, and enables customized hardware configurations. The LoI strategy also gives Anthropic negotiating power with data center operators, who are competing for anchor tenants. If the company executes well, it could lock in favorable terms and build a moat against competitors. I have seen similar moves in the crypto space—when Bitcoin ETF custodians like Coinbase and Fidelity upgraded their multi-signature architectures in 2024, they reduced single points of failure. That was a genuine improvement, even if the initial hype overstated the security gains. Similarly, Anthropic’s infrastructure push is a genuine step forward, but the magnitude is uncertain. However, the bulls overlook the execution risk. Operating a distributed network of data centers requires expertise in hardware, networking, and power management. Anthropic is a software company, not a colocation provider. It can hire talent, but that takes time and money. Meanwhile, OpenAI and Google already have massive infrastructure footprints. OpenAI has Microsoft’s Azure cloud at its disposal, with guaranteed capacity. Google designs its own TPUs and operates its own data centers. Anthropic is playing catch-up, and the LOIs are a signal of ambition, not of capability. Code is law. Logic is lethal. The logic here is that intent without execution is a liability. Fourth, the investment implications. For the crypto and tech community, this story is a double-edged sword. On one hand, it validates the thesis that AI infrastructure is a growth sector. Data center REITs, GPU manufacturers, and power utilities could benefit from the capital wave. On the other hand, it amplifies the risk of a capital bubble. When a pre-profit company signs 80 LOIs, it creates a demand signal that could distort pricing across the supply chain. If the bubble bursts, the fallout will cascade. I have seen this before—in 2020, when DeFi protocols promised sustainable yields, the data showed otherwise. The speculative frenzy led to massive losses. The same pattern could emerge in AI infrastructure if investors ignore the fundamentals. Fifth, the compliance and security angle. The LOIs do not address data sovereignty, carbon footprint, or physical security. Anthropic has positioned itself as a safety-first AI company, yet its infrastructure strategy is opaque. Where are these data centers located? Are they powered by renewable energy? What is the disaster recovery plan? These questions matter. In my 2024 Bitcoin ETF audit, I found that custody solutions had residual single points of failure in key management. The same scrutiny applies here. Verification precedes trust. The ledger does not forgive. Finally, the takeaway. The Anthropic LOI story is a test of market discipline. The number 70-80 is a headline, not a balance sheet. Investors should demand verifiable data: signed contracts, capacity commitments, and financial projections. Until then, treat the LOIs as what they are—non-binding expressions of interest. The market is already pricing in an outcome that may not materialize. I have spent 25 years watching the intersection of technology and finance. The patterns repeat. The hype always precedes the hangover. Follow the coins, not the claims. Sanity checks the chain. In this case, the chain is the supply chain. Audit everything. Trust nothing. The only thing that matters is what gets delivered, not what gets promised.