Ed Zitron: Everything you know about AI investing is wrong

Anthropic’s draft IPO prospectus shows $16.3 billion in first-half revenue against an $80.7 billion net loss, including about $72 billion in non-cash charges on convertible notes, and roughly $450 billion in mostly non-cancelable cloud commitments. EZPR CEO Ed Zitron argues that OpenAI and Anthropic are load-bearing for the broader market, says no ratings agency could honestly rate them above triple-C, and contends they should be allowed to fail.

Analysis

The key market question is not whether the reported loss is large, but how much represents cash burn versus accounting remeasurement—and whether Anthropic can fund its commitments without repeatedly tapping public or private capital. The non-cash note charge should not be treated as equivalent to operating cash outflow; investors need the cash-flow statement, note terms, and liquidity runway before drawing a solvency conclusion. Conversely, long-dated cloud commitments can turn a growth slowdown into a fixed-cost and financing problem if they are difficult to renegotiate.

This creates a two-sided exposure for cloud providers: contracted demand may support near-term infrastructure utilization, but concentrated customer obligations become a credit and revenue-quality risk if an AI lab retrenches. The prospectus could also reset investor scrutiny of OpenAI and other frontier-model developers, shifting focus from revenue growth to cash conversion, dilution, and committed compute costs. The article’s ratings-agency claim is opinion, not evidence of an actual downgrade or imminent default.

Near term, the story is a sentiment and IPO-pricing risk, not by itself a confirmed credit event. Over 1–3 months, prospectus updates and disclosure of cash burn, commitment counterparties, and financing terms are the catalysts. Over 6–18 months, the structural test is whether enterprise AI revenue scales fast enough to absorb compute commitments. The bearish thesis weakens if audited disclosures show ample liquidity, manageable termination/repricing provisions, and improving cash economics.

Key Decisions for Investors

• Do not trade the reported net loss as a cash-loss proxy. Before underwriting default risk, verify operating cash burn, unrestricted cash, convertible-note conversion/settlement terms, and the timing and cancellation provisions of cloud commitments.
• Treat any Anthropic IPO filing or pricing as a catalyst for repricing AI risk, not proof of distress. If disclosures show heavy fixed commitments and weak cash conversion, consider limited-premium put spreads on broad technology or semiconductor exposure over the 1–3 month filing window; size as a hedge, since the article alone is not enough to justify an outright short.
• Map cloud-contract exposure before taking a provider trade: identify counterparties and concentration first. If bookings are concentrated and renegotiable only under stress, cloud-provider upside from committed demand may be overstated; if commitments are diversified and collectible, the same contracts could support utilization. Avoid assuming any named provider is exposed without disclosure.
• Watch for falsification or confirmation in audited cash-flow disclosure, revisions to compute commitments, financing terms, and cloud-company commentary on customer concentration or AI-related capex. A cash-burn improvement alongside stable commitments would undermine the near-term credit-bear case; evidence of cancellations, refinancing pressure, or worsening cash conversion would strengthen it.