Ed Zitron: The AI Bubble is Bleeding Cash, Here Are The Receipts – Monetary Matters with Jack Farley

Ed Zitron: The AI Bubble is Bleeding Cash, Here Are The Receipts – Monetary Matters with Jack Farley — Transcript

Market views

Bearish

OPENAI— OpenAI lost $21 billion in 2025 on $13 billion revenue with no path to profitability. Variable costs worsen with scale, sales/marketing inflated to hide true inference costs, and company faces $300+ billion in compute commitments it cannot afford.
ANTHROPIC— Anthropic has similar economics to OpenAI with costs increasing proportionally to revenue. Their ARR calculation methodology is misleading (daily subscribers x12 + 4 weeks API spend x13), and they face the same unsustainable burn rate and compute commitment issues.
MSFT— Microsoft's Azure revenue is heavily dependent on OpenAI's $17 billion spend, representing a material portion of their RPO from a customer that cannot afford to pay long-term. This creates significant balance sheet risk as hyperscaler becomes increasingly debt-heavy.
META— Meta has no coherent AI strategy, spending $14+ billion on disconnected initiatives without clear monetization. Reorganized AI department four times, and incremental ad improvements don't justify massive GPU investments. Zuckerberg cannot be fired, creating governance risk.
NVDA— NVIDIA's revenue growth is sustained by circular financing with unprofitable AI labs and unrealistic market expectations. The company has backstopped deals and created artificial demand, setting up for disappointment if guidance doesn't meet inflated expectations in coming quarters.
AILABS— AI labs collectively face structural unprofitability with 89% of top AI revenue concentrated in two massively loss-making companies. Token-based billing created temporary revenue bursts that are now reversing as enterprises implement strict usage limits after burning through budgets.
CLOUDGPU— AI compute demand is illusory - the majority comes from just Anthropic, OpenAI, and Meta. Natural enterprise demand is only $1-2 billion, not the hundreds of billions needed to justify planned data center capacity. Circular financing masks lack of real customers.
DATACENTER— Over 100 gigawatts of AI data centers are planned but lack corresponding demand. Many announced partnerships (OpenAI-Broadcom 10GW, OpenAI-AMD 6GW) never materialized. GPUs will sit fallow as unprofitable AI labs cannot sustain their compute commitments.
VC— Venture capital is at its limit for AI funding, evidenced by private equity and credit funds entering the space. OpenAI and Anthropic need $150 billion in next 6-12 months but face increasingly difficult fundraising as their economics become public through IPO disclosures.

Not investment advice.

Mentioned

Market views