MacroVoices #526 Matt Barrie: Pay To PrAI

Episode Summary

Executive Summary: This episode centers on two major macro risks: a potential AI valuation and business-model bubble, and a rapidly escalating Iran conflict with major energy-market implications. Matt Berry argues AI inference economics are structurally unprofitable and may force a shift to pay-per-token pricing, while Anas Al-Hajji warns the Middle East conflict could trigger oil, LNG, and broader global recessionary shocks. The postgame translates these themes into trades and chart levels.

Main Topics: AI valuation bubble and unsustainable unit economics (Priority: 5/5): Matt Berry argues the AI industry is burning enormous capital, with consumer and developer subscriptions priced below true inference costs. He says the current model depends on heavy subsidization, vendor financing, and circular capex spending that may not be sustainable. OpenAI mega-round and circular financing (Priority: 5/5): The discussion dissects OpenAI's reported $122 billion fundraising, emphasizing that much of it is in-kind or contingent vendor financing rather than pure cash, and that the round reflects an attempt to bridge to an IPO while scaling compute infrastructure. Pay-per-token future and software as a slot machine (Priority: 4/5): Berry predicts AI will move from flat-rate subscriptions to token-based pricing because current plans lose money. He warns this will make software development feel like gambling, with unpredictable costs and outcomes, especially for power users. AI, private credit, and SaaS stress (Priority: 4/5): The episode links AI disruption to private credit risk, arguing that debt-financed data-center and SaaS exposure could deteriorate as AI pressures software margins and forces more borrowing to fund growth. Iran conflict escalation and energy-market shock (Priority: 5/5): Anas Al-Hajji argues the war is not winding down and that President Trump's speech signaled a longer conflict. He warns that attacks on civilian power infrastructure, desalination, or nuclear facilities could trigger severe humanitarian and market consequences. Oil, LNG, and global recession/stagflation risk (Priority: 5/5): Al-Hajji says the conflict could keep oil and LNG prices elevated until demand destruction or recession intervenes. He expects shortages across oil, LNG, NGLs, fertilizer, methanol, and other products, with Asia and emerging markets especially vulnerable. Trade ideas and market positioning (Priority: 3/5): Patrick Ceresna frames a public-market expression of private-credit stress via BIZD put options, while the postgame reviews technical levels in equities, dollar, crude, gold, uranium, copper, and Treasury yields in light of the geopolitical shock.

Key Arguments: AI is not yet a profitable consumer or developer business at current flat-rate pricing; the $20 and $200 plans likely cost close to or more than they generate. The OpenAI mega-round is partly vendor financing and in-kind compute, not a straightforward cash equity raise, suggesting the industry is trying to buy time until IPO. AI model competition and low switching costs mean each new generation burns more tokens for more complex tasks, worsening economics rather than improving them. A pay-per-token model is likely inevitable, but it may alienate users because costs become unpredictable and software development turns into a high-variance, slot-machine-like workflow. AI will likely create a productivity boom for highly skilled users while displacing or compressing lower-skill roles, intensifying inequality and the K-shaped economy. The Iran conflict is not de-escalating; Trump’s speech implied a longer war and possible attacks on civilian power infrastructure, which could provoke Iranian retaliation against desalination and nuclear assets. The Strait of Hormuz situation is fundamentally an insurance and logistics problem, not just a naval one; ships will not move without affordable insurance even if escorted. Oil and LNG shortages could persist until demand destruction or recession forces consumption lower; the crisis could push the global economy into stagflation or recession. Private credit is vulnerable because it is exposed to debt-financed AI infrastructure and SaaS businesses whose margins may be impaired by AI-driven competition. The market may be underpricing geopolitical tail risk, especially the possibility of attacks on civilian nuclear facilities, which would be a major escalation and a blow to the nuclear renaissance.

Data Points: OpenAI fundraising round: $122 billion - Discussed as the largest private fundraising round ever, with much of it described as vendor financing/in-kind support. OpenAI pre-money valuation: $730 billion - Used to frame the implied valuation before the new round. OpenAI revenue run rate: about $25 billion annually / $2 billion per month - Berry cited this as current revenue while arguing losses remain massive. OpenAI expected loss: $14 billion this year - Berry said the company is still deeply unprofitable. OpenAI burn rate: $70 million per day this year; $156 million per day next year - Berry used these figures to illustrate the scale of cash consumption. AI industry capex: about $600 billion per year - Berry said hyperscalers are spending at this scale on AI infrastructure. Projected AI capex by 2030: $5.2 trillion - Berry cited this as a hypothesized future run rate. Training run cost: north of $100 million to approaching $500 million per run - Berry described the expense of training frontier models. Claude $20 plan implied cost: $15 to $20, maybe $18 or more - Berry said even Claude itself would estimate the plan is near break-even or loss-making. VibRank monthly burn: $51,000 in a single month on a $200 plan - Example of extreme power-user token consumption. Median global income: about $2,500 per year - Berry used this to argue $240/year AI subscriptions are expensive globally. BIZD year-to-date performance: down 15%+ - Patrick cited this as the private-credit proxy used for the trade idea. BIZD option trade: May 15, 2026 $13 put at about $1.10 - Patrick proposed this as a convex way to express downside in private credit. S&P 500 move: down 24 bps to 65.75 (as reported in intro) - Macro scoreboard at Wednesday close. WTI crude move: up 1,085 bps to 100.12 in intro; later around 104.50-105 after Trump speech - Oil rallied sharply on the geopolitical escalation. Gold move: up 500 bps to 48.13 in intro; later fell from about 48.17 to 45.79 intraday before rebounding - Gold reacted to shifting inflation and war expectations. U.S. 10-year yield: 4.37% - Macro scoreboard at Wednesday close. Oil shortage estimate: 10 to 12 million barrels per day short - Al-Hajji said mitigation still leaves a large supply gap. Demand destruction threshold: Brent around $160 - Al-Hajji’s model for when demand destruction becomes severe. Freelancer marketplace size: 87 million people - Berry described the scale of Freelancer.com’s cloud workforce. Gene editing challenge prize: $7.5 million - Berry cited a current Moonshot Innovation Challenge with NIH/NASA-related work. Baraka nuclear plant risk: 1 operating civilian nuclear power plant in the UAE - Discussed as a potential retaliation target if Iran’s civilian power infrastructure is hit.

Pivotal Quotes: "the more you use the product, the more you lose the money" — Matt Berry: Explaining why current AI inference economics are structurally unprofitable. "the inevitable destination for these subscription models ... is that they're going to have to move to a per token pricing" — Matt Berry: Describing the likely shift away from flat-rate AI subscriptions. "this is a long war and this is going to continue" — Dr. Anas Al-Hajji: His assessment after President Trump's address on the Iran conflict.

Implications: Listeners should expect continued volatility in AI equities, private credit, oil, LNG, and inflation-sensitive assets. The episode argues that AI monetization may need a reset, while the Middle East conflict could keep energy prices elevated and raise recession/stagflation risk globally.