The week in three numbers: Aug 2 two AI transparency laws take effect the same day, in the EU and California · 8.2% the share of revenue Infosys now calls "AI" — while cutting its growth forecast · 1 the number of SEC filings this week that tie a layoff to AI.

This week's question: who actually keeps the money AI creates? The event is the fact; the money read is our analysis, labeled separately.

Named companies' AI value claims against what shows in the P&L, on one ladder. None of the eleven companies tracked has reached Stage 4 — until one does, we won't treat a public AI-value claim as evidence of a durable moat.

The evidence ladder: 0 · Narrative · 1 · Operational (a quantified activity number) · 2 Financial claim (a dollar figure asserted) · 3 P&L-attributed (AI isolated in reported results) · 4 Sustained (Stage 3 holds four straight quarters).

Four rows added this week, and note their shape: every new row quantified a cost or a contract, and none isolated an AI profit. S&P Global and Ecolab both report Tuesday — the first real chance in months to move a row up, or to mark two rows down.

The rent is moving to the things you cannot copy: power, and proprietary data. A new NBER paper gives the sharpest evidence yet. Using 380 trillion tokens of actual AI usage across more than four hundred models, the authors find a persistent stock-market premium for AI-exposed companies worth about 64 basis points a week — but it concentrates almost entirely at the frontier: paid, closed-source, heavy usage. It is not there for casual or open-weight use, and not there in emerging markets. Read plainly: markets pay for capability that is scarce and pay nothing for capability that is abundant.

The week's business news fits that shape uncomfortably well. Domo sold substantially all of its assets to Progress Software for $400 million — while carrying roughly $246 million of net cash and more than $900 million of loss carryforwards in the shell. LivePerson, whose combined customer base includes 25 of the Fortune 100, told shareholders its outstanding debt "currently exceeds the total value of the transaction," after its board contacted 66 potential counterparties. Distribution into the largest companies in the world was worth little once the underlying capability commoditized. (Domo 8-K, LivePerson 8-K)

Follow the money down rather than up and it keeps going. GE Vernova raised its free-cash-flow guidance from $6.5–7.5 billion to $11.5–12.5 billion, on data-center orders now "over $5 billion year-to-date, more than double our 2025 total." Hut 8 signed a second 15-year, $9.8 billion lease for 352 megawatts of IT capacity with a 3% annual escalator — on a site that used to be an aluminum smelter. IREN signed $2.8 billion in new AI-cloud contracts carrying "customer prepayments representing approximately 45% of the associated GPU capital expenditure." And PJM's independent market monitor puts data centers at $6.3 billion — 38% — of the $16.4 billion in charges from the latest capacity auction, which lands on everyone else's electricity bill. (The auction figure is the market monitor's president on the record; his written analysis is not published yet — Reported, not Primary.) (GE Vernova 8-K, Hut 8 8-K, IREN, PJM IMM, via Utility Dive)

The honest counter-case: the NBER result is an associational premium in equity prices — it measures what investors expect, not profit anyone booked, and expectations have been wrong before. Fed chair Kevin Warsh told Congress this month that productivity growth has been strong "predating gains from AI adoption" — the strength in the data so far is not AI's to claim. (Testimony, Jul 14)

Mostly "anyone can buy it" → you are funding a cost line. Mostly "only us, tied to our business" → that is where advantage forms, and it deserves the funding the tool itself does not.

The live split between serious people, and where we come down.

[How we label evidence: Primary source · Corroborated · Reported · Vendor claim · Analysis.] · Written & edited by Mario Suarez · Independent analysis · Every link and date verified before send.

AI Above the Cut is a weekly decision brief for executives — VP-and-up leaders in business who want signal over noise. It covers the outcomes and impact of AI rather than its engineering, and asks a standing question of every development: who keeps the rents? Each Sunday we read a fixed spine of the field's highest-signal voices — operators, researchers, and independent skeptics like Andrew Ng, Ethan Mollick, Simon Willison, Nathan Lambert, the AI Snake Oil team, Erik Brynjolfsson, and Cassie Kozyrkov — plus a rotating edge of specialists (Chip Huyen, Jack Clark, Ben Thompson, Benedict Evans, and others) and the primary research, regulator, and lab feeds. We tag every source — vendor, researcher, operator, investor, regulator, economist, or skeptic — and check strong claims across categories, so we curate evidence, implementation, and disagreement rather than celebrity.

The brief comes in two speeds: a fast skim — the single most important development, three concrete moves, and the week's decision-relevant signals plus one Below the Cut counter-signal — then the Margin-Proof Tracker, our standing scorecard of AI-value claims against reported P&L evidence, and a longer Synthesis that connects the moves, takes a position, and names the tests we're watching. We optimize for quality over influence, link to the source rather than the hype around it, and flag anything unconfirmed. No "10 AI tools you need today."