The Daily Receipts

Epoch AI modeled the total cost of ownership of a typical 1-gigawatt AI data center built by a US hyperscaler around Nvidia's GB200 NVL72 systems, and found the up-front capital expenditure alone comes to $38 billion for one hyperscale AI data center

A January 2026 IMF staff discussion note tracking the spread of AI-related job skills across US local labor markets found that in occupations highly exposed to AI but with little scope to complement it, five years after AI-skill demand appeared, employment was 3.6% lower than in other regions

A National Bureau of Economic Research working paper, drawing on the Federal Reserve Banks of Atlanta and Richmond's CFO Survey of roughly 750 corporate executives, found that by the end of 2025 more than 40% of US companies still had not invested in AI at all, as of year-end 2025

TrendForce reports TSMC is racing to keep up with AI chip orders from Nvidia, AMD and Broadcom -- monthly 3-nanometer wafer starts, already near 150,000 in the first half of 2026, could reach 180,000 wafers a month by early Q4 2026

AI models can now win gold at the International Mathematical Olympiad, yet Stanford's 2026 AI Index found they still fumble a task most children master early. Across ClockBench's 180 clock designs and 720 questions, the top model, GPT-5.4 High, read analog clocks correctly only 50.6% of the time, versus 90.1% for humans

Equinix, the world's largest data-center operator, ties its buildout of hyperscale capacity directly to demand it links explicitly to the adoption of AI. Its capital expenditure in the quarter ended 30 June 2026 reached Equinix's Q2 2026 capital expenditures: $1.578 billion, up 59.6% year-over-year