Saturday, September 19, 2026 · Daily edition
IEA’s 2025 data-centre surge, Minneapolis Fed job transformation, and NIST’s NVD AI agent
Today’s ledger follows an IEA brief on how fast data-centre electricity actually rose in 2025, a Minneapolis Fed magazine feature on who wins inside a job when records work is automated, NIST’s AI-agent enrichment work on the National Vulnerability Database, a first-trimester pregnancy-risk study across more than half a million pregnancies, and the World Bank’s free introductory AI training campaign for adults in Sub-Saharan Africa.
IEA: data-centre electricity jumped 17% in 2025 — bottlenecks now define the scramble
What happened. The International Energy Agency published a news brief on energy and AI that puts observed 2025 growth, not another 2030 forecast, at the center. Data-centre electricity demand rose about 17 percent in 2025, and AI-focused sites climbed faster still — both well ahead of 3 percent global electricity-demand growth. Five large technology companies spent more than $400 billion in 2025, driven by data-centre investment, and are set for a further 75 percent increase in 2026. The named bottlenecks are gas turbines, transformers, advanced chips and IT components, plus planning and regulatory systems that slow grid connections. Tech firms accounted for about 40 percent of corporate renewable power-purchase agreements signed in 2025. Conditional offtake between data-centre operators and small modular reactor projects grew from 25 GW at end-2024 to 45 GW today. Where connections lag, developers are advancing onsite natural-gas generation — largely in the United States — and onsite batteries are becoming critical for hard-swinging AI loads.
What to watch. After a stretch of 2030 paths, queue counts, and ride-through design, the near-term power story is how fast AI load already moved — and whether turbines, transformers, wires, PPAs, SMRs, and onsite gas-plus-batteries can keep up. This is observed growth and bottlenecks, not another long-run TWh path.
Read the IEA brief →
Minneapolis Fed: automating records work can raise average wages — and still hurt the records specialists
What happened. The Federal Reserve Bank of Minneapolis For All magazine featured Freund and Mann’s Institute working paper on job transformation, specialization, and the labor-market effects of AI. Occupations are bundles of tasks; workers are bundles of skills. The authors use large language models to organize about 20,000 O*NET occupation-specific tasks into 38 clusters. Processing and analyzing records looks like the task most likely to be automated — think financial analysts and information and record analysts. The counter-intuitive wage path: if that task disappeared completely, occupations that do a lot of it would see wage gains on average, because time frees up for customer-facing coordination, communication, and negotiation. Workers strong at records work but weaker on those other tasks are likelier to switch jobs and see wages fall. Colleagues less specialized in records are likelier to stay and gain. Moderate exposure helps workers on average; high exposure hurts them, with large dispersion inside the same job title. Returns to social skills rise; returns to analytical skills fall; lower earners gain more than higher earners in the model.
What to watch. The jobs beat is who wins inside the same title when the records task disappears — not a household fear survey, not a posting-decline note, and not a 2026 layoff census.
Read the Minneapolis Fed magazine feature →
Read the Freund & Mann working paper →
NIST puts an AI agent on the National Vulnerability Database
What happened. NIST hosted an Information Technology Laboratory AI webinar on 17 September covering development of an AI-agent enrichment workflow for the National Vulnerability Database. NVD is the U.S. government’s standards-based vulnerability-management repository used across public and private sectors. Rising volume and complexity of discovered vulnerabilities — plus AI tools that help find and exploit them — make timely, actionable enrichment harder. NIST has begun work on an agentic workflow to help enrich vulnerability information; the webinar covered approach, architecture, implementation issues, and early results with the tool at NVD. In parallel, an NVD-modernization Request for Information published 12 August still seeks comments through 13 October 2026 on scalability, automation, interoperability, transparency, and utility.
What to watch. The policy beat is operational cybersecurity infrastructure — an AI agent on the government’s vulnerability ledger — not a labeling statute and not another evaluation-framework docket.
Read the NIST webinar page →
Read the NVD modernization RFI →
First-trimester machine learning beat existing early pregnancy risk tools — and still misfit one country
What happened. News-Medical reported on 18 September a Journal of Medical Internet Research multicenter study on machine-learning first-trimester antenatal risk prediction for adverse maternal and neonatal outcomes. Models using information available in the first 14 weeks analyzed data from more than half a million pregnancies across Sweden, Chile, and Singapore and generally outperformed the early risk-assessment methods already used in each setting. Discrimination improved substantially in Sweden and Chile; Singapore’s gain was smaller but still statistically significant. Calibration did not travel evenly: Swedish and Singaporean models showed reasonable agreement between predicted and observed risks, while the Chilean model was less well calibrated. Social and demographic factors ranked among the most important predictors in some populations — information traditional assessments can miss. The authors frame the work as decision support for closer monitoring or earlier intervention, not a clinician replacement.
What to watch. Early-pregnancy risk tools can look strong at population scale and still fail to calibrate across countries. This is a retrospective model-development study, not a bedside outcome trial.
Read the News-Medical recap →
Open the JMIR study →
World Bank: The Job I Hope For — free introductory AI training for Sub-Saharan African adults
What happened. The World Bank published the campaign brief The Job I Hope For, inviting young people across Africa to reflect on hoped-for jobs and receive free introductory AI training. Eligibility is adults 18 or older who are citizens of named Sub-Saharan African countries; the campaign is not directed at minors. Registration closes 7 October 2026, with three-part online sessions in English and French in October. Modules cover generative-AI fundamentals, effective prompts, and AI do’s and don’ts. Registration also collects hoped-for job, three needs, and how AI fits career ambitions, in English, French, or Portuguese. After training, aggregated answers are to be published as a snapshot of youth aspirations; personal data is to be deleted six months after the campaign ends. The brief notes that more than 600 million young people will enter Africa’s labor market between now and 2050.
What to watch. The education beat is beginner AI-skills access for working-age youth — opportunity design with a hard registration clock, not a learning-outcomes trial and not another connected-campus pilot.
Read the World Bank campaign brief →
Also in today’s ledger
• 2025 electricity growth is not a 2030 TWh path — and task transformation is not a pink-slip census.
• An NVD agent workflow is not a labeling statute.
• A multi-country risk model is not a bedside RCT — and a free intro course is not a learning-gain print.
|