World Macro Radar Global Macro, Explained

Macro Focus

AI Macro Focus

A public layer-by-layer report on the AI system, from applications to suppliers.

Latest update: 2026-04-07
View Archive

How To Read This Report

This report follows the AI value chain from applications through labs, cloud, data centers, power, semiconductors, and suppliers. Read each layer in four parts: Core metrics for the current state, What changed for the latest movement, AI system impact for what it means inside the chain, and Human system impact for what it may mean for work, capital, and society. Public market expression is an equal-weight basket for that layer. The score uses spot, 5-day, 30-day, and 90-day performance versus SPY, with VIX as the risk backdrop; the plain-English status matters more than the raw number.

Application - Consumer

Core metrics

  • Public market expression: weak
  • Market score: 29.0/100
  • Spot / 5d: -1.0% / +1.6%
  • 30d vs SPY: -11.9% vs -3.9%
  • 90d vs SPY: -9.5% vs +1.1%
  • VIX: 24.5
  • Consumer AI usage (blended proxy): 850 million users in April 2026, unchanged 0.0% vs previous official disclosure

What changed

  • Public basket check: latest session down 1.0%; 5d relative: 0.1 percentage points behind SPY.

AI system impact

  • Consumer AI usage (blended proxy) is 850 million users in April 2026. Even without a fresh daily change, that still points to a very large installed demand base for inference and distribution. Public markets are not fully confirming that story right now, which is a sign that investors still see fragility, crowding, or slower payback in this layer.

Human system impact

  • This is where AI becomes ordinary behavior: searching, writing, studying, and asking for help. Today's read says that habit is already large, even if the daily social effects are hard to measure cleanly. That caution favors the biggest distribution platforms over smaller consumer challengers that still need to prove they can turn usage into durable profit.

Application - Enterprise

Core metrics

  • Public market expression: neutral
  • Market score: 46.8/100
  • Spot / 5d: +0.9% / +1.4%
  • 30d vs SPY: +3.8% vs -3.9%
  • 90d vs SPY: -11.2% vs +1.1%
  • VIX: 24.5
  • Enterprise AI adoption (blended proxy): 8 million seats in Q2 2026, unchanged 0.0% vs previous official disclosure

What changed

  • Public basket check: latest session up 0.9%; 5d relative: 0.3 percentage points behind SPY.

AI system impact

  • Enterprise AI adoption (blended proxy) is 8 million seats in Q2 2026. That still matters because enterprise rollouts are what turn model capability into recurring software and cloud spend. Public markets are giving a mixed read, so the operating story is clearer than the valuation signal.

Human system impact

  • This is the route by which AI moves from demo to workflow. When enterprise adoption rises, the real questions become budgeting, job design, training, and which teams are expected to do more with the same headcount. The market signal is mixed, so the wider business and social consequences remain unsettled.

Foundation Labs

Core metrics

  • Public market expression: weak
  • Market score: 30.3/100
  • Spot / 5d: -0.7% / +2.2%
  • 30d vs SPY: -12.7% vs -3.9%
  • 90d vs SPY: -6.3% vs +1.1%
  • VIX: 24.5
  • Model launches / upgrades: Mistral released Mistral 3, showing that frontier-model development outside the US is still active.
  • Platform / app launches: No major named lab product launch was logged in the last 30 days.
  • Enterprise / channel partnerships: No major named lab distribution or enterprise deal was logged in the last 30 days.
  • Reach / adoption / ecosystem scale: Anthropic expanded its Google Cloud TPU usage and pointed to continued business-customer growth. This is directional rather than precise, because the disclosure does not tell us by how much business demand increased.
  • OpenAI revenue estimate (secondary source): about $2.0bn a month, reported on March 2026

What changed

  • Public basket check: latest session down 0.7%; 5d relative: 0.5 percentage points ahead of SPY.
  • Commercial scale still looks real: Anthropic pointed to continued business demand, and secondary reporting still put OpenAI near $2.0bn of monthly revenue.
  • Frontier product competition is still active outside the US: Mistral released another model.

AI system impact

  • The lab layer is still doing two things at once: shipping new models and proving customers will pay. That combination keeps demand for cloud compute and advanced chips from easing. Public markets are not fully confirming that story right now, which is a sign that investors still see fragility, crowding, or slower payback in this layer.

Human system impact

  • Today's lab read matters because commercialization is no longer hypothetical. If frontier products keep improving and still attract paying demand, more managers have a reason to test them against work currently done by people. That caution can narrow which labs can keep funding model development at full scale and which ones fall back into dependence on larger partners.

Cloud

Core metrics

  • Public market expression: weak
  • Market score: 30.3/100
  • Spot / 5d: -0.5% / +1.2%
  • 30d vs SPY: -9.1% vs -3.9%
  • 90d vs SPY: -14.7% vs +1.1%
  • VIX: 24.5
  • Demand / growth: Microsoft reported 40% growth in its latest cloud disclosure. That is a strong company signal, but it is too narrow to stand in for the whole cloud layer.
  • Product / platform launches: No major named item was logged in the last 30 days.
  • Customer / partnership wins: No major named item was logged in the last 30 days.
  • Capacity / network expansion: No major named item was logged in the last 30 days.

What changed

  • Public basket check: latest session down 0.5%; 5d relative: 0.5 percentage points behind SPY.
  • The clearest cloud change in the current data remains Microsoft's 40% Azure growth print; the rest of the cloud layer has not yet produced an equally strong named update.

AI system impact

  • The cloud read is positive but narrow: one major provider is still growing very fast. That is enough to keep the infrastructure story expansionary, but not enough to say the whole cloud layer is accelerating in sync. Public markets are not fully confirming that story right now, which is a sign that investors still see fragility, crowding, or slower payback in this layer.

Human system impact

  • If AI is mostly delivered through a few cloud platforms, then most firms adopt it as renters rather than owners. That affects pricing power, dependency, and how much smaller firms can build for themselves. That caution still favors incumbent platforms with the balance sheets to keep building while everyone else rents capacity from them.

Data Centers

Core metrics

  • Public market expression: neutral
  • Market score: 47.9/100
  • Spot / 5d: +0.6% / +3.4%
  • 30d vs SPY: +1.5% vs -3.9%
  • 90d vs SPY: -2.2% vs +1.1%
  • VIX: 24.5
  • Capacity under development: Equinix said it had about 1.0 GW of additional powered land under control in February 2026, alongside 52 major expansion projects. For one operator this is meaningful, but it is not enough to stand in for the whole market.
  • Named expansion signal: No additional large named data-center expansion item was logged in the last 30 days.

What changed

  • Public basket check: latest session up 0.6%; 5d relative: 1.8 percentage points ahead of SPY.
  • No comparably large new buildout disclosure arrived in the last 30 days, so this layer still rests mainly on the latest stored operator update.

AI system impact

  • The data-center layer still says physical expansion is continuing, but the evidence base here is narrow. We can say construction is still moving; we cannot yet say by how much at the full market level. Public markets are giving a mixed read, so the operating story is clearer than the valuation signal.

Human system impact

  • This is where AI stops being abstract and becomes local: land deals, construction, utility negotiations, water, and permitting fights all start to matter to real communities. The market signal is mixed, so the wider business and social consequences remain unsettled.

Power

Core metrics

  • Public market expression: soft
  • Market score: 44.0/100
  • Spot / 5d: +0.2% / +1.1%
  • 30d vs SPY: -1.0% vs -3.9%
  • 90d vs SPY: +17.3% vs +1.1%
  • VIX: 24.5
  • Power-ready queue capacity: LBNL's latest queue data shows 195.1 GW of active capacity with signed interconnection agreements across the main tracked regions (ERCOT 90.3 GW, PJM 30.6 GW, CAISO 74.3 GW). That is a large amount of already-advanced queue capacity competing for real buildout.
  • Grid interconnection timing backdrop: LBNL's completed-project data says request to operation took about 4.5 years in 2024 completed projects. This is backdrop, not a daily spot metric, but it shows that big power-linked projects still move on multi-year timelines.
  • Power cost: the short-run wholesale proxy built from ERCOT North and California NP15 was $16/MWh reported on April 01, 2026. This is a volatile market-price signal, not the long-term contracted power price a data center would actually try to lock in.
  • Contracted power and generation mix: the latest large disclosed AI-linked contract in the dataset was 0.8 GW, and that newly contracted supply was 100% low-carbon. This is useful as a read on what big AI buyers are trying to secure, not as a full market total.

What changed

  • Public basket check: latest session up 0.2%; 5d relative: 0.6 percentage points behind SPY.
  • Near-term wholesale power costs moved down 45.9% versus March 31, 2026, though this is a noisy short-run market signal rather than a full long-term power-access answer.
  • Microsoft's Constellation deal still anchors the contracted-power read: 835 MW of clean baseload under a 20-year agreement.
  • The queue data still says ERCOT is far ahead of PJM on already-advanced capacity, which helps explain why Texas looks easier to build in than Northern Virginia on this measure.

AI system impact

  • Power still looks like the binding physical constraint. Money and demand are not enough on their own; projects need signed agreements, real queue progress, and actual electricity. Public markets are not fully confirming that story right now, which is a sign that investors still see fragility, crowding, or slower payback in this layer.

Human system impact

  • This is where AI expansion runs into the wider human system most directly, because grid access and electricity prices can spill into utility politics, industrial policy, and local conflict over scarce infrastructure. That caution can make it harder to finance the generation and grid projects needed to relieve the bottleneck, even when demand is obvious.

Semiconductors

Core metrics

  • Public market expression: neutral
  • Market score: 50.0/100
  • Spot / 5d: +0.3% / +3.1%
  • 30d vs SPY: -6.7% vs -3.9%
  • 90d vs SPY: +14.4% vs +1.1%
  • VIX: 24.5
  • Advanced chip manufacturing demand (TSMC revenue proxy): NT$317.7bn in February 2026

What changed

  • Public basket check: latest session up 0.3%; 5d relative: 1.4 percentage points ahead of SPY.
  • SK hynix said HBM demand remained strong in its latest results, showing that AI memory demand is still tight.

AI system impact

  • Advanced chip manufacturing demand (TSMC revenue proxy) is NT$317.7bn in February 2026. Chip supply still sets the pace for how quickly the rest of the AI system can expand. Public markets are giving a mixed read, so the operating story is clearer than the valuation signal.

Human system impact

  • A system this dependent on a handful of chipmakers leaves profit, leverage, and strategic risk concentrated in a very small number of firms and places. The market signal is mixed, so the wider business and social consequences remain unsettled.

Suppliers / Inputs

Core metrics

  • Public market expression: firm
  • Market score: 56.0/100
  • Spot / 5d: -1.5% / +2.7%
  • 30d vs SPY: -4.9% vs -3.9%
  • 90d vs SPY: +46.9% vs +1.1%
  • VIX: 24.5
  • Chip-equipment demand (ASML bookings proxy): EUR 13.2bn in Q1 2026

What changed

  • Public basket check: latest session down 1.5%; 5d relative: 1.0 percentage points ahead of SPY.
  • ASML reported EUR 13.2bn of net bookings in Q1 2026.

AI system impact

  • Chip-equipment demand (ASML bookings proxy) is EUR 13.2bn in Q1 2026. Upstream equipment demand still supports further chip-capacity expansion. Public markets are reinforcing that read, which makes it easier for this part of the chain to attract capital and keep expanding.

Human system impact

  • When tool supply stays concentrated, countries and firms without direct access fall further behind. That turns a technical bottleneck into a strategic one. When markets reward this layer, the dominant toolmakers gain more room to expand capacity and stay ahead of would-be challengers.

30d named / stored event count: 15