These are examples of what Noesis answers look like — tables and charts
built from primary sources, dated, with the most useful follow-up
questions. The live answer engine is in early access.
What are the top 10 stocks in QQQ — and what's their weight?
The Invesco QQQ Trust tracks the Nasdaq-100. Its top holdings are
dominated by mega-cap tech, and the concentration is striking: the top
ten names alone are over 54% of the entire
fund.
#
Ticker
Company
Weight
Relative
1
AAPL
Apple
9.0%
2
MSFT
Microsoft
8.0%
3
NVDA
NVIDIA
7.5%
4
AMZN
Amazon
5.7%
5
AVGO
Broadcom
5.0%
6
META
Meta Platforms
4.9%
7
GOOG
Alphabet (A + C)
5.6%
8
TSLA
Tesla
3.0%
9
COST
Costco
2.8%
10
NFLX
Netflix
2.5%
Top 10 cumulative
54.0%
Noesis read
QQQ's top-10 concentration is its biggest factor — and its biggest
risk. A single decade-long shift in semiconductors, cloud, or
consumer tech can move the whole fund. If you hold QQQ, you mostly
own these ten names.
Source: Invesco QQQ Trust public fund disclosures, late 2024. Weights
shift daily as prices and the Nasdaq-100 rebalance.
Which countries consume the most crude oil — and how much do they make themselves?
The world burns about 102 million barrels of crude
oil per day. The top consumers and producers are rarely the same set,
which is why oil is the single biggest geopolitical lever in modern
economics.
Country
Consumption (mbd)
Production (mbd)
Net import (mbd)
Self-sufficient?
United States
19.4
13.2
+6.2
Net importer
China
15.5
4.1
+11.4
Net importer
India
5.4
0.7
+4.7
Net importer
Russia
3.5
10.7
-7.2
Yes
Saudi Arabia
3.7
10.4
-6.7
Yes
Japan
3.3
0.0
+3.3
Net importer
Brazil
2.6
3.6
-1.0
Yes
South Korea
2.6
0.0
+2.6
Net importer
Canada
2.5
5.5
-3.0
Yes
Germany
2.0
0.1
+1.9
Net importer
Noesis read
The US is the world's largest oil consumer and the largest
producer — and still a net importer of crude. China imports the
most by volume (over 11 mbd). Only Russia, Saudi Arabia, Canada,
and Brazil among the top consumers run a surplus. Energy
self-sufficiency is not a binary — it's a balance sheet.
Source: U.S. Energy Information Administration (EIA) International Energy Statistics
and the International Energy Agency (IEA) Oil Market Report, 2023 averages.
Figures cover crude oil and condensates; values rounded.
How is the world's largest economies' electricity made?
Coal still dominates Asia, gas anchors the United States, France runs
on nuclear, and Germany is in the middle of the transition. One
chart, ten countries, eight sources.
China
United States
India
Japan
Germany
France
Brazil
United Kingdom
Russia
South Korea
CoalGasNuclearHydroWindSolarOilOther
Noesis read
China and India are still over half coal. France's grid is the
cleanest of any G7, almost entirely nuclear + hydro. Germany and
the UK lead the gas-to-wind pivot. The US is in the middle —
cleaner than it looks because it swapped coal for gas, but still
burning a lot of fossil fuels for power.
Source: International Energy Agency (IEA) Electricity Information 2024 release,
covering 2023 generation by source. "Other" includes biomass, waste, and
minor categories. Values rounded to whole percentages.
How much energy does an AI query use — small picture and big picture?
A single LLM query is tiny. A planet's worth of LLM queries is huge.
Both are true.
Per query
~0.3 Wh
A typical chat answer is roughly one Google search.
Light-bulb equivalent
~2 min
A 10W LED bulb running for two minutes equals one query.
ChatGPT, per day
~300 MWh
Powers ~25,000 US homes for a day.
Average US home
~30 kWh/day
≈ 100,000 LLM queries equal one home for one day.
Training one big model
~50 GWh
One frontier training run ≈ a small US city for a week.
AI share of US grid by 2030
8–12%
Projected — all data centers today are ~4%.
Noesis read
The per-query cost is genuinely small — running a model is cheap
compared to making one. The aggregate is where the story lives:
billions of queries per day plus an ongoing arms race to train
bigger models means AI is on track to be a single-digit-percent
chunk of US electricity demand within five years. Whether that
"breaks the grid" depends almost entirely on where the new data
centers get built.
Source: IEA Electricity 2024 report (data-center forecasts), Schneider Electric
and industry estimates for per-query energy, and public training-energy disclosures
from leading AI labs. Per-query figures are inherently rough — they depend on the
model, the prompt, and the data-center efficiency (PUE).
Is the AI / memory-chip rally a bubble — and where are we in the cycle?
"History doesn't repeat, but it often rhymes." Even if this is a
bubble, the foundation has real economic value. The useful question
isn't "bubble or not" — it's which phase rhymes and
what would change the read.
The canonical bubble curve (Rodrigue, after Minsky)
Position is a hypothesis, not a measurement. The point of the curve is to ask
"what would have to be true for us to be earlier (or later) than this?"
Five historical analogues — how the current rally rhymes
Era
Why it rhymes
Peak-to-trough
Recovery
Rhyme
Dot-com 1995–2000
Sector capex super-cycle; "new economy" narrative; real productivity revolution; concentration in mega-cap tech
Nasdaq −78%
~15 yrs to new high, 2015
Strong
Japan asset bubble 1985–1989
Liquidity-driven; valuation extremes; cross-holdings + concentration. Today's rates are higher — a partial rhyme.
Nikkei −80%
~34 yrs to new high, 2024
Medium
Railroad mania US, 1860s–1893
Real revolutionary infrastructure; massive capex; monopoly fears; multiple boom-busts within a real revolution
~−50% per panic
~5–10 yrs per cycle
Strong on the "real revolution" parallel
Nifty Fifty 1968–1973
Mega-cap concentration; "growth at any price"; durable quality names priced for perfection
~−50% cohort
~7–8 yrs to prior peak
Strong on concentration
Semi cycle 1995–2001
Memory boom + capex super-cycle inside the larger dot-com — the most direct rhyme for memory specifically
Memory −80% Capex-cyclicals −70%
~3–5 yrs
Strongest for memory chips
Signals — where each phase ends, and where we are
Signal
Dot-com peak Mar 2000
Japan 1989
Today rough estimate
Top-name fwd P/E
100–200×
~70× (broad)
30–40× leaders
Sector capex / GDP
4× trend
2.5× trend
3–4× trend
IPO frenzy
Extreme
Extreme
Moderate, rising
Retail participation
Extreme
Extreme
High, contested
Narrative saturation
Total
Total
High
Real free cash flow
Lagging price
Lagging
Higher — meaningful FCF
Insider selling
Heavy
Heavy
Moderate–Heavy
Margin debt / mcap
Peak
Peak
Elevated, not extreme
Noesis read
The current rally rhymes strongly with railroads and dot-com — a genuine
technological revolution that markets are pricing aggressively. It rhymes
weakly with tulips: there is real free cash flow, especially in
compute infrastructure.
Where we likely sit: awareness → early mania. Several
extreme-end signals are not yet present (parabolic blow-off, IPO frenzy at
dot-com intensity). Others are present (concentration, narrative saturation,
retail interest).
The fear-vs-greed fight phase — every dip is bought, bears
get stopped out, anxieties get dismissed — historically precedes the
final mania leg by 6–12 months. The 2025 corrections that were bought
hard fit this pattern. The transition into blow-off typically arrives when
sentiment reaches near-unanimous bullishness.
Two markers historically signaled the end of cycles like this:
Real earnings break the narrative. Watch top-3 hyperscaler capex guidance and NVDA-like Q-over-Q growth. Quarter-over-quarter flat-lining while valuations remain extreme is the inflection.
Fed policy turns. Lower rates have powered every modern bubble (Japan 1985, dot-com 1995, housing 2003). Re-acceleration of cuts extends; hawkish pivot compresses.
Duration: "obvious froth → blow-off top" historically takes 12–24 months.
Blow-off → despair: 9–18 months. Recovery for the survivors (the real
revolutions): single-digit years for the leaders, sometimes decades for
indices. The strategic implication: even if you believe we're in
awareness → mania, you may have ~12–24 months before peak — but the end is
sudden.
What would change the read
↑ Mania confidence if: parabolic move in semi indices · IPO frenzy approaches dot-com intensity · mainstream "this time is different" narratives saturate · retail margin balances accelerate faster than market cap.
↓ Bubble concern if: earnings keep beating · real FCF growth continues at current pace · breadth widens (small/mid-cap participation) · capex efficiency improves (revenue per $ of capex rises).
Sources: Robert J. Shiller, Irrational Exuberance (CAPE methodology);
Charles P. Kindleberger, Manias, Panics, and Crashes (Minsky-cycle
taxonomy); Reinhart & Rogoff, This Time Is Different; BIS bubble
research; NBER working papers on asset cycles; Macrotrends and historical
Nasdaq / Nikkei price series. Current-cycle signals are illustrative estimates
(training data), not live measurements — the live engine wires real-time feeds.
See KAN-114 for the live build.
What is the 4% rule for retirement — and does it still hold?
The 4% rule says you can withdraw 4% of your initial portfolio in
year one, adjust the dollar amount for inflation each year after, and
have a high probability of your money lasting 30 years. Here's how
the math sits at different withdrawal rates.
Withdrawal rate (yr 1)
30-yr success rate
Visual
3.0%
99%
3.5%
96%
4.0%
89%
4.5%
75%
5.0%
56%
5.5%
38%
Noesis read
4% is still a defensible starting point for a 30-year horizon with
a 50/50 stock/bond mix. Several updated studies (Morningstar,
among others) argue the safer number today is closer to
3.3–3.7%, because expected returns are lower than the late-20th-
century data the original rule was built on. The honest answer:
4% is a good rule of thumb; treat it as the upper end of
comfortable, not a guarantee. Stress-test your own plan in the
Retirement Planner →.
Source: Bengen, W. (1994) Determining Withdrawal Rates Using Historical Data;
Cooley, Hubbard, Walz (Trinity Study, 1998); Morningstar 2024 State of
Retirement Income. Success rates are illustrative — actual outcomes depend
on allocation, fees, and the specific historical (or Monte-Carlo) regime modeled.
Which US states are the best for doing business right now?
CNBC's 20th annual study scored all 50 states across 138 metrics
in 10 categories. In 2026, Infrastructure became the
top-weighted category for the first time — including a new "ease
of permitting" metric — reflecting the reality of AI-era data-center and
advanced-manufacturing site selection. Economy dropped from #1 to #2.
Here's how the rankings landed.
#
State
Best category rank
Watch out for
1
Ohio
Cost of Doing Business #1
No red flags — balanced top-quartile
2
North Carolina
Workforce #3, Access to Capital #5
Cost of Living #35
3
Virginia
Education #5, Tech #6, Quality of Life #7
Cost of Doing Business #26
4
Texas
Workforce #1, Access to Capital #2, Tech #3
Quality of Life #49
5
Minnesota
Quality of Life #4, Education #10
Cost of Doing Business #31
6
Michigan
Cost of Doing Business #4
Business Friendliness #27, Education #29
7
Georgia
Access to Capital #8, Workforce #9
Quality of Life #46
8
Florida
Workforce #2, Access to Capital #4
Cost of Living #48
9
Tennessee
Business Friendliness #7, Workforce #7
Quality of Life #50, Education #45
10
Indiana
Cost of Doing Business #2, Cost of Living #6
Quality of Life #48
11
Washington
Tech #4, Workforce #6
Cost of Doing Business #47
12
Illinois
Education #7
Business Friendliness #44
13
Pennsylvania
Tech #5, Access to Capital #7
Business Friendliness #42
14
Arizona
Workforce #4
Education #43
15
Massachusetts
Education #1, Access to Capital #6
Cost of Doing Business #49
Bold = #1 in that category nationally. Full 50-state × 10-category
breakdown archived in the Noesis knowledge base.
Category #1 winners
Workforce
Texas
Cost of Doing Business
Ohio
Technology & Innovation
California
Access to Capital
California
Business Friendliness
North Dakota
Education
Massachusetts
Quality of Life
Vermont
Cost of Living
West Virginia
The Noesis read
The 2026 Infrastructure re-weighting is the most
consequential methodology shift since 2020. As hyperscaler AI capex
tops $750B in 2026 — projected up to $920B by Goldman — the states
already leading on Cost of Doing Business AND with real infrastructure
investment programs (Ohio, Indiana, Michigan, Georgia, North Carolina)
consolidate a structural site-selection advantage.
The sharpest single-state paradox: California is #1 in
both Technology & Innovation AND Access to Capital — yet #50 in Cost of
Living. The venture + academic + industrial cluster remains structurally
dominant even as operational costs push corporations out. Similar
paradox: Massachusetts #1 Education, #49 Cost of Doing Business.
Brain-cost tradeoffs have diminishing returns when friction dominates.
Arkansas was named Most Improved in 2026 — the "Walmart
HQ effect" pulling workers to Bentonville for low costs + rising
quality of life. Consistent with the ongoing Sun Belt + Middle America
workforce migration.
What would change the read next year: if bottom-quartile
Business Friendliness states (NJ #50, CA #47, NY #49) pass real regulatory
reform, they close the gap. If they don't — and Infrastructure remains
top-weighted — expect the top-10 to consolidate further around Ohio, TX,
NC, VA, MI, IN.
Source: CNBC's America's Top States for Business 2026, 20th
annual edition, published 2026-07-09 by Scott Cohn. Methodology: 138
metrics across 10 categories, 2,500 max points.
The Noesis live answer engine is in early access. Submit a question
you want answered the way these are — primary sources, full picture,
no hand-waving — and you'll be first in.
