How do you read an energy headline like a researcher?

Tag, check and weigh three real energy stories

News lens N1 of Energy Systems, Step by Step: read three real energy stories with seven questions (scale, time, physics, who pays, boundary, counterfactual, uncertainty), run a check with units, spot units and scale errors, and write a three-sentence take.
  • Foundations to Analyst · L1–L2
  • 20 min
  • After lesson 1

The seven questions

Question Ask The trap it catches
1. Scale How big is it next to the whole? A big-sounding number that is a small share, or a national figure read as a global one.
2. Time Power or energy? Which hour, which year? GW read as GWh, or a 2030 projection read as today's fact.
3. Physics Do the units, losses and limits make sense? Nameplate (the maximum rated power) treated as typical output, or a unit that cannot be right.
4. Who pays Who pays, who benefits, who carries the risk? Savings for one group that other customers end up paying for.
5. Boundary What is counted, and what is left out? A total that quietly leaves out part of the system.
6. Counterfactual Compared with what? A change measured against a baseline that would never have happened.
7. Uncertainty How sure, says who, with what range? One number from one party, with no range.

Story 1 · Data centers as a load · United States

“U.S. data centers could use nearly 12% of the nation's electricity by 2030” Our paraphrase of how such reports get headlined.

A national-laboratory report estimates that data centers used 192 TWh in 2024, or 4.7% of U.S. electricity. Its Reference Case puts 2030 use at 649 TWh, within a range of 521–843 TWh. Serving that load would take about 148 GW of grid connections. Cryptocurrency mining is not counted.

A researcher's notes on story 1
  • 192 TWh in 2024 (Time): Energy over a whole year, not power. Spread over 8,760 hours it is about 22 GW of average demand.
  • 4.7% of U.S. electricity (Scale): A share of one country's use. Read it next to the whole, and never mix it with a global share.
  • 2030 use at 649 TWh (Time): A model's projection six years ahead, not a measurement: 2024 is the report's last historical year.
  • within a range of 521–843 TWh (Uncertainty): The authors' own range: its top is 1.6 times its bottom. The IEA's early-2025 projection, about 430 TWh, sits below it, as the report notes (p. 11).
  • about 148 GW of grid connections (Physics): Capacity, not energy: the report turns 649 TWh into GW by assuming connections are used at 50% on average.
  • Cryptocurrency mining is not counted (Boundary): What counts as a data center is a choice; a study that includes mining reports a larger number.
  • Left unanswered (Who pays): Who pays for the new lines and plants that serve large loads depends on each utility's tariff rules, and the story does not say. Analyst lesson A4 covers large-load tariffs.
  • A researcher's take: I accept the direction, strong growth in U.S. data-center use by 2030, but not the exact 11.8%. A new estimate built on lower chip shipments or lower server power would change my mind. Confidence: medium, since credible projections still differ by about 50%.

Source: LBNL, United States Data Center Energy Usage Report: 2025 Update, pp. 5, 8–11 (PDF) (published 18 June 2026)

Story 2 · Storage · United States

“A record 15 GW of batteries joined the U.S. grid in 2025” Our paraphrase of how the data release was headlined.

Developers added a record 15 GW of battery storage in 2025, according to preliminary data. At the start of the year, 18.2 GW had been planned. Solar led with 27.2 GW of 53 GW in total. The count covers utility-scale plants of 1 MW and up.

A researcher's notes on story 2
  • a record 15 GW of battery storage (Time): GW is power: how fast the batteries can deliver, not how much energy they hold. Energy in GWh needs the duration, which the story does not give.
  • according to preliminary data (Uncertainty): Preliminary counts come from a monthly inventory and are revised when the final annual data arrive.
  • 18.2 GW had been planned (Uncertainty): Planned is not built: actual additions were about 82% of the amount first planned (15 ÷ 18.2).
  • Solar led with 27.2 GW (Physics): A GW of solar makes less energy in a year than a GW of a plant that runs most hours. Compare capacity factors (average output ÷ rated power), not GW.
  • 53 GW in total (Scale): Big for one year. For scale, U.S. use averaged about 480 GW across 2025 (4.20 trillion kWh ÷ 8,760 h).
  • The count covers utility-scale plants of 1 MW and up (Boundary): Home batteries and rooftop solar sit outside this count.
  • Left unanswered (Who pays): Who pays for the batteries, and who earns from them, depends on market rules and contracts; the story does not say. Analyst lesson A2 covers contracts and risk.
  • A researcher's take: I accept that a record amount of battery power, about 15 GW, came online in 2025, as a preliminary count. Final annual data showing a much lower figure would change my mind. Confidence: high for the power; the energy stored cannot be known from this story.

Source: EIA, Today in Energy: 2025 additions, preliminary (20 February 2026); EIA, Today in Energy: additions planned for 2025 (24 February 2025); EIA, Energy Explained: utility-scale means 1 MW or more (checked 2 November 2026); EIA, Energy Explained: U.S. use 4.20 trillion kWh in 2025 (data for 2025); EIA, Today in Energy: battery duration = MWh ÷ MW (25 March 2022)

Story 3 · Reliability · Spain and Portugal

“Renewables caused the Iberian blackout” Our paraphrase of a claim that circulated soon after; no single outlet is quoted.

At 12:33 CEST (11:33 in Portugal) on 28 April 2025, the continental power systems of Spain and Portugal collapsed. Early posts and headlines named one cause: renewables. The European grid operators' expert panel found many interacting factors. Among them: gaps in voltage and reactive-power control, power swings (oscillations), and generators tripping off one after another (a cascade).

A researcher's notes on story 3
  • the continental power systems of Spain and Portugal (Scale): The mainland grids of two countries went dark, not one plant or one city: that is why a formal investigation followed.
  • Early posts and headlines named one cause (Time): The claim came before the evidence: the expert panel's factual report took 158 days, its final report 326.
  • renewables (Uncertainty): A single cause from early, partial information. Ask who says it and what data they had.
  • many interacting factors (Boundary): Control rules, regulation practices and protection settings are part of the system, not only the power plants.
  • gaps in voltage and reactive-power control (Physics): Frequency responds to any gap between supply and demand (lesson 0). Voltage also depends on local network conditions and reactive-power control, so a yearly energy total cannot diagnose this blackout.
  • Left unanswered (Counterfactual): Compared with what? The headline skips the test: would the same plants with different control settings, or different plants with the same settings, have held? The panel's event-by-event analysis is how such questions get answered.
  • A researcher's take: I reject the claim that renewables alone caused the blackout. An investigation tracing the collapse to one cause would change my mind. Confidence: high that there was no single cause, because the expert panel found many interacting factors.

Source: ENTSO-E, 28 April 2025 Iberian blackout: expert panel reports (factual report 3 October 2025; final report 20 March 2026)

The three checks with units, worked
  • Story 1: 649 TWh × 1,000 = 649,000 GWh a year; ÷ 8,760 h ≈ 74.1 GW of average power. At 50% average use of the connections (the report's assumption), that needs about 148 GW of connections.
  • Story 2: 15 GW × 4 h (an illustrative duration) = 60 GWh. That energy equals about 7.5 minutes of U.S. use at the 2025 average of about 480 GW, but 15 GW is only about 3% of that average power. Actual additions were about 82% of the 18.2 GW first planned.
  • Story 3: the expert panel's factual report came 158 days after the blackout, and its final report 326 days after: about 11 months.
Units or scale? Six sentences, with answers
  • “The new battery can store 100 MW.” A unit error. Stored energy is in MWh; 100 MW is how fast it charges or discharges. A 100 MW, four-hour battery stores 400 MWh.
  • “U.S. data centers used about 192 TWh of electricity in 2024.” Reads right. TWh over a year is an amount of energy: LBNL's estimate for 2024.
  • “A 13 kWh home battery can run an average U.S. home for a month.” A scale error. An average home bought about 899 kWh a month (EIA, 2022 data), so 13 kWh lasts about 10 hours.
  • “The plant's two 500 MW units will generate 1,000 MW every hour of the year.” A physics error. Nameplate (the maximum rated power) is a ceiling, not an output: plants run below it and stop for maintenance. Average output ÷ nameplate is the capacity factor (lesson 5).
  • “In 2025 the United States used about 4,200 TWh of electricity.” Reads right. 4.20 trillion kWh is 4,200 TWh (EIA, data for 2025).
  • “A town of 10,000 homes needs about 10 GW of power.” A scale error. At EIA's 10,791 kWh a home a year (2022 data), 10,000 homes use about 108 GWh a year: about 12 MW on average. Even a peak a few times higher is far below 10 GW.
The evidence ladder, with the stories' claims placed
  1. Announced: Someone says it will happen.
  2. Funded: The money is committed.
  3. Permitted: Regulators have approved it.
  4. Under construction: Work has started on site.
  5. Operating: It runs and is counted as in service.
  6. At design output: It delivers what it was built to deliver.
  • “15 GW of batteries entered service in 2025 (preliminary count)”: Operating, as counted so far. Whether each battery delivers its rated hours is the next rung up.
  • “18.2 GW of batteries planned for 2025, as reported at the start of the year”: Below operating: plans in EIA's inventory sit on different rungs. Actual additions were about 82% of this planned total.
  • “649 TWh of U.S. data-center use in 2030”: Not on the ladder at all: a projection is a model's output, not a stage of anything being built.

With JavaScript on, you tag each story's phrases yourself, run a check with units, and compare with a researcher's notes.

Headlines are paraphrased in our own words; no outlet is quoted. Facts come from the primary sources linked in each story (LBNL via the U.S. Department of Energy, EIA, ENTSO-E), as of the dates shown, checked on 2 November 2026. Battery durations and connection use in the checks are illustrative.

What you just did, as a checklist and a ladder

The lens
Seven questions, always in the same order: scale, time, physics, who pays, boundary, counterfactual, uncertainty.
Power or energy
GW is a rate and GWh an amount. Multiply a year's TWh by 1,000 to get GWh, then divide by 8,760 hours to get average GW.
The evidence ladder
Announced, funded, permitted, under construction, operating, at design output; a projection is not on the ladder.
Your take
The claim you accept or reject, what would change your mind, and your confidence.
Go deeper · AnalystThree more checks for the next tier
  • Evidence status. Which rung of the ladder is the claim on? A plan, a permit and a running plant are different news.
  • Jurisdiction and date. Rules belong to a place and a year; a tariff or credit that applies in one state, or before a rule change, may not apply elsewhere.
  • Which emissions question? A share of a grid's emissions, the short-run change from adding a load, the long-run change through what gets built, and what a buyer may claim are four questions with four methods.
SteelmanWhy the 2030 data-center projection could come out lower, or higher

The report's own sensitivity scenarios give the strongest case each way (LBNL 2025 Update, p. 5):

  • Lower: fewer IT installations than forecast (578 TWh in 2030), or shorter working lives for specialised chips (590 TWh).
  • Higher: more specialised graphics chips installed (664 TWh), or servers that draw more power when idle and run busier (782 TWh).

Combined, the authors' bounds are 521–843 TWh. A careful reader reports the range and the drivers, not one number.

Go deeper · ModelerWhich model would test each claim?
  • Story 1: a planning model that adds the new load hour by hour and asks what must be built to serve it (M3, capacity expansion).
  • Story 2: a storage dispatch model, where the duration limits how long each battery can discharge (lesson 9 on the course map).
  • Story 3: a power-flow and dynamics study of voltage and protection; the course shows when an energy-only answer fails (A17) but stops short of dynamic simulation.

Modeler's note: at the next tier the drill changes from "which question?" to "which model would test this, and what result would prove it wrong?"

Go deeperWhat this lesson leaves out
  • The three stories were chosen because their primary sources are public and well documented; many headlines are harder to trace.
  • The lens tells you what to ask, not what is true. Answers come from data, models and investigations.
  • Numbers are dated. Estimates get revised, including past years.

Check yourself

Myth or fact? “A fleet of 15 GW of batteries stores 15 GWh.”

  • Myth
  • Fact
Answer

Myth. Only if every battery lasts exactly one hour: energy is power times duration, and the duration is often missing from the story.

Myth or fact? “A projection's range guarantees that the true value will fall inside it.”

  • Myth
  • Fact
Answer

Myth. A range comes from the scenarios its authors chose; reality can land outside it, and later reports revise past years too.

Myth or fact? “A blackout during hours of high renewable output proves that renewables caused it.”

  • Myth
  • Fact
Answer

Myth. Timing is not cause. Investigations rebuild the sequence of events; the Iberian expert panel found many interacting factors.

Myth or fact? “If data centers use 4.7% of U.S. electricity, they use about 4.7% of the world's.”

  • Myth
  • Fact
Answer

Myth. A share depends on the whole you divide by; a national share and a global share answer different questions.

A 2 GW load runs at full power for a whole year. How much energy is that?

  • 2 GWh
  • 17.5 TWh
  • 17,520 TWh
  • 2 TWh
Answer

17.5 TWh: 2 GW × 8,760 h = 17,520 GWh. GW is a rate; a year's amount is in GWh or TWh.

A story says a policy saved money “compared with a year without it”. Which lens question is that phrase about?

  • Scale
  • Boundary
  • Counter­factual
  • Uncertainty
Answer

Counterfactual: it names the baseline. Next, ask whether that baseline would really have happened.

Numbers, sources and what’s next

Numbers to remember

Number What it means Source
8,760 h Hours in a year (not a leap year): 1 GW running all year is 8.76 TWh. arithmetic
192 TWh U.S. data-center electricity use in 2024, about 4.7% of the total. LBNL 2025 Update, p. 8 · published June 2026
15 GW U.S. utility-scale battery power added in 2025, a record (preliminary). EIA, Today in Energy · February 2026
4.20 trillion kWh U.S. electricity use in 2025, about 480 GW on average. EIA, Energy Explained · data for 2025
326 days From the Iberian blackout to the expert panel's final report. ENTSO-E · final report 20 March 2026

Next

CK-F · Foundations checkpoint: estimate an unfamiliar situation

Practise on real news

Pick one item from this site's Field Notes briefings, or from today's news. Run the seven questions on it, then write your three-sentence take.

  • Concepts from lesson 0 (power and energy, frequency) and lessons 1 and 5 (units, capacity factor).
  • Next: A2 and A4 (who pays), R2 (what is already known).

Sources and further reading

Spotted an error, or stuck? Write to hlli@ksu.edu. Corrections are made in place and dated here.

Text licensed CC BY 4.0, code MIT: teachers and students may reuse and adapt it, with credit to Honglin Li.