3 The Modern Energy Context
“Energy transitions are protracted affairs that take decades to accomplish, and the greater the scale of the prevailing uses and conversions, the longer the substitution will take.” – Vaclav Smil
“I think the economic, security, and moral reasons for moving away from fossil fuels are overwhelming. The question is how fast we can do it.” – Stephen Chu
The laws of thermodynamics haven’t changed since David MacKay wrote Sustainable Energy Without the Hot Air in 2008. The Carnot limit still governs heat engines. Conservation of energy remains inviolate. The Shockley-Queisser limit still bounds single-junction solar cells.
But almost everything else has changed: the economics, the technology costs, the policy landscape, and the nature of electricity demand itself. This chapter examines the modern energy context through three lenses: the ongoing debate about transition timescales, what MacKay got right and wrong, and the new challenges that have emerged since his analysis.
3.1 Heat Pumps vs. Heat Engines: A Thermodynamic Comparison
Before diving into the policy debates, let’s complete our thermodynamic foundation by carefully distinguishing heat engines from heat pumps. This distinction is central to understanding why electrification of heating is so important.
3.1.1 Heat Engines: Converting Temperature Difference to Work
A heat engine extracts work from the flow of heat from hot to cold:
::::::{.cell config=“{”securityLevel”:“loose”}” layout-align=“default”}
flowchart TB
H["HOT RESERVOIR (T<sub>H</sub>)"]
E["HEAT ENGINE"]
C["COLD RESERVOIR (T<sub>C</sub>)"]
W(("W"))
H -->|"Q<sub>H</sub> (heat in)"| E
E -->|"Q<sub>C</sub> (heat out)"| C
E -->|"work out"| W
::::::
By the First Law: \(Q_H = W + Q_C\)
The efficiency is: \[ \eta = \frac{W}{Q_H} = \frac{Q_H - Q_C}{Q_H} = 1 - \frac{Q_C}{Q_H} \]
And the Carnot limit constrains this: \[ \eta \leq \eta_{Carnot} = 1 - \frac{T_C}{T_H} \]
Key insight: Heat engine efficiency is always less than 100%. You cannot convert all the heat from a hot reservoir into work; some heat must be rejected to the cold reservoir. This is a fundamental consequence of the Second Law, not an engineering limitation.
Examples of heat engines: - Coal, gas, and nuclear power plants (thermal → electrical) - Internal combustion engines (chemical → thermal → mechanical) - Jet engines and rockets (chemical → thermal → kinetic) - Geothermal plants (geothermal heat → electrical)
3.1.2 Heat Pumps: Using Work to Move Heat “Uphill”
A heat pump does the opposite: it uses work to move heat from cold to hot:
::::::{.cell config=“{”securityLevel”:“loose”}” layout-align=“default”}
flowchart BT
C["COLD RESERVOIR (T<sub>C</sub>)"]
P["HEAT PUMP"]
H["HOT RESERVOIR (T<sub>H</sub>)"]
W(("W"))
C -->|"Q<sub>C</sub> (heat extracted)"| P
P -->|"Q<sub>H</sub> (heat delivered)"| H
W -->|"work in"| P
::::::
By the First Law: \(Q_H = W + Q_C\)
The Coefficient of Performance (COP) for heating is: \[ COP_{heating} = \frac{Q_H}{W} = \frac{Q_H}{Q_H - Q_C} \]
And the Carnot limit for a heat pump is: \[ COP \leq COP_{Carnot} = \frac{T_H}{T_H - T_C} \]
Key insight: COP can exceed 1, often by a factor of 3-5. This isn’t magic or perpetual motion. The heat pump doesn’t create energy; it moves heat that was already in the cold reservoir (outdoor air, ground, water) to the hot reservoir (your house). The work input makes this thermodynamically possible.
3.1.3 Why This Matters for Energy Policy
Consider heating a house to 20°C (293 K) when it’s 5°C (278 K) outside:
Option 1: Burn natural gas directly - 95% efficient modern furnace - 1 kWh of gas → 0.95 kWh of heat - COP = 0.95
Option 2: Resistance electric heating - 100% efficient (by First Law) - 1 kWh electricity → 1 kWh of heat - COP = 1.0 - But: if electricity came from a 50% efficient gas plant, primary energy COP = 0.5
Option 3: Heat pump - Carnot COP: \(\frac{293}{293-278} = 19.5\) - Real COP: 3.5-4.5 (achievable with good equipment) - 1 kWh electricity → 4 kWh of heat - Even with 50% efficient gas generation: primary energy COP = 2.0
The heat pump delivers twice as much useful heat per unit of primary energy as burning gas directly. This is why electrification of heating via heat pumps (not resistance heating) is central to decarbonization strategy.
3.1.4 Ground-Source vs. Air-Source Heat Pumps
Ground-source heat pumps (GSHPs) extract heat from the earth: - Ground temperature at 3-6 meters depth is remarkably stable: 10-15°C year-round - Smaller temperature difference means higher COP - Typical heating COP: 4-5 - Higher installation cost (requires drilling or trenching) - Lower operating cost - Best for new construction where ground loops can be installed
Air-source heat pumps (ASHPs) extract heat from outdoor air: - Temperature varies widely with weather - COP drops as outdoor temperature falls - Typical heating COP: 2.5-4 (varies with conditions) - Lower installation cost - Higher operating cost in cold climates - Modern “cold climate” units work down to -15°C or below
For a home requiring 15,000 kWh of heating per year:
Natural gas furnace (95% efficient, gas at $1.20/therm = \(0.041/kWh):\)$ = = $647 $$
Heat pump (COP = 3.5, electricity at \(0.15/kWh):\)$ = = $643 $$
At these prices, the heat pump roughly breaks even on operating costs. With higher gas prices or lower electricity rates, heat pumps win decisively.
But the comparison changes in cold climates where COP drops: - At COP = 2.5: Annual cost = $900 - At COP = 2.0: Annual cost = $1,125
This is why heat pump economics depend heavily on climate and rate structures.
3.2 The Smil Debate: Can We Transition Fast Enough?
Vaclav Smil, the Czech-Canadian polymath at the University of Manitoba, has shaped how many think about energy transitions. Bill Gates has said he learns more from Smil than almost anyone else. But Smil’s conclusions about the pace of energy transition are deeply uncomfortable for many climate advocates.
3.2.1 Smil’s Core Argument
Smil’s position, articulated across dozens of books and papers, can be summarized in four points:
1. Historical transitions take generations
Every major energy transition (wood to coal, coal to oil, oil to natural gas) took 50-70 years to unfold. The shift from 5% to 25% market share typically requires 30-40 years even for technologies that are clearly superior to their predecessors.
Coal began displacing wood around 1800; it didn’t reach 50% of global primary energy until the 1910s. Oil began its rise around 1900; it didn’t surpass coal until the 1960s. Natural gas began significant growth in the 1950s; it still hasn’t surpassed oil.
2. Scale creates inertia
The global energy system represents $20+ trillion in installed capital: - ~60,000 operating fossil fuel power plants - ~1.4 billion vehicles with internal combustion engines - Hundreds of thousands of kilometers of pipelines - Refineries, tankers, storage facilities, distribution networks - Buildings designed for gas heating - Industrial processes optimized for fossil fuel inputs
This infrastructure has decades of useful life remaining. Smil argues it would be “economically unthinkable” for nations and corporations to abandon these investments prematurely.
3. Renewables face physical limits
Wind and solar have inherent characteristics that constrain their deployment: - Intermittency: They produce power only when the sun shines or wind blows - Low power density: Solar farms produce 5-20 W/m2; wind farms 1-3 W/m2. Fossil fuel plants produce 200-2,000 W/m2. Replacing fossil generation requires vastly more land. - Material intensity: Wind turbines and solar panels require more steel, concrete, copper, and specialty materials per unit of energy delivered than fossil fuel plants - Storage requirements: Addressing intermittency requires storage that remains expensive and limited
4. Energy demand keeps growing
Global primary energy consumption grew approximately 2% annually for decades before the pandemic disruption. Developing nations (home to 6+ billion people) reasonably aspire to energy consumption levels comparable to rich countries. Even with efficiency improvements, total energy demand is likely to continue rising.
3.2.2 The Optimist Response
Critics of Smil’s position offer several counterarguments:
1. This time may be different
Previous energy transitions were not driven by existential urgency or coordinated policy. Climate change creates political, economic, and moral pressures that didn’t exist when coal replaced wood. Wars and economic mobilizations have demonstrated that societies can transform rapidly when motivated.
2. Learning curves are unprecedented
Solar PV costs have fallen 99% since 1976 and 89% since 2010. Battery costs have fallen 93% since 2010. Wind turbine costs have fallen 70% over two decades. These learning rates (roughly 20% cost reduction per doubling of cumulative production) exceed any previous energy technology.
Extrapolating from historical transitions may miss this discontinuity. Previous transitions were slow partly because the new technology wasn’t clearly superior economically. Solar and wind are now the cheapest sources of new electricity in most of the world.
3. S-curves accelerate past the inflection point
Technology adoption follows S-curves: slow initial growth, rapid acceleration through the middle, then saturation at the top. Smil’s historical examples may be measuring transitions that were still in early stages.
Solar and wind may now be past their inflection points. Global solar capacity has grown from 40 GW in 2010 to over 1,500 GW in 2024, a 37× increase. If this growth continues even at half the historical rate, solar alone could dominate global electricity generation within two decades.
4. Infrastructure can be stranded deliberately
The $20 trillion in fossil fuel infrastructure is not sacred. Germany closed its nuclear plants in about a decade. The UK shut its coal fleet even faster. Countries have nationalized industries, restructured economies, and mobilized for wars. The question is political will, not technical possibility.
Moreover, infrastructure depreciates. A coal plant built in 1990 will need replacement regardless of climate policy. The choice is whether to replace it with another fossil plant or something cleaner, and economics increasingly favor the latter.
5. The alternative is worse
Climate damages are cumulative and potentially nonlinear. Each year of delayed transition means more warming, more damage, higher eventual costs. The “realistic” slow timeline may actually be more expensive than the “unrealistic” fast one when damages are counted.
3.2.3 Where Smil Has a Point
Regardless of which perspective you find more convincing, Smil highlights real constraints that optimists sometimes underestimate:
Materials and supply chains: Manufacturing a billion EVs requires mining and processing lithium, cobalt, nickel, copper, and rare earths at scales unprecedented in human history. Building that mining and processing capacity takes years, faces environmental opposition, and involves geopolitical complications.
Workforce and skills: Installing solar panels, heat pumps, EV chargers, and grid infrastructure requires trained workers. The electrician shortage in the United States is already acute. Training programs don’t scale overnight.
Permitting and planning: New transmission lines take 7-10 years to permit in the United States. Offshore wind projects face decade-long approval processes. Even solar farms encounter local opposition. The constraint is often not technology or economics but bureaucracy and politics.
Grid integration: High renewable penetration creates technical challenges (the duck curve, frequency stability, voltage regulation) that require grid upgrades, storage, and new operational practices. Grid operators are learning to manage these issues, but the learning takes time.
3.2.4 The Critical Question
The debate between Smil and the optimists ultimately hinges on a question that cannot be answered with physics alone:
Is the skeptical view enabling or paralyzing?
If Smil is right that rapid transition is impossible, then accepting this reality might lead to better policy: focusing on achievable goals, investing in adaptation alongside mitigation, and avoiding promises that cannot be kept.
But if Smil is wrong, if rapid transition is possible but discouraged by the belief that it isn’t, then his skepticism becomes a self-fulfilling prophecy. The transitions that “can’t happen” don’t happen partly because people who could make them happen believe they can’t.
This textbook doesn’t resolve this debate. Instead, we provide analytical tools so you can evaluate claims from all perspectives. When someone says “we can decarbonize by 2050” or “we can’t possibly decarbonize by 2050,” you should be able to:
- Identify the physical, economic, and political assumptions underlying the claim
- Evaluate whether those assumptions are reasonable
- Understand what would have to change for the claim to be true or false
- Recognize where uncertainty is genuine vs. where claims exceed available evidence
3.3 What MacKay Got Right, and What Changed
David MacKay published Sustainable Energy: Without the Hot Air in 2008, making it freely available online. The book became a touchstone for quantitative energy analysis. Sixteen years later, how does it hold up?
3.3.1 MacKay’s Method: Still Valid
MacKay’s core approach was to calculate: - How much energy do we use, in what forms, for what purposes? - How much renewable energy could we physically harvest? - Do the numbers add up without wishful thinking?
This method remains sound. The laws of physics haven’t changed. The importance of quantitative rigor is greater than ever in an era of competing claims and motivated reasoning.
MacKay’s insistence on units, conversions, and sanity checks is exactly what energy discourse needs. “If everyone does a little, we’ll achieve only a little” remains a corrective to symbolic environmentalism.
3.3.2 Technology Costs: Dramatically Wrong
MacKay didn’t emphasize costs; he focused on physical potential. But to the extent he discussed economics, 2008-era assumptions have been demolished:
| Technology | MacKay Era (~2008) | Today (2025) | Change |
|---|---|---|---|
| Solar PV ($/W installed) | $8.00 | $0.70 | -91% |
| Utility solar LCOE ($/kWh) | $0.36 | $0.043 | -88% |
| Li-ion batteries ($/kWh) | $1,200 | $108 | -91% |
| Onshore wind LCOE ($/kWh) | $0.10 | $0.034 | -66% |
| LED lighting ($/k lumen) | $30 | $0.50 | -98% |
MacKay’s physical analysis of solar potential was correct: the sun delivers abundant energy, and PV can capture it. His implicit assumption that solar would remain expensive was wrong. This is perhaps the most consequential prediction failure in energy analysis history, not because MacKay was foolish, but because hardly anyone foresaw how quickly costs would fall.
3.3.3 Policy Landscape: Transformed
The policy context of 2008 and 2025 are almost unrecognizable:
2008: - The Paris Agreement didn’t exist - The U.S. had minimal clean energy tax credits - China was just beginning its solar manufacturing push - EVs were curiosities (Tesla Roadster had just launched) - “Clean coal” was a serious policy proposal - Natural gas was expensive; shale had not yet transformed U.S. energy
2025: - 195 countries have signed the Paris Agreement - The U.S. Inflation Reduction Act offers $369+ billion in clean energy incentives - China manufactures >80% of global solar panels and >75% of batteries - EVs are the fastest-growing vehicle segment globally - Coal is in structural decline in most wealthy countries - Natural gas is abundant and cheap (for better and worse)
3.3.4 What Hasn’t Changed
Despite revolutionary cost declines, several of MacKay’s observations remain valid:
Physics and thermodynamics: Efficiency limits, energy density hierarchies, and conservation laws are eternal. You still cannot extract more than the Carnot limit from a heat engine. Batteries still cannot match hydrocarbon energy density. The sun still doesn’t shine at night.
Intermittency: Solar and wind remain variable. Storage and grid management remain essential challenges. The costs of managing intermittency have fallen, but the challenge hasn’t disappeared.
Land area requirements: Solar and wind require significant land area per unit of average power. Smil’s power density concerns remain valid even as costs have fallen. The land exists, but using it involves siting, permitting, and competing uses.
The need for quantitative analysis: Energy claims still require numerical verification. “Green” is not a number. “Sustainable” is not a number. MacKay’s insistence on arithmetic over adjectives is more relevant than ever.
3.3.5 The U.S. Electrification Puzzle
MacKay’s calculations suggested that electrification of transport and heating was physically feasible and would become economically attractive. Technology has proven him right: - EVs are now cost-competitive with ICE vehicles in total cost of ownership - Heat pumps are cost-competitive with gas furnaces in many climates - The grid can handle electric loads with proper management
Yet U.S. adoption lags other wealthy nations: - EV market share: ~9% in U.S. vs. ~25% in EU, ~35% in China (2024) - Heat pump adoption: ~15% of U.S. homes vs. ~60% in some European countries - Home solar: growing but still <5% of homes
Why? This is where the Framework for Change becomes useful. The path from Principle to Technology to Product is clear. The obstacles lie in Policy and Outcome:
- Split incentives: Renters can’t install solar or heat pumps; landlords don’t pay utility bills
- Infrastructure gaps: Charging networks, grid capacity, skilled installers
- Information barriers: Most consumers don’t know heat pumps exist or how they work
- Financing: Upfront costs remain barriers despite lifetime savings
- Cultural factors: Truck culture, range anxiety, resistance to change
The technology exists. The economics work. Something else is in the way, and understanding that “something else” is essential for effective policy.
3.4 AI and the New Energy Demand
Just as the world debates how to decarbonize electricity, a new source of demand is emerging: artificial intelligence and data centers.
3.4.1 The Scale of Data Center Energy
In 2024, global data centers consumed approximately 415 TWh of electricity, about 1.5% of global electricity consumption. To put this in perspective: - 415 TWh ≈ the annual electricity consumption of the United Kingdom - 415 TWh ≈ output of about 50 large nuclear reactors running continuously - 415 TWh ≈ all U.S. solar generation in 2023
The United States accounts for the largest share: 183 TWh in 2024, more than 4% of U.S. electricity consumption.
3.4.2 AI as the Growth Driver
Not all data center growth is AI, but AI has become the dominant driver. The International Energy Agency calls AI “the most important driver of this growth.”
Key projections:
| Year | Global Data Center Electricity | U.S. Share | AI Share of Data Centers |
|---|---|---|---|
| 2024 | 415 TWh | 183 TWh (44%) | 15-20% |
| 2026 | 590 TWh | 280 TWh | ~30% |
| 2030 | 945 TWh | ~400 TWh | 40-50% |
The growth rate is striking: ~12-15% annually, about 4× faster than overall electricity demand growth.
3.4.3 Why AI Is Energy-Intensive
Training and running AI models requires massive computation:
Training large models: Teaching a model like GPT-4 or Claude requires thousands of GPUs running for months. Estimates suggest training a frontier model can consume 50-100 GWh, equivalent to the annual electricity use of 5,000-10,000 U.S. homes. Next-generation models may require 10× more.
Inference (running models): Every AI query requires computation. A ChatGPT query uses roughly 10× the electricity of a Google search, perhaps 0.001-0.01 kWh per query. With billions of queries daily, this adds up.
Cooling: AI chips run extremely hot. NVIDIA’s H100 GPU consumes 700W; data centers with thousands of these chips require extensive cooling infrastructure, which itself consumes substantial power.
3.4.4 Geographic Implications
Data center construction is concentrated in regions with: - Reliable power supply - Abundant cooling (or cool climates) - Fiber connectivity - Permissive regulations
This is creating localized grid stress. In Northern Virginia, data centers account for >25% of electricity demand. In Ireland, they exceed 20% of national consumption.
Some tech companies are responding by signing power purchase agreements (PPAs) for dedicated clean energy: - Google’s partnership with Fervo Energy for geothermal (24/7 carbon-free) - Microsoft’s contract with Constellation to restart Three Mile Island nuclear plant - Amazon’s investments in utility-scale solar and wind
3.4.5 The Efficiency Question
AI hardware efficiency is improving rapidly. Each new GPU generation offers better performance per watt. Model architectures are becoming more efficient, achieving comparable capabilities with less computation.
But will efficiency outpace demand growth? Historical experience suggests caution. More efficient cars led to more driving. More efficient computing enabled applications (like AI) that increased total compute demand. This “rebound effect” or “Jevons paradox” means efficiency gains don’t automatically reduce consumption.
The energy implications of AI remain deeply uncertain. AI might: - Increase demand through direct power consumption and enabling energy-intensive activities - Decrease demand by optimizing grid operations, building efficiency, industrial processes - Both, with the net effect depending on which applications dominate
If 5 billion AI queries per day each use 0.005 kWh: \[ \text{Daily consumption} = 5 \times 10^9 \times 0.005 = 25 \text{ GWh/day} \] \[ \text{Annual consumption} = 25 \times 365 = 9.1 \text{ TWh/year} \]
That’s just for inference of current models at current query volumes. Training and the growth of both models and usage could multiply this by 10× or more by 2030.
For comparison: 9 TWh is about 0.2% of U.S. electricity generation. But 90 TWh would be 2%, comparable to adding a medium-sized country’s electricity demand.
3.5 Power Density: The Smil Perspective
Vaclav Smil has long emphasized power density, power produced per unit land area (W/m2), as a critical metric that energy discussions often overlook.
3.5.1 The Power Density Gap
Different energy sources deliver vastly different power densities:
| Source | Typical Power Density (W/m2) | Notes |
|---|---|---|
| Solar PV (average) | 5-20 | Varies with latitude, weather |
| Wind farm | 1-3 | Land between turbines often usable |
| Biomass/biofuels | 0.1-0.6 | Includes growing area |
| Hydro (reservoir) | 1-50 | Varies enormously by site |
| Natural gas plant | 200-2,000 | Plant footprint only |
| Coal plant | 100-1,000 | Excluding mining |
| Nuclear plant | 500-1,000 | Plant footprint |
The gap between renewable and fossil power density is roughly 100×. This means replacing a 1 GW gas plant (at 500 W/m2 = 2 km2) with solar (at 10 W/m2) requires 100 km2, which is 50× more land.
3.5.2 Why Power Density Matters
1. Land use competition: Solar and wind farms compete with agriculture, ecosystems, and other uses. While the absolute land requirement for 100% renewable electricity is modest (roughly the area of Maine for the entire U.S.), local siting conflicts are real.
2. Transmission requirements: Low power density means generation is spread over large areas, requiring extensive transmission infrastructure to collect power and deliver it to demand centers.
3. Material intensity: Spread-out generation requires more materials per unit of useful energy: more concrete, steel, copper, aluminum for foundations, towers, wiring.
4. Visual and ecological impacts: Utility-scale solar and wind have visual impacts that some communities resist. They also affect wildlife habitat and migration patterns.
3.5.3 The Counterargument
Critics of the power density framing note:
1. The land exists: The U.S. has 3.8 million km2 of agricultural land, much of which could host solar or wind with minimal interference. Rooftops alone could provide substantial generation.
2. Dual use is possible: Solar panels can coexist with grazing, pollinator habitat, or parking lots. Wind turbines allow continued farming between towers.
3. Fossil fuel power density is misleading: A gas plant’s footprint excludes the land disturbed by drilling, pipelines, and gas processing. Coal excludes mining. Nuclear excludes uranium mining and enrichment. Full-lifecycle land use narrows the gap.
4. Land isn’t the constraint: In practice, economics, permitting, and grid integration (not land availability) limit renewable deployment.
Smil’s power density analysis is physically correct and highlights real constraints. Whether those constraints are binding depends on geography, policy, and values that physics alone cannot determine.
3.6 Trilemma Check: Putting It Together
How do the issues discussed in this chapter map to the Energy Trilemma?
3.6.1 Security
Transition speed and security: Rapid transition could reduce dependence on volatile fossil fuel markets but creates new dependencies on clean energy supply chains (especially China for solar and batteries). Slow transition prolongs fossil fuel dependence.
AI and grid security: Concentrated data center demand creates potential single points of failure. But data centers can also provide grid services (demand response, backup power).
Smil’s concern: Rushing transitions might compromise reliability. Maintaining fossil backup capacity during transition has security value.
3.6.2 Equity
Transition costs and benefits: Who pays for stranded fossil fuel assets? Who benefits from clean energy jobs? Geographic mismatches create winners and losers.
AI energy use: Tech companies consuming vast electricity while global energy poverty persists raises equity questions. Who benefits from AI, and who pays the energy costs?
Heat pump access: Low-income households often can’t afford heat pump installation despite long-term savings. Policy design must address upfront cost barriers.
3.6.3 Sustainability
Transition pace: Every year of delay means more cumulative emissions and more locked-in warming. Smil’s caution about pace has sustainability costs if it slows action.
AI sustainability: If AI accelerates clean energy deployment (through better materials, grid optimization, etc.), it could have net positive climate impact despite direct energy consumption. The net effect is uncertain.
Land use: Solar and wind have lower lifecycle emissions but higher land requirements. Trading carbon emissions for land use change involves sustainability tradeoffs.
3.7 Summary
The modern energy context is characterized by:
Heat pumps vs. heat engines: Heat pumps achieve COP > 1 by moving rather than creating heat, making electric heating potentially more efficient than direct combustion.
The Smil debate: Legitimate disagreement exists about how quickly energy transitions can occur. Historical precedents suggest 50-70 years; learning curves and policy suggest acceleration is possible.
MacKay’s legacy: His physics remains correct, but cost declines have exceeded all 2008-era expectations. The technology-to-policy gap, not physics, now constrains deployment.
AI energy demand: A major new source of electricity demand with uncertain but potentially large growth trajectory.
Power density: Renewable sources require more land per unit power than fossil alternatives, a real constraint though not necessarily a binding one.
The tools of this course (quantitative estimation, framework analysis, trilemma evaluation) are essential for navigating this complex and contested landscape.
3.8 Readings
- smil_power_density.txt: Vaclav Smil’s Power Density Primer (understand W/m2 as key metric)
- smil_net_zero.txt: Smil on net-zero by 2050 (the skeptical case)
- rmi_renewable.txt: RMI’s Renewable Revolution (2023), the optimistic case (read summary + Chapter 1)
3.9 Exercises
Heat Pump Comparison: Calculate the primary energy consumption for heating a home requiring 20,000 kWh/year under three scenarios:
- Natural gas furnace (95% efficient)
- Air-source heat pump (COP = 3.0) with electricity from a 50% efficient gas plant
- Air-source heat pump (COP = 3.0) with electricity from solar PV
Which uses the least primary energy? Which produces the least CO2?
Historical Transitions: Smil argues that energy transitions take 50-70 years. Research the coal-to-oil transition:
- When did oil reach 5%, 25%, and 50% of global primary energy?
- Do Smil’s numbers hold for this case?
- What factors might make the renewable transition faster or slower?
Land for New York: Using power density figures, calculate the land area needed to supply New York City’s electricity (~50 TWh/year) entirely from:
- Solar PV at 15 W/m2 average
- Onshore wind at 2 W/m2 average
- A combination of 70% solar, 30% wind
How do these areas compare to NYC’s land area (780 km2)? What does this imply about where NYC’s power must come from?
AI Energy Projection: If AI-related data center electricity grows at 20% annually from a 2024 base of 80 TWh:
- What will AI electricity consumption be in 2030?
- How many GW of new generation capacity does this represent (at 50% capacity factor)?
- If this is 50% solar and 50% nuclear, how much of each needs to be built?
The Debate: Read sections from both Smil (smil_net_zero.txt) and RMI (rmi_renewable.txt). In 500 words, identify:
- One point where they clearly agree
- Their core disagreement
- What evidence would help resolve the disagreement
MacKay Update: MacKay estimated UK energy consumption at ~125 kWh/person/day. Research current UK energy consumption. Has it changed? What drove any changes?
3.10 Trilemma Analysis
Security: How does the pace of energy transition affect energy security? Consider both the risks of rapid transition (new dependencies, grid reliability) and slow transition (continued fossil fuel dependence, climate damages).
Equity: AI data centers concentrate in wealthy regions while consuming electricity that could serve other needs. Is this an equity concern? How might it be addressed?
Sustainability: If Smil is right that rapid transition is infeasible, what are the sustainability implications? If the optimists are right but we act as if Smil is right, what are the implications?