Synthesis
Where everyone agrees
These show up in essentially every serious 2050 forecast. Treat them as background, not as scenario differentiators:
- Demographic deceleration is locked in. UN, OECD, McKinsey, Zeihan all converge: global fertility at/below replacement by ~2050, world population peaks around 2080 at 10–10.4B (or earlier and lower per Zeihan). Workers per retiree falls everywhere except SSA. Disagreement is on the consequence, not the trajectory.
- The center of economic gravity moves south and east. China big but plateauing; India overtakes; sub-Saharan Africa goes from 3% to 13% of world output. Even pessimists agree.
- Energy decarbonizes — at some speed. Disagreement is whether it’s 35% fossil by 2050 (BCG Climate Coalition), <10% (IEA NZE), or still dominant (Smil). Everyone agrees the direction.
- AI compute is the swing variable. Even the skeptics (Marcus, Smil) frame their case as “AI doesn’t deliver.” It’s now the load-bearing technology in every forecast.
Where the smart forecasters split
This is where the load is. Each axis is one I’d interrogate before adopting any specific 2050 plan.
Axis 1 — AI takeoff: gradual or punctuated?
| Position | Implication for 2050 |
|---|---|
| BCG AI Abundance | 5%/yr growth, 25% fewer work hours, AI as productivity tool |
| McKinsey | 3.3%/yr productivity unlock, sector by sector |
| Andreessen / e/acc | accelerate; ceiling unknown; regulation = the danger |
| Kokotajlo (AI 2027) | superintelligence by 2029; geopolitics catches up after |
| Hanson (Age of Em) | distinct emulation-driven economy; ≫1000% growth episodes |
| Marcus | LLMs don’t scale to AGI; need a paradigm break first |
| Ord | conditional on alignment success — otherwise this axis is moot |
BCG’s four scenarios all assume the gradual end of this spectrum. If Kokotajlo or Hanson is right by 50%, none of BCG’s scenarios apply.
Axis 2 — Globalization: trade share at 2050
| Position | Trade as % of GDP |
|---|---|
| Status quo extension | ~55–60% |
| BCG Climate Coalition / AI Abundance | ~55–65% |
| BCG Battling Blocs | 35% |
| NIC Separate Silos | bloc-regionalized, no global figure |
| Zeihan | substantially lower; some countries lose access entirely |
BCG’s 35% floor is Zeihan’s expected case, not his downside. The pessimistic geopolitical voices are systematically more pessimistic than BCG’s pessimistic scenario.
Axis 3 — Energy: how cheap and how fast?
| Position | 2050 picture |
|---|---|
| Smil | fossil still ≥60% of primary energy; transition stalls |
| BCG Battling Blocs | slow transition |
| IEA NZE2050 | electricity 50% final, ⅔ useful; carbon removal scales |
| BCG Climate Coalition | fossil 35%, low-carbon near-total in electricity |
| BCG AI Abundance | ”abundant low-carbon energy” via AI-designed materials |
| Handmer | 95% of energy downstream of solar PV by 2042 |
Handmer’s position is the one BCG entirely omits: energy stops being a constraint, by economics, well before 2050. That’s a different world.
Axis 4 — Inequality and political stability
BCG Digital Darwinism: top 1% holds 50% of wealth. WEF: inequality is the most-interconnected risk. Ord: distribution of catastrophe risk is itself unequal. None of the scenarios examine what level of inequality breaks legitimacy and triggers something other than persistence. The Hickel/degrowth scenario lives here.
Axis 5 — Catastrophe as scenario vs. as risk
BCG models risks; NIC’s Tragedy and Mobilization models the catastrophe happening. Ord assigns 1-in-6 to existential catastrophe this century. The honest forecaster has to choose between modeling a median 2050 (assumes you got there) and modeling the path-with-catastrophes (assumes some scenarios don’t have a “2050” worth describing).
Five scenarios BCG doesn’t have
Useful for stress-testing strategy against worlds BCG’s four can’t generate:
- Solar-deflation world. Handmer + Andreessen. Energy ≈free by 2040; everything energy-intensive (DAC, desalination, vertical agriculture, propulsion) becomes trivially cheap. Climate solves itself by economics. AI compute is no longer a bottleneck. New constraint: land, materials, biology.
- Demographic-collapse world. Zeihan + UN low-fertility variant + China <0% potential growth (OECD). Aging societies can’t sustain pension systems or military deterrence; trade collapses because demand evaporates in rich aging countries; SSA inherits demographic weight but lacks infrastructure to monetize it for decades. Global GDP per-capita might rise; total GDP shrinks.
- Misaligned-AI world. Ord + Kokotajlo. Takeoff happens, alignment doesn’t. Either a fast catastrophe (1-in-10 per Ord) or a slow disempowerment (“gradual disempowerment” line of thinking). 2050 may not be a useful planning horizon.
- Degrowth-by-choice world. Hickel. Rich countries voluntarily shrink material throughput. GDP per capita falls in OECD, rises in South. Political legitimacy is rebuilt around sufficiency rather than growth. Implausible to BCG’s frame because BCG assumes elite preferences continue.
- Hot-war world. None of BCG’s scenarios model great-power kinetic war over Taiwan, Korean peninsula, eastern Europe, or Arctic. Battling Blocs is the closest, but it explicitly stays cold. A 7%-of-GDP defense world that has fought is a different scenario from one that’s deterring.
What I’d interrogate before adopting any single forecast
- What does this forecast assume about AI capability in 2035? That’s the load-bearing year. If your forecast doesn’t have a clear AI assumption, it’s a forecast of a pre-AI world.
- What does it assume about US-China? Cooperation, cold competition, hot conflict, or one side decisively wins.
- What does it assume happens first — a triggering catastrophe or a managed transition? NIC’s Tragedy and Mobilization vs. Climate Coalition differ only on the trigger event.
- What does it leave constant that other forecasters move? BCG holds fertility roughly constant; Zeihan moves it. BCG holds great-power conflict cold; NIC’s Separate Silos doesn’t. BCG holds the alignment problem solved by default; Ord doesn’t.
Next moves for this directory
If extending this, the high-value additions would be:
- A Chinese-state 2049 vision (CCP white papers, Five Year Plans) — currently summarized but no primary source in
sources/ - The Indian view (NITI Aayog long-range)
- A serious AI-doom case beyond Ord (e.g. Anthropic responsible-scaling policy, MIRI)
- An African 2050 vision (AU Agenda 2063)
- A specifically biology/longevity axis — none of the above forecasters take radical life-extension seriously, but if it happens, it dominates demography