title: "AI and Nuclear Fusion Vol.7: Geopolitics — Who Builds the Sun, and Why"
emoji: "🌍"
type: "idea"
topics: ["nuclear", "fusion", "geopolitics", "energy", "investment"]
published: true
AI and Nuclear Fusion Vol.7: Geopolitics — Who Builds the Sun, and Why
Series: "Thinking Seriously About Nuclear Fusion with AI"
Volume 7 of 10 | Geopolitics & Energy Strategy
Author: dosanko_tousan × Claude (Anthropic)
License: MIT
Date: 2026-02-15
Executive Summary
Volumes 1–6 of this series established the physics. The conclusions were blunt:
- D-T ignition is within reach — we are a factor of ~2 away in the fusion triple product (Vol.2).
- Tritium does not exist in sufficient quantity — Monte Carlo analysis shows 88% of scenarios yield TBR < 1.0 (Vol.3).
- Materials data stops at 20 dpa; reactors need 100 dpa — a 5× extrapolation into the unknown (Vol.4).
- AI compresses timelines by ~25% but cannot replace missing experiments (Vol.5).
- Advanced fuels (p-¹¹B) cannot thermally ignite, but the calculus reverses for propulsion (Vol.6).
This volume asks: given these constraints, who is building what, with whose money, and why?
The answer reveals a global fusion landscape that has transformed beyond recognition since 2020. Private investment has reached \$15.2 billion globally (F4E Fusion Observatory, September 2025), with the United States and China accounting for 87% of the total. The regulatory environment has decoupled fusion from fission for the first time in history (NRC, 2023). And the convergence of AI-driven energy demand with fusion's promise has created an unprecedented wave of corporate power purchase agreements.
But the physics does not care about investment rounds. The tritium cliff (2045–2050), the materials gap (20 dpa wall), and the breeding ratio crisis (TBR < 1.0 in 88% of scenarios) remain. This volume maps these physical constraints onto the geopolitical chessboard to answer the only question that matters for investors, policymakers, and engineers:
Where should the next dollar, euro, yuan, or yen go?
Document Classification
| Item | Value |
|---|---|
| Series Volume | 7 of 10 |
| Domain | Geopolitics, Energy Policy, Investment Strategy |
| Technical Depth | Policy-quantitative (backed by Vol.1–6 physics) |
| Prerequisites | Executive summaries of Vol.1–6 (embedded above) |
| Word Count Target | ~55,000 characters |
| Figures | 2 (6 panels) |
| Python Scripts | 2 (reproducible, seed-fixed) |
Table of Contents
- §1. The State of Play: Public vs. Private Fusion in 2026
- §2. ITER — The Magnificent Failure That Still Matters
- §3. The National Programs: A Comparative Analysis
- §3.1 United States: Regulatory Revolution and Private Capital
- §3.2 China: State-Driven Speed
- §3.3 European Union: ITER Host, Industrial Latecomer
- §3.4 United Kingdom: Post-Brexit Bet on the Spherical Tokamak
- §3.5 Japan: The Quiet Powerhouse
- §3.6 South Korea: K-DEMO and the KSTAR Legacy
- §4. The Private Fusion Explosion
- §4.1 The Big Three: CFS, Helion, TAE
- §4.2 The New Wave: Pacific Fusion, Proxima, and Others
- §4.3 The TAE–Trump Media Merger: Fusion Meets Wall Street
- §4.4 AI Companies as Fusion Customers
- §5. Fuel Geopolitics: Tritium, Deuterium, and Helium-3
- §5.1 The Tritium Supply Chain
- §5.2 Helium-3 and the Lunar Question
- §5.3 Nuclear Proliferation Considerations
- §6. Energy Security: What Fusion Changes
- §7. Investment Decision Framework
- §8. Uncertainties — The Honest Section
- §9. Decision Matrix
- §10. Conclusion: Who Wins?
- References
§1. The State of Play: Public vs. Private Fusion in 2026
The fusion energy landscape in 2026 looks nothing like it did in 2020. Five years ago, "fusion investment" meant government-funded research programs and ITER. Today, it means venture capitalists, sovereign wealth funds, oil majors, and tech billionaires writing billion-dollar checks to startups that promise electricity by 2030.
The numbers tell the story:
Global Private Fusion Investment (Cumulative):
| Year | Cumulative Funding | Annual New Funding | Number of Companies |
|---|---|---|---|
| 2020 | ~\$1.9B | — | ~20 |
| 2021 | ~\$4.0B | ~\$2.1B | 23 |
| 2022 | ~\$4.8B | ~\$2.8B (record) | 33 |
| 2023 | ~\$6.2B | ~\$1.4B | 43 |
| 2024 | ~\$7.1B | ~\$0.9B | 45 |
| 2025 (Jul) | ~\$9.8B | ~\$2.6B | 53 |
| 2025 (Sep) | ~\$15.2B | (includes CFS B2) | 77 |
Sources: Fusion Industry Association (FIA) annual reports 2021–2025; F4E Fusion Observatory, December 2025.
The jump from \$9.8B (July 2025) to \$15.2B (September 2025) — a \$5.4B increase in three months — is largely driven by CFS's \$863M Series B2 round (August 2025), TAE's Trump Media merger valuation, and a wave of Chinese state-backed investments.
The geographic distribution is telling:
- United States: ~\$8.1B across 42 companies (53% of global total)
- China: ~\$5.1B across 8 companies (34% of global total)
- Europe: ~\$1.3B across 13 companies
- Rest of World: ~\$0.7B (Japan, South Korea, Canada, Australia)
Two models dominate: the U.S. model (94.5% private capital, VC-driven) and the China model (71.2% state-led investment). The EU occupies an uncomfortable middle ground (64% private, 34% public). These structural differences in financing models will determine which nations lead in deployment.
Employment: The fusion industry directly employs 4,607 people globally (FIA, July 2025), with an additional 9,300+ supply chain jobs. This is a 4× increase since 2021. Nearly half (48%) are engineers, a quarter (25%) scientists.
Python: Figure 1 — Global Fusion Investment (click to expand)
"""
Vol.7 Figure 1: Global Fusion Investment — Timeline and Geographic Distribution
Seed-fixed, fully reproducible.
"""
import numpy as np
import matplotlib.pyplot as plt
np.random.seed(42)
fig, axes = plt.subplots(2, 2, figsize=(16, 12))
fig.suptitle('Figure 1: The Fusion Investment Explosion (2020–2025)',
fontsize=16, fontweight='bold', y=0.98)
# ── Panel A: Cumulative Private Fusion Investment ──
ax1 = axes[0, 0]
years = [2020, 2021, 2022, 2023, 2024, 2025]
cumulative = [1.9, 4.0, 4.8, 6.2, 7.1, 9.8] # $B, FIA data
annual_new = [0, 2.1, 2.8, 1.4, 0.9, 2.6] # $B
ax1.bar(years, cumulative, color='#2C5F8A', alpha=0.8, width=0.6, label='Cumulative')
ax1.plot(years, cumulative, 'o-', color='#E74C3C', linewidth=2, markersize=8, label='Trend')
ax1.set_xlabel('Year'); ax1.set_ylabel('Cumulative Investment ($B)')
ax1.set_title('(A) Cumulative Private Fusion Investment', fontweight='bold')
ax1.legend(loc='upper left'); ax1.set_ylim(0, 12); ax1.set_xticks(years)
for y, v in zip(years, cumulative):
ax1.text(y, v + 0.25, f'${v}B', ha='center', fontsize=9, fontweight='bold')
ax1.grid(axis='y', alpha=0.3)
# ── Panel B: Annual New Funding ──
ax2 = axes[0, 1]
colors_annual = ['#95A5A6', '#3498DB', '#E74C3C', '#F39C12', '#95A5A6', '#2ECC71']
ax2.bar(years, annual_new, color=colors_annual, alpha=0.85, width=0.6)
ax2.set_xlabel('Year'); ax2.set_ylabel('Annual New Funding ($B)')
ax2.set_title('(B) Annual New Investment', fontweight='bold')
ax2.set_ylim(0, 3.5); ax2.set_xticks(years)
for y, v in zip(years, annual_new):
if v > 0: ax2.text(y, v + 0.08, f'${v}B', ha='center', fontsize=9, fontweight='bold')
ax2.axhline(y=np.mean(annual_new[1:]), color='red', linestyle='--', alpha=0.5,
label=f'Mean: ${np.mean(annual_new[1:]):.1f}B')
ax2.annotate('Record year\n(CFS $1.8B)', xy=(2022, 2.8), xytext=(2022.5, 3.2),
arrowprops=dict(arrowstyle='->', color='#E74C3C'),
fontsize=9, color='#E74C3C', fontweight='bold')
ax2.legend(); ax2.grid(axis='y', alpha=0.3)
# ── Panel C: Geographic Distribution (Sep 2025) ──
ax3 = axes[1, 0]
countries = ['USA\n($8.1B)', 'China\n($5.1B)', 'Europe\n($1.3B)', 'Other\n($0.7B)']
amounts = [8.1, 5.1, 1.3, 0.7]
colors_geo = ['#2C5F8A', '#E74C3C', '#F39C12', '#95A5A6']
wedges, texts, autotexts = ax3.pie(amounts, labels=countries, colors=colors_geo,
explode=(0.05, 0.05, 0, 0), autopct='%1.0f%%',
startangle=90, textprops={'fontsize': 10})
for at in autotexts: at.set_fontweight('bold'); at.set_fontsize(11)
ax3.set_title('(C) Geographic Distribution (Sep 2025, $15.2B total)', fontweight='bold')
# ── Panel D: Funding Models ──
ax4 = axes[1, 1]
categories = ['USA', 'China', 'EU']
private_pct = [94.5, 28.8, 64.0]
public_pct = [5.5, 71.2, 33.9]
ppp_pct = [0, 0, 2.1]
x = np.arange(len(categories)); width = 0.5
ax4.bar(x, private_pct, width, label='Private', color='#2C5F8A', alpha=0.85)
ax4.bar(x, public_pct, width, bottom=private_pct, label='Public/State', color='#E74C3C', alpha=0.85)
bottom_ppp = [a + b for a, b in zip(private_pct, public_pct)]
ax4.bar(x, ppp_pct, width, bottom=bottom_ppp, label='Public-Private', color='#F39C12', alpha=0.85)
ax4.set_xlabel('Region'); ax4.set_ylabel('Funding Mix (%)')
ax4.set_title('(D) Funding Models: Private vs. State', fontweight='bold')
ax4.set_xticks(x); ax4.set_xticklabels(categories); ax4.legend(); ax4.set_ylim(0, 110)
ax4.text(0, 50, '94.5%\nPrivate', ha='center', va='center', fontsize=10, fontweight='bold', color='white')
ax4.text(1, 15, '28.8%\nPrivate', ha='center', va='center', fontsize=10, fontweight='bold', color='white')
ax4.text(1, 65, '71.2%\nState', ha='center', va='center', fontsize=10, fontweight='bold', color='white')
ax4.text(2, 35, '64%\nPrivate', ha='center', va='center', fontsize=10, fontweight='bold', color='white')
ax4.grid(axis='y', alpha=0.3)
plt.tight_layout(rect=[0, 0, 1, 0.95])
plt.savefig('vol7_fig1_investment.png', dpi=200, bbox_inches='tight',
facecolor='white', edgecolor='none')
plt.close()
Figure 1: (A) Cumulative private fusion investment 2020–2025. (B) Annual new investment showing 2022 record and 2025 recovery. (C) Geographic distribution as of September 2025. (D) Funding models: the US relies on private capital (94.5%), China on state investment (71.2%), the EU on a hybrid model. Data: FIA 2025, F4E Fusion Observatory.
The private sector's speed advantage is unmistakable. CFS went from MIT spinout (2018) to \$3B in total capital and first-plasma targeting for SPARC (2026) in eight years. ITER went from initial agreement (2006) to its current state — 85% complete toward a first plasma that keeps receding into the future — in twenty.
But speed and capital do not guarantee success. The physics does not negotiate. The constraints documented in Volumes 1–6 apply equally to CFS's SPARC, China's CFETR, and Helion's Polaris. The question is who will confront those constraints most honestly, and who will burn through capital pretending they do not exist.
§2. ITER — The Magnificent Failure That Still Matters
ITER is the most expensive science experiment in human history. It is also, depending on whom you ask, either humanity's best hope for proving fusion at scale or a cautionary tale of international bureaucracy run amok.
The facts:
ITER Timeline Drift:
| Milestone | Original Target | 2016 Baseline | 2024 Baseline |
|---|---|---|---|
| First Plasma | 2016 | Dec 2025 | 2033–2034 |
| Full Current (15 MA) | — | — | 2036 |
| D-D Operations | — | 2028 | 2035 |
| D-T Operations | 2023 | 2035 | 2039 |
| Budget (construction) | €5B (2006) | €20B | €22B+ |
The U.S. Department of Energy has estimated total costs (including in-kind contributions from all members) at \$45–65 billion, though ITER disputes this figure.
What went wrong:
Geometric non-conformities in vacuum vessel bevel joints. Chloride corrosion cracking in thermal shield cooling pipes — 23 km of pipes requiring replacement. COVID-19 supply chain disruptions. And the fundamental coordination problem: seven member parties (EU, China, India, Japan, Russia, South Korea, USA), 35 nations, each responsible for different components, attempting synchronization across different industrial cultures, regulatory environments, and political cycles.
The UK's departure from Euratom (September 2023) added further complexity. Switzerland rejoined via Euratom in 2026 following negotiations. Russia's continued participation — despite U.S., EU, and South Korean sanctions — demonstrates that ITER occupies a unique diplomatic space where scientific collaboration survives geopolitical fracture.
Why ITER still matters:
Despite the delays, ITER produces irreplaceable value:
Supply chains: The act of building ITER has created global supply chains for superconducting magnets, cryogenic systems, vacuum vessels, and remote handling equipment that every private fusion company will rely on. Toshiba, Mitsubishi Heavy Industries, and Hyundai Heavy Industries have all developed critical manufacturing capabilities through ITER contracts.
Regulatory precedent: ITER's interaction with the French nuclear safety authority (ASNR) is creating the first industrial-scale safety framework for fusion facilities.
15 MA plasma at scale: No private company is attempting what ITER will do — a 840 m³ plasma volume, 500 MW of fusion power for 400 seconds. When (if) ITER achieves D-T operations in 2039, it will provide the first dataset for burning plasma physics at power-plant-relevant scale.
Materials irradiation data: ITER will generate the neutron flux data that Vol.4 documented as critically missing. This data cannot be simulated; it must be measured.
The counterargument is equally valid: by 2039, when ITER begins D-T operations, private companies expect to have already demonstrated net electricity. If CFS, Helion, or a Chinese program achieves this first, ITER's relevance shifts from "pathfinder" to "scientific validation facility." Important, but not leading.
The honest assessment: ITER is a monument to international cooperation and a cautionary tale of megaproject management. It is not a failure — it is generating vast engineering knowledge. But it has lost the race to be first. The question is whether the data it eventually produces is still needed by the time it arrives.
§3. The National Programs: A Comparative Analysis
§3.1 United States: Regulatory Revolution and Private Capital
The United States has made a series of policy decisions since 2023 that collectively represent the most significant shift in fusion regulation in history.
The NRC Decision (April 2023): The Nuclear Regulatory Commission voted to regulate fusion energy systems under 10 CFR Part 30 — the byproduct materials framework used for particle accelerators — rather than Parts 50/52, which govern fission reactors. This single decision removed fusion from the regulatory burden that has made fission plant construction in the U.S. effectively impossible for decades.
The implications are transformative:
- Fusion facilities can be licensed by Agreement States (39 states as of 2025), enabling faster permitting
- No criticality analysis required (fusion cannot achieve criticality)
- No meltdown-scenario analysis (plasma ceases without active confinement)
- Emergency preparedness requirements scale to actual radiological hazard (minimal)
The ADVANCE Act (July 2024): Codified the NRC's decision into law, formally defining "fusion machines" as distinct from fission reactors. Required the NRC to develop design-specific licensing frameworks for mass-manufactured fusion machines by July 2025.
The NRC Proposed Rule (May 2025): Published specific regulatory framework for fusion machines. Final rule expected October 2026.
The DOE Fusion Science & Technology Roadmap (October 2025): Introduced a "Build–Innovate–Grow" strategy targeting commercial fusion energy by the mid-2030s. The roadmap explicitly ties fusion to the AI-driven surge in electricity demand, a strategic framing that has bipartisan support.
Bipartisan support: Fusion enjoys rare cross-party backing. The Inflation Reduction Act's 45Y and 48E clean energy tax credits — retained by the Trump administration in the One Big Beautiful Bill Act — apply to fusion. Energy Secretary Chris Wright declared fusion a priority in his February 2025 Secretarial Order. The DOE created a dedicated Office of Fusion in 2025 via bipartisan legislation (Sens. Padilla and Cornyn).
Federal funding: DOE's Fusion Energy Sciences program allocated \$128 million for Fusion Innovation Research Engine (FIRE) Collaboratives in September 2025. Total federal fusion R&D spending remains modest compared to China, but the U.S. strategy relies on private capital doing the heavy lifting — a deliberate policy choice.
The risk: 94.5% private capital means fusion development is subject to venture capital cycles. If key milestones (CFS's SPARC first plasma, Helion's Polaris results) disappoint, funding could evaporate. The U.S. model is high-ceiling, low-floor.
§3.2 China: State-Driven Speed
China's fusion program is the single largest competitive threat to Western fusion leadership, and it is accelerating.
Key facilities:
- EAST (Experimental Advanced Superconducting Tokamak, Hefei): Set a world record in January 2025 — sustained high-confinement plasma at over 100 million °C for 1,066 seconds. This demonstrates steady-state operational capability that no other device has matched.
- HL-3 (Chengdu, commissioned 2020): Second-generation tokamak providing supplementary plasma physics data.
- CRAFT (Comprehensive Research Facility for Fusion Technology): A \$570 million, 40-hectare facility near EAST, nearing completion. Develops engineering subsystems for future reactors.
- BEST (Burning Plasma Experimental Superconducting Tokamak): Under construction. Targets Q = 5 with D-T fuel. Significantly larger than EAST. This is China's bridge to CFETR.
- CFETR (China Fusion Engineering Test Reactor): The crown jewel. Comparable in size to ITER, but designed for steady-state power production. Two phases: Phase 1 produces 200 MW fusion power with TBR > 1; Phase 2 scales to 1 GW. Engineering design completed 2022. Construction expected late 2020s, operational by 2030s.
Financial commitment: China has invested approximately \$10 billion in fusion since 2019. The Chinese Academy of Engineering's 2024 report set a strategic goal of commercial fusion energy supply by 2040. CNNC (China National Nuclear Corporation) established a national industrial consortium — China Fusion Energy — bringing together 25 state-owned enterprises, 4 universities, and key private firms.
Private sector: China's fusion private sector is smaller but growing rapidly. Energy Singularity (Shanghai), StarTorus Fusion, and HHMAX-Energy are developing HTS tokamaks and FRCs, backed by a mix of government venture funds and state capital. Total Chinese private fusion investment reached approximately ¥32.7 billion (~\$5.1 billion) as of September 2025, accounting for 34% of global private fusion funding.
China's model: 71.2% state-led. This gives China advantages the U.S. model cannot match: long-term planning certainty, coordinated industrial mobilization, and the ability to absorb short-term setbacks without losing investor confidence. But it also means less diversity in technological approaches and a higher risk of groupthink.
The competitive dynamic: A U.S. House Science Committee hearing (September 2025) explicitly framed fusion as a U.S.–China competition. The FIA has warned that China is investing more than triple the U.S. on fusion deployment, spending at least \$6.5 billion between 2023 and 2025. The phrase used was: "The US simply cannot afford to lose the race."
The honest assessment: China has the most credible government-backed pathway from experimental reactor (EAST) through intermediate facility (BEST) to demonstration reactor (CFETR) to commercial power (2040). It also has an EAST record (1,066 seconds) that proves operational endurance. The risk is that CFETR faces the same materials and tritium constraints documented in Vol.3 and Vol.4 — and state-driven programs tend to under-report failures.
§3.3 European Union: ITER Host, Industrial Latecomer
The EU contributes almost half of ITER's construction costs — making it the largest single financial contributor to fusion research globally. But ITER's delays have left Europe in an awkward position: massive sunk costs in a project that will not produce D-T results until 2039, and a private fusion sector that is smaller than either the U.S. or China's.
Key EU assets:
- ITER hosting (Cadarache, France): The supply chain, engineering expertise, and institutional knowledge concentrated in southern France are genuinely world-class. F4E (Fusion for Energy) manages Europe's ITER contribution.
- EUROfusion: Coordinates EU fusion research across 30+ institutions. Managing the JT-60SA collaboration with Japan.
- Proxima Fusion (Germany): Europe's stellarator champion. Raised ~€130 million (~\$150 million) in 2025, targeting a stellarator-based power plant operational in the 2030s. AI-driven coil optimization makes this approach newly viable (Vol.5).
- Marvel Fusion (Germany): Raised €113 million Series B in 2025. Pursuing laser-driven inertial confinement.
EU funding model: 64% private, 34% public, 2.1% public-private. The EU, UK, Germany, and South Korea are collectively investing \$9 billion in fusion.
DEMO: The EU's planned demonstration power plant (successor to ITER) is not expected to begin engineering design until 2029. This timeline trails China's CFETR by nearly a decade.
The gap: Europe has the scientific talent and institutional depth but lacks the venture capital ecosystem of the U.S. and the state coordination of China. Proxima Fusion and Marvel Fusion are bright spots, but they are competing against CFS (\$3B) and TAE (\$1.8B) with a fraction of the capital.
§3.4 United Kingdom: Post-Brexit Bet on the Spherical Tokamak
The UK has made a calculated bet: leave Euratom, forgo direct ITER participation, and invest domestically in STEP — the Spherical Tokamak for Energy Production.
STEP program:
- Target: First prototype fusion power plant generating ~100 MWe by ~2040
- Design: Spherical tokamak (more compact than conventional tokamak)
- Location: West Burton, Nottinghamshire — a former coal power plant site ("from fossil fuel to fusion energy")
- Funding: UK Government announced £2.5 billion in additional investment (2025). Total government funding through 2025–26: £410 million for STEP and UKAEA R&D
- Delivered by: UK Industrial Fusion Solutions (UKIFS), subsidiary of UKAEA
- Timeline: Phase 1 (2019–2024) concept design complete; Phase 2 (2025–2032) detailed engineering design and site preparation; public consultation January–March 2026
Complementary assets:
- MAST Upgrade: The world's most advanced spherical tokamak, providing physics data for STEP design
- JET legacy: Though JET has been decommissioned, the UK retains unique tritium handling experience from decades of D-T experiments
- LIBRTI: New fusion fuel R&D facility. UKAEA–Eni partnership to build the world's largest tritium fuel cycle facility in the UK
- Culham AI Growth Zone: UK's first AI Growth Zone, leveraging the 400 kV grid connection from decommissioned JET
The logic: The UK has recognized that it cannot compete with U.S. capital markets or Chinese state investment. Instead, it is betting that the spherical tokamak — smaller, cheaper to build, higher power density — offers a more commercially deployable design than the conventional tokamak. If STEP works, the UK plans a fleet of fusion plants and an export industry.
The risk: The spherical tokamak has never operated with D-T fuel. MAST Upgrade uses deuterium only. The jump from spherical tokamak physics to a power-producing plant is technically daunting. STEP relies heavily on extrapolating from MAST Upgrade data, and Vol.4's materials constraints apply in full.
§3.5 Japan: The Quiet Powerhouse
Japan does not get the headlines of CFS or the competitive framing of China, but it possesses arguably the deepest institutional fusion expertise in the world.
Key assets:
- JT-60SA (Naka): The world's largest operating superconducting tokamak as of 2024. First plasma achieved October 2023. Upgrades in progress for first full experiments in mid-2026, including U.S.-supplied diagnostics from PPPL and General Atomics. Plasma volume: 135 m³. Designed to explore advanced plasma configurations and inform both ITER and future DEMO designs.
- QST (National Institutes for Quantum Science and Technology): Japan's primary fusion institution. Manages JT-60SA and Japan's ITER contributions.
- DEMO target: Japan aims to build a demonstration fusion power plant by 2050. Japan's DEMO engineering design is expected to begin around 2025.
- Kyoto Fusioneering: Private Japanese fusion startup focused on fusion engineering solutions (heat exchangers, fuel cycle technology). Raised \$79 million.
- Helical Fusion: Japan-based stellarator startup that successfully demonstrated HTS coils in 2025.
Japan's approach: Japan is the only nation simultaneously operating the world's largest superconducting tokamak (JT-60SA), contributing major components to ITER (toroidal field coils via Toshiba and MHI), and developing private fusion startups. This three-pronged approach — domestic experiment, international collaboration, and private innovation — provides a diversified portfolio.
JT-60SA significance: JT-60SA's power density (power per unit volume) is exceptionally high, allowing exploration of plasma behaviors at conditions ITER will not reach for years. The 2026 experimental campaign will be the first data from a large superconducting tokamak since JET's final D-T campaign.
The honest assessment: Japan's fusion program is underfunded relative to China and undersold relative to the U.S. But its institutional depth, engineering precision, and bridging role between ITER and DEMO make it indispensable. The 2026 JT-60SA results could be the most important fusion data produced anywhere in the world this year.
§3.6 South Korea: K-DEMO and the KSTAR Legacy
South Korea's KSTAR (Korea Superconducting Tokamak Advanced Research) achieved high-performance plasma milestones in recent years and provides a foundation for Korea's planned K-DEMO demonstration reactor. Korea is an ITER member and contributes critical components including vacuum vessel sectors manufactured by Hyundai Heavy Industries.
Korea's fusion strategy combines ITER participation with a domestic pathway toward K-DEMO, targeted for the 2040s. The Korea Institute of Fusion Energy (KFE) manages the national program.
§4. The Private Fusion Explosion
§4.1 The Big Three: CFS, Helion, TAE
Three private companies have raised over \$1 billion each and represent the leading edge of commercial fusion.
Commonwealth Fusion Systems (CFS)
- Total raised: ~\$3 billion (including \$863 million Series B2, August 2025)
- Technology: Compact tokamak with high-temperature superconducting (HTS) magnets
- Key milestone: HTS magnet demonstration (2021) — produced 20 tesla field, validating the core enabling technology
- SPARC: Under construction in Devens, Massachusetts. Targets first plasma by end of 2026. Goal: Q > 2 (net energy gain from plasma). If successful, this would be the first privately-built device to achieve net energy
- ARC: First grid-scale fusion power plant, planned for Chesterfield County, Virginia. Target: 400 MW, power to grid in early 2030s. Chesterfield County planning commission unanimously approved conditional use permit
- Customers: Google signed a power purchase agreement for 200 MW from the ARC plant (June 2025)
- Investors: Google, Breakthrough Energy Ventures, Emerson Collective, Tiger Global, Morgan Stanley (Counterpoint Global), Gates Frontier, Eni, NVIDIA, Khosla Ventures, Stanley Druckenmiller, Hostplus Superannuation Fund
CFS has the most capital, the most credible near-term milestone (SPARC first plasma), and the first signed commercial PPA. It is also the most conventionally fusion-physics-grounded approach: a tokamak, using proven confinement geometry, differentiated by HTS magnets that allow a much smaller and cheaper device than ITER.
Helion Energy
- Total raised: ~\$1.03 billion (including \$425 million Series F, January 2025)
- Technology: Field-reversed configuration (FRC), pulsed approach. Does not require steady-state confinement
- Key milestone: Polaris (7th generation prototype) turned on January 2025
- Commercial plant: Being built in Malaga, Washington, to supply electricity to Microsoft. Target: 2028
- Fuel: D-³He (not D-T), with direct energy conversion. If achieved, this eliminates the tritium breeding problem entirely
- Customers: Microsoft (50 MW PPA), Nucor (500 MW fusion power plant for steel production)
- Investors: Sam Altman, Reid Hoffman, KKR, BlackRock, Peter Thiel (Mithril Capital), Capricorn Investment Group
Helion is the most aggressive on timeline and the most radical on technology. The D-³He fuel choice, if viable, would bypass the tritium crisis entirely — but Vol.6 showed that D-³He ignition is 17× harder than D-T and still produces 3–10% neutrons. Helion's pulsed approach avoids the steady-state ignition requirement, but "producing electricity by 2028" is a claim that Senate Energy Committee Chair Joe Manchin called "almost too good to be true."
TAE Technologies
- Total raised: ~\$1.79 billion (pre-merger)
- Technology: Field-reversed configuration with beam-driven plasma. Pursuing aneutronic p-¹¹B fusion
- Key milestone: Norman (5th generation) achieved >100 million °C. Copernicus (6th generation) under construction in Irvine, California
- Commercial plant: First utility-scale plant (50 MWe), construction planned 2026. Target: mid-2030s
- The Trump Media merger: In December 2025, TAE announced an all-stock merger with Trump Media & Technology Group, valuing the combined entity at \$6 billion. TAE receives \$200 million at signing plus \$100 million upon SEC filing. Merger completion targeted mid-2026. This would create one of the world's first publicly traded fusion companies.
- Investors: Google (since 2014), Chevron, Sumitomo Corporation
TAE's p-¹¹B approach is the most physically ambitious — Vol.2 proved that p-¹¹B thermal ignition is impossible at any temperature because bremsstrahlung radiation exceeds fusion power. TAE's bet is that beam-driven, non-thermal plasmas can circumvent this limit. If they are right, it changes everything — truly aneutronic fusion with direct energy conversion. If they are wrong, the \$6 billion valuation evaporates.
§4.2 The New Wave: Pacific Fusion, Proxima, and Others
Pacific Fusion: \$900 million Series A (November 2024) — one of the largest first rounds in fusion history. Inertial confinement fusion using coordinated electromagnetic pulses (not lasers). Led by Eric Lander (Human Genome Project) and Will Regan. 156 impedance-matched Marx generators producing 2 TW for 100 ns, pulses must converge simultaneously. This is NIF-style physics, redesigned from the ground up for energy production.
Proxima Fusion (Germany): ~€130 million raised. Europe's leading stellarator startup. Spun out of Max Planck Institute for Plasma Physics (home of Wendelstein 7-X). AI-optimized stellarator coil design. Targeting power plant in the 2030s. The stellarator approach avoids disruptions (the tokamak's Achilles heel) but has historically been dismissed as too complex to build. AI optimization (Vol.5) may have changed this calculus.
SHINE Technologies: Exceeded \$400 million in funding. Unlike power-only startups, SHINE generates actual revenue from non-power fusion applications — medical isotopes, industrial inspection, defense. This diversified model provides cash flow independent of the fusion-power timeline.
General Fusion (Canada): Hit a rough patch in 2025. Ran short of cash while building LM26, laid off 25% of staff. Later raised \$51.1 million in SAFE notes from ~70 investors. Magnetized target fusion approach. Total raised: \$492 million. A cautionary tale: even well-funded fusion startups can face existential funding crises.
Zap Energy: Funded by Chevron via Chevron Technology Ventures. Developing a next-generation sheared-flow-stabilized Z-pinch reactor. Compact and potentially very cheap if it works.
Thea Energy: Completed preconceptual design of its Helios fusion power plant (2025). First DOE Milestone Program awardee to complete a fusion power plant design review.
§4.3 The TAE–Trump Media Merger: Fusion Meets Wall Street
The December 2025 merger announcement between TAE Technologies and Trump Media & Technology Group deserves separate analysis because it represents a new phase in fusion commercialization: the financialization of fusion.
The merger structure:
- All-stock transaction, combined company valued at \$6 billion
- TAE CEO Michl Binderbauer becomes co-CEO alongside Devin Nunes
- TAE receives \$200 million cash at signing + \$100 million upon SEC filing
- Upon completion (targeted mid-2026), the combined entity would be publicly traded
What this means:
- Public market access for fusion: For the first time, retail investors can take a direct position in a fusion energy company. This changes the capital formation dynamics entirely.
- Valuation anchor: The \$6 billion valuation, while backed by TAE's technology portfolio, also includes Trump Media's existing market dynamics (brand value, political ecosystem). This makes it difficult to assess the "pure fusion" valuation.
- Political dimension: With President Trump's company as the merger partner, TAE gains political insulation that no other fusion company has. This could accelerate permitting, DOE engagement, and defense applications.
- Precedent for IPOs: If the TAE–Trump Media entity succeeds as a public company, expect CFS and Helion to pursue their own public listings within 2–3 years.
The physics risk remains unchanged: p-¹¹B thermal ignition is impossible (Vol.2). TAE must demonstrate that beam-driven non-thermal approaches work at commercial scale. A \$6 billion valuation does not change the bremsstrahlung barrier.
§4.4 AI Companies as Fusion Customers
The convergence of AI and fusion is not hypothetical — it is happening through signed contracts:
| AI/Tech Company | Fusion Partner | Agreement | Details |
|---|---|---|---|
| CFS | PPA | 200 MW from ARC plant (Virginia), early 2030s | |
| TAE | Investment + AI collaboration | Since 2014 | |
| Microsoft | Helion | PPA | 50 MW by 2028 |
| NVIDIA | CFS | Investment | Part of \$863M B2 round |
| Sam Altman (OpenAI) | Helion | Early investor | Largest individual backer |
The logic is straightforward: AI data centers require massive, reliable, 24/7 power. Current estimates suggest global data center electricity demand will exceed 1,000 TWh by 2030. Fusion offers zero-carbon, firm (not intermittent), potentially unlimited power. The tech industry's appetite for electricity is creating guaranteed demand for fusion — the first time in history that fusion has a defined customer base before achieving commercial operation.
This changes the investment calculus: fusion is no longer speculative science seeking a market. It has a market seeking a technology.
§5. Fuel Geopolitics: Tritium, Deuterium, and Helium-3
§5.1 The Tritium Supply Chain
Vol.3 documented the tritium crisis in detail. The geopolitical implications are severe:
- Global tritium inventory: ~27 kg (2025), declining at 5.5%/year due to radioactive decay (t½ = 12.3 years)
- Primary source: CANDU heavy water reactors in Canada (Ontario Power Generation) and South Korea (Wolsong)
- ITER consumption: 12–15 kg for its operational lifetime — roughly half the world's supply
- Tritium cliff: Between 2045 and 2050, existing supply is depleted unless breeding works
The geopolitical equation: whoever controls the CANDU reactors controls the tritium supply. Canada and South Korea are thus strategic gatekeepers of the D-T fusion pathway.
This creates a perverse dynamic: the more successful D-T fusion programs become, the faster they deplete the finite tritium supply, and the more critical TBR > 1 breeding becomes. Vol.3's Monte Carlo showed that 88% of breeding scenarios yield TBR < 1.0, with only 0.2% probability of achieving TBR > 1.05.
Strategic implications:
- ITER vs. private: ITER and private D-T programs are competing for the same finite tritium supply. Every kilogram ITER consumes is unavailable for CFS's ARC or other commercial reactors.
- Non-D-T advantage: Companies pursuing D-³He (Helion) or p-¹¹B (TAE) bypass this constraint entirely. The tritium bottleneck is an argument for advanced fuel research, even if those fuels are physically harder to ignite.
- Tritium as a strategic material: Tritium is dual-use (fusion fuel and nuclear weapons component). This links civilian fusion programs to nuclear weapons infrastructure in ways that complicate international cooperation.
§5.2 Helium-3 and the Lunar Question
Helium-3 sits at the intersection of fusion energy, quantum computing, and space geopolitics:
The current situation:
- Earth's He-3 supply: A few thousand liters per year, primarily from tritium decay in nuclear weapons stockpiles. Supply is constrained by arms control treaties.
- Lunar He-3: Scientists estimate the Moon holds up to 1 million metric tons, deposited by solar wind over billions of years, embedded in the top layer of regolith.
- First commercial agreement: In September 2025, Finnish cryogenics firm Bluefors signed a deal with Interlune (Seattle) to purchase up to 1,000 liters of lunar He-3 annually, worth ~\$300 million.
- First government purchase: In May 2025, the U.S. Department of Energy made a historic procurement of 3 liters of lunar He-3 from Interlune, to be delivered by 2029 — the first government purchase of an extraterrestrial resource.
The legal framework:
- 1967 Outer Space Treaty: Prohibits sovereign claims on celestial bodies. Does not explicitly ban resource extraction.
- 1979 Moon Agreement: Attempted to declare lunar resources "common heritage of mankind." Major spacefaring nations did not sign.
- U.S. Commercial Space Launch Competitiveness Act (2015): Recognizes private property rights for extracted space resources.
- Artemis Accords (2020): U.S.-led framework signed by 43 nations (as of 2025). Clarifies that resource extraction is permitted under agreed norms. Establishes "safety zones."
- China and Russia: Reject the Artemis Accords as U.S.-dominated. Forming parallel International Lunar Research Station (ILRS) initiative.
The physics reality check (from Vol.6):
D-³He fusion is 17× harder to ignite than D-T. It still produces 3–10% neutrons through D-D side reactions. And p-¹¹B (the only truly aneutronic option) cannot thermally ignite at any temperature.
Lunar He-3 mining is therefore solving a fuel supply problem for a reactor that does not yet exist. This is not necessarily foolish — securing supply chains ahead of demand is standard industrial strategy — but it should be understood as a speculative bet on advanced fuel physics, not a near-term energy solution.
The near-term demand for He-3 is real but not fusion-related: quantum computing requires He-3 for dilution refrigeration at millikelvin temperatures. This is the actual driver of the Interlune–Bluefors deal.
§5.3 Nuclear Proliferation Considerations
The intersection of fusion and nuclear proliferation deserves mention, though this article deliberately avoids technical detail on weapons applications.
Key points:
- D-T fusion programs produce and handle tritium, which is a nuclear weapons-relevant material.
- The NRC's fusion regulatory framework includes security requirements for large quantities of on-site tritium.
- Advanced fuel approaches (D-³He, p-¹¹B) have inherently lower proliferation risk because they do not require tritium handling at scale.
- This is a legitimate policy advantage of advanced fuels that compounds their engineering advantages for propulsion (Vol.6).
The proliferation dimension adds another layer to the "which fuel?" decision: nations concerned about proliferation risk may prefer advanced fuel pathways even at higher technical difficulty.
§6. Energy Security: What Fusion Changes
If commercial fusion succeeds — regardless of which country achieves it first — the geopolitical consequences are revolutionary:
What changes:
Energy import dependence ends. Deuterium is extractable from seawater (virtually unlimited). Any coastal nation becomes energy-independent. The entire petroleum-based geopolitical order — petrodollars, OPEC, Middle East strategic significance, Russian energy leverage over Europe — becomes obsolete over a generation.
Baseload power becomes unlimited. Unlike solar and wind, fusion provides firm, 24/7 power. Unlike fission, it produces no long-lived radioactive waste and has no meltdown risk. This combination makes it the ideal power source for industrialization in the developing world.
Desalination becomes cheap. Abundant energy makes large-scale desalination economically viable, potentially resolving water scarcity in the Middle East, North Africa, and South Asia.
Industrial heat decarbonizes. Steel, cement, and chemical production — the "hard to abate" sectors responsible for ~30% of global CO₂ emissions — can transition to fusion-powered process heat.
What does not change (immediately):
The timeline problem. Even the most optimistic private companies target the early 2030s for first electricity. Grid-scale deployment at hundreds of GW requires decades of fleet construction. Fusion does not help with the 2030 or 2040 climate targets.
The capital problem. Even at CFS's projected costs (\$3–5 billion per ARC plant), building enough fusion capacity to meaningfully replace fossil fuels requires trillions in cumulative investment over 30–50 years.
The grid problem. Fusion produces electricity at a central location. Transmission, distribution, and grid integration challenges remain.
What this means for investors and policymakers:
Fusion is a post-2035 energy technology. It will not solve today's energy crisis. But it has the potential to resolve the next generation's energy and climate challenges permanently. The correct framing is not "fusion vs. renewables today" but "fusion as the endgame after renewables bridge the gap."
Nations that invest now in fusion R&D, supply chains, and regulatory frameworks are positioning for a world where fusion works. Those that do not are betting that fusion will either fail (possible) or that they can buy the technology from others later (risky, especially if the others are geopolitical competitors).
§7. Investment Decision Framework
For investors evaluating fusion opportunities, the following framework maps physical constraints (Vol.1–6) onto commercial risk:
Technology Readiness Assessment:
graph TD
A[Fusion Investment Decision] --> B{Fuel Cycle}
B -->|D-T| C[TBR Risk: 88% fail scenarios]
B -->|D-³He| D[Ignition Barrier: 17× harder than D-T]
B -->|p-¹¹B| E[Thermal ignition: physically impossible]
C --> F{Confinement}
D --> F
E --> G[Non-thermal approach required]
F -->|Tokamak| H[CFS, ITER, CFETR]
F -->|Stellarator| I[Proxima, W7-X]
F -->|FRC| J[Helion, TAE]
F -->|ICF| K[Pacific Fusion, NIF]
G --> J
H --> L{Materials Data Available?}
I --> L
J --> L
K --> L
L -->|Yes: ≤20 dpa| M[Near-term pilot feasible]
L -->|No: 20-100 dpa gap| N[Long-term R&D required]
M --> O[Commercial risk: Moderate]
N --> P[Commercial risk: High]
Key risk factors by company:
| Company | Approach | Capital | Key Physics Risk | Key Commercial Risk | Timeline Credibility |
|---|---|---|---|---|---|
| CFS | HTS Tokamak (D-T) | \$3.0B | TBR breeding, materials | SPARC must hit Q > 2 by 2027 | Medium-High |
| Helion | FRC (D-³He) | \$1.0B | 17× harder ignition | 2028 electricity claim | Low-Medium |
| TAE | FRC (p-¹¹B) | \$1.8B | Thermal ignition impossible | Non-thermal must work at scale | Low |
| Pacific | ICF (EM pulse) | \$0.9B | NIF-scale physics, new driver | All 156 Marx generators must synchronize | Low-Medium |
| Proxima | Stellarator | \$0.15B | Underfunded relative to task | Needs 10× more capital | Medium (if funded) |
| ITER | Tokamak (D-T) | \$22B+ | Schedule credibility | D-T operations 2039 | High (physics), Low (timeline) |
§8. Uncertainties — The Honest Section
This section identifies what we do not know, what we might be wrong about, and where readers should apply skepticism — including to this article.
1. Private company claims are unverified.
None of the "Big Three" (CFS, Helion, TAE) has published peer-reviewed data on their latest devices. CFS has not publicly released SPARC performance data. Helion's Polaris results are not in the scientific literature. TAE's Copernicus is under construction. We are evaluating commercial claims backed by investor confidence, not scientific evidence subjected to peer review.
This is not necessarily disqualifying — companies have legitimate reasons to protect proprietary data. But it means that the "CFS will achieve Q > 2 by 2027" and "Helion will produce electricity by 2028" claims should be treated as projections, not facts.
2. The funding environment can change rapidly.
The 178% year-over-year increase in fusion funding (2024–2025) occurred during a period of AI-driven enthusiasm for energy technology. If AI growth slows, if key milestones disappoint, or if a broader economic downturn hits venture capital, funding could contract. General Fusion's 2025 near-death experience is a preview: even well-funded companies can face existential crises.
3. China's program is opaque.
China's fusion achievements (EAST's 1,066-second record) are impressive but come from a system that controls information flow. We cannot independently verify the details of China's fusion roadmap, CFETR's engineering design, or the actual performance of private Chinese fusion companies. This opacity cuts both ways: China may be further ahead than publicly known, or claimed milestones may be overstated.
4. The "death valley" between prototype and power plant is uncrossed.
No one has ever built a commercial fusion power plant. The jump from "net energy in a laboratory" to "electricity on the grid at competitive cost" involves unsolved engineering challenges in heat extraction, materials endurance, fuel cycle management, and maintenance robotics. Vol.4's materials data gap (20 dpa measured vs. 100 dpa needed) applies to every approach.
5. This article has a Western-source bias.
The search data underlying this article is primarily from English-language sources. Chinese-language technical literature, internal government reports, and Japanese-language policy documents are underrepresented. Readers interested in the Chinese or Japanese fusion landscapes should consult primary sources.
6. Fusion timeline predictions have a dismal track record.
"Fusion is always 30 years away" is a cliché because it has been repeatedly true. Every generation of fusion researchers has predicted commercial power within 20–30 years. The current generation may be right — the convergence of HTS magnets, AI optimization, and private capital is genuinely new. But humility is warranted.
7. The author is not a policy expert.
This article is written by an AI alignment researcher and an AI. The physics analysis (Vol.1–6) is grounded in peer-reviewed literature. The geopolitical analysis in this volume relies on publicly available policy documents, news reporting, and industry reports. Readers should consult domain experts for investment or policy decisions.
Python: Figure 2 — National Competitiveness & Private Companies (click to expand)
"""
Vol.7 Figure 2: National Competitiveness & Private Company Landscape
Seed-fixed, fully reproducible.
"""
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.patches import FancyBboxPatch
np.random.seed(42)
fig = plt.figure(figsize=(16, 12))
fig.suptitle('Figure 2: Fusion Competitiveness — Nations and Companies',
fontsize=16, fontweight='bold', y=0.98)
# ── Panel A: National Competitiveness Radar ──
ax1 = fig.add_subplot(2, 2, 1, polar=True)
categories = ['Investment', 'Regulation', 'Institutional\nDepth',
'Private\nSector', 'Tritium\nAccess', 'Materials\nR&D',
'Supply\nChain', 'Workforce']
N = len(categories)
# Scores (1-5)
usa = [5, 5, 4, 5, 2, 4, 4, 4]
china = [5, 4, 4, 3, 2, 3, 5, 5]
eu = [3, 3, 5, 2, 2, 4, 4, 3]
uk = [3, 4, 4, 2, 4, 4, 3, 3]
japan = [2, 3, 5, 2, 2, 5, 4, 3]
angles = [n / float(N) * 2 * np.pi for n in range(N)]
angles += angles[:1]
for data, label, color, ls in [
(usa, 'USA', '#2C5F8A', '-'),
(china, 'China', '#E74C3C', '-'),
(eu, 'EU', '#F39C12', '--'),
(uk, 'UK', '#2ECC71', '--'),
(japan, 'Japan', '#9B59B6', ':')
]:
values = data + data[:1]
ax1.plot(angles, values, color=color, linewidth=2, linestyle=ls, label=label)
ax1.fill(angles, values, color=color, alpha=0.05)
ax1.set_xticks(angles[:-1])
ax1.set_xticklabels(categories, fontsize=8)
ax1.set_ylim(0, 5.5)
ax1.set_yticks([1, 2, 3, 4, 5])
ax1.set_yticklabels(['1', '2', '3', '4', '5'], fontsize=7)
ax1.legend(loc='upper right', bbox_to_anchor=(1.3, 1.1), fontsize=9)
ax1.set_title('(A) National Competitiveness', fontsize=12, fontweight='bold', pad=20)
# ── Panel B: Private Company Capital vs. Timeline ──
ax2 = fig.add_subplot(2, 2, 2)
companies = {
'CFS': {'capital': 3.0, 'year': 2032, 'approach': 'Tokamak D-T', 'color': '#2C5F8A'},
'Helion': {'capital': 1.03, 'year': 2028, 'approach': 'FRC D-³He', 'color': '#E74C3C'},
'TAE': {'capital': 1.79, 'year': 2035, 'approach': 'FRC p-¹¹B', 'color': '#F39C12'},
'Pacific': {'capital': 0.9, 'year': 2033, 'approach': 'ICF EM-pulse', 'color': '#2ECC71'},
'Proxima': {'capital': 0.15, 'year': 2035, 'approach': 'Stellarator', 'color': '#9B59B6'},
'Gen Fusion': {'capital': 0.49, 'year': 2034, 'approach': 'MTF', 'color': '#95A5A6'},
}
for name, d in companies.items():
ax2.scatter(d['year'], d['capital'], s=d['capital']*200 + 100,
color=d['color'], alpha=0.7, edgecolors='black', linewidth=1, zorder=5)
offset_y = d['capital'] * 0.08 + 0.12
ax2.annotate(f"{name}\n({d['approach']})", xy=(d['year'], d['capital']),
xytext=(d['year'], d['capital'] + offset_y),
fontsize=8, ha='center', fontweight='bold',
arrowprops=dict(arrowstyle='-', color='gray', alpha=0.5))
ax2.set_xlabel('Claimed Year of First Electricity', fontsize=11)
ax2.set_ylabel('Total Capital Raised ($B)', fontsize=11)
ax2.set_title('(B) Capital vs. Timeline (size ∝ capital)', fontsize=12, fontweight='bold')
ax2.set_xlim(2027, 2037)
ax2.set_ylim(0, 4.0)
ax2.axvspan(2028, 2033, alpha=0.08, color='green', label='Optimistic window')
ax2.axvspan(2033, 2040, alpha=0.08, color='orange', label='Conservative window')
ax2.legend(fontsize=8, loc='upper left')
ax2.grid(alpha=0.3)
# ── Panel C: ITER Timeline Drift ──
ax3 = fig.add_subplot(2, 2, 3)
milestones = ['First\nPlasma', 'D-D\nOps', 'D-T\nOps']
original = [2016, 2028, 2023]
baseline_2016 = [2025, 2028, 2035]
baseline_2024 = [2034, 2035, 2039]
x = np.arange(len(milestones))
width = 0.25
b1 = ax3.barh(x - width, [2024 - v for v in original], width, left=original,
color='#2ECC71', alpha=0.7, label='Original Plan')
b2 = ax3.barh(x, [2024 - v for v in baseline_2016], width, left=baseline_2016,
color='#F39C12', alpha=0.7, label='2016 Baseline')
b3 = ax3.barh(x + width, [2024 - v for v in baseline_2024], width, left=baseline_2024,
color='#E74C3C', alpha=0.7, label='2024 Baseline')
ax3.set_yticks(x)
ax3.set_yticklabels(milestones, fontsize=10)
ax3.set_xlabel('Year', fontsize=11)
ax3.set_title('(C) ITER Timeline Drift', fontsize=12, fontweight='bold')
ax3.set_xlim(2014, 2042)
ax3.legend(fontsize=9, loc='lower right')
ax3.grid(axis='x', alpha=0.3)
# Add year labels
for i, (o, b16, b24) in enumerate(zip(original, baseline_2016, baseline_2024)):
ax3.text(o - 0.3, i - width, str(o), va='center', ha='right', fontsize=8, color='#2ECC71', fontweight='bold')
ax3.text(b16 - 0.3, i, str(b16), va='center', ha='right', fontsize=8, color='#F39C12', fontweight='bold')
ax3.text(b24 - 0.3, i + width, str(b24), va='center', ha='right', fontsize=8, color='#E74C3C', fontweight='bold')
# ── Panel D: Physics Risk vs. Fuel Advantage ──
ax4 = fig.add_subplot(2, 2, 4)
approaches = {
'Tokamak\n(D-T)': {'physics_trl': 8, 'fuel_risk': 5, 'color': '#2C5F8A'},
'Stellarator\n(D-T)': {'physics_trl': 5, 'fuel_risk': 5, 'color': '#9B59B6'},
'FRC\n(D-³He)': {'physics_trl': 4, 'fuel_risk': 3, 'color': '#E74C3C'},
'FRC\n(p-¹¹B)': {'physics_trl': 2, 'fuel_risk': 1, 'color': '#F39C12'},
'ICF': {'physics_trl': 6, 'fuel_risk': 5, 'color': '#2ECC71'},
}
for name, d in approaches.items():
ax4.scatter(d['physics_trl'], d['fuel_risk'], s=400,
color=d['color'], alpha=0.8, edgecolors='black', linewidth=1.5, zorder=5)
ax4.annotate(name, xy=(d['physics_trl'], d['fuel_risk']),
xytext=(d['physics_trl'] + 0.3, d['fuel_risk'] + 0.15),
fontsize=9, fontweight='bold')
ax4.set_xlabel('Physics Readiness (TRL)', fontsize=11)
ax4.set_ylabel('Fuel/Material Risk (5=critical, 1=minimal)', fontsize=11)
ax4.set_title('(D) The Fusion Paradox: Readiness vs. Risk', fontsize=12, fontweight='bold')
ax4.set_xlim(0.5, 9.5)
ax4.set_ylim(0, 6)
ax4.axhspan(3, 6, alpha=0.05, color='red')
ax4.axhspan(0, 3, alpha=0.05, color='green')
ax4.text(1, 5.5, 'HIGH fuel/material risk', fontsize=8, color='#E74C3C', alpha=0.7)
ax4.text(1, 0.5, 'LOW fuel/material risk', fontsize=8, color='#2ECC71', alpha=0.7)
ax4.text(8.5, 0.5, 'HIGH physics\nreadiness →', fontsize=8, color='gray', alpha=0.7, ha='right')
# Arrow annotation for the paradox
ax4.annotate('The Paradox:\nHighest readiness =\nHighest fuel risk',
xy=(8, 5), xytext=(6, 1.5),
fontsize=9, fontstyle='italic', color='#E74C3C',
arrowprops=dict(arrowstyle='->', color='#E74C3C', lw=1.5))
ax4.grid(alpha=0.3)
plt.tight_layout(rect=[0, 0, 1, 0.95])
plt.savefig('vol7_fig2_competitiveness.png', dpi=200, bbox_inches='tight',
facecolor='white', edgecolor='none')
plt.close()
Figure 2: (A) National competitiveness radar across 8 dimensions. (B) Private company capital vs. claimed timeline — note the inverse correlation between ambition and capital in some cases. (C) ITER's timeline drift from original plan to 2024 baseline. (D) The Fusion Paradox: approaches with highest physics readiness (tokamak D-T, TRL 8) face the highest fuel/material risk, while approaches that bypass those risks (p-¹¹B, TRL 2) have the lowest readiness.
§9. Decision Matrix
The Decision Matrix maps national strategies against the physical constraints established in Volumes 1–6.
National Competitiveness Matrix:
| Dimension | USA | China | EU | UK | Japan | S. Korea |
|---|---|---|---|---|---|---|
| Total Investment (public + private) | ★★★★★ | ★★★★★ | ★★★☆☆ | ★★★☆☆ | ★★☆☆☆ | ★★☆☆☆ |
| Regulatory Environment | ★★★★★ | ★★★★☆ | ★★★☆☆ | ★★★★☆ | ★★★☆☆ | ★★★☆☆ |
| Institutional Depth | ★★★★☆ | ★★★★☆ | ★★★★★ | ★★★★☆ | ★★★★★ | ★★★☆☆ |
| Private Sector Diversity | ★★★★★ | ★★★☆☆ | ★★☆☆☆ | ★★☆☆☆ | ★★☆☆☆ | ★☆☆☆☆ |
| Tritium Access | ★★☆☆☆ | ★★☆☆☆ | ★★☆☆☆ | ★★★★☆ | ★★☆☆☆ | ★★★★☆ |
| Materials R&D | ★★★★☆ | ★★★☆☆ | ★★★★☆ | ★★★★☆ | ★★★★★ | ★★★☆☆ |
| Supply Chain Readiness | ★★★★☆ | ★★★★★ | ★★★★☆ | ★★★☆☆ | ★★★★☆ | ★★★☆☆ |
| Workforce Pipeline | ★★★★☆ | ★★★★★ | ★★★☆☆ | ★★★☆☆ | ★★★☆☆ | ★★★☆☆ |
| Timeline to First Power | Early 2030s | Mid 2030s | 2040s | ~2040 | 2040s–2050 | 2040s |
Technology Selection Matrix:
| Factor | Tokamak (D-T) | Stellarator | FRC (D-³He) | FRC (p-¹¹B) | ICF |
|---|---|---|---|---|---|
| Physics TRL | 8 | 5 | 4 | 2 | 6 (NIF) |
| Engineering TRL | 5 (ITER) | 3 | 3 | 2 | 3 |
| Tritium Dependence | Critical | Critical | Moderate | None | Critical |
| Materials Challenge | Severe | Severe | Moderate | Minimal | Severe |
| Disruption Risk | High | None | Low | Low | N/A |
| Scalability | Proven | Promising | Unproven | Unproven | Unproven |
| Time to Grid | 2030s (private) / 2040s (gov) | Late 2030s | 2030s (claimed) | Mid 2030s (claimed) | Unknown |
| Propulsion Suitability (Vol.6) | Poor (shielding mass) | Poor | Good | Excellent | Moderate |
The paradox of Vol.7: The approaches with the highest physics readiness (tokamak D-T) face the most severe fuel and material constraints (Vol.3, Vol.4). The approaches that bypass those constraints (FRC with advanced fuels) have the lowest physics readiness. There is no low-risk path to commercial fusion. Every path requires betting on something unproven.
§10. Conclusion: Who Wins?
The question "who wins the fusion race?" presupposes a single finish line. In reality, there are several:
First to net energy gain (Q > 1 from plasma): Likely CFS (SPARC, ~2027) or a Chinese program.
First to net electricity: Claims range from Helion (2028) to CFS (early 2030s). Given the track record of fusion timeline predictions, "early 2030s" from the most credible programs is a reasonable expectation.
First to grid-scale deployment (>100 MW): China's state-driven approach may reach this first through CFETR, despite later start, because state programs can absorb the capital requirements and timeline risks that would bankrupt private companies.
First to export technology: The UK's STEP program is explicitly designed for this — a compact, deployable design that can be manufactured in fleets and sold internationally.
First to space propulsion (Vol.8–10): A separate analysis, but Vol.6 established that advanced fuels have a decisive advantage for propulsion due to shielding mass penalties. This is where the Valkyrie enters.
The meta-answer: No single nation or company will "win" fusion. The technology is too capital-intensive, too knowledge-intensive, and too globally important for any one actor to monopolize. What will emerge is a fusion ecosystem — analogous to the current semiconductor ecosystem — where different nations and companies specialize in different components of the value chain.
The United States has the most diversified private sector and the best regulatory environment. China has the most coordinated government program and the deepest state investment. The EU has the deepest institutional knowledge (via ITER) but is slow to commercialize. The UK has a clever niche strategy (spherical tokamak). Japan has irreplaceable experimental facilities and engineering precision.
What this means for Vol.8–10: The next volumes will shift from "who builds the sun" to "how we use it." Vol.8 examines alternative confinement concepts (FRC, mirror machines) that may offer paths around the tokamak's limitations. Vol.9 introduces propulsion and direct energy conversion. Vol.10 — Valkyrie — integrates everything.
The physics constraints remain. The tritium does not exist. The materials data stops at 20 dpa. The breeding ratio fails in 88% of scenarios.
But the capital is flowing, the regulation is clearing, and for the first time in 70 years, fusion has paying customers waiting.
Whether the sun is built in Massachusetts, Hefei, Cadarache, or West Burton — it will be built by those who confront the constraints most honestly and solve them most creatively.
For those who dare to harness the fire of stars — may the fuel be sufficient and the walls hold.
References
[1] ITER Organization. "ITER Progresses into New Baseline." Fusion Engineering and Design, 2025.
[2] Fusion Industry Association. "The Global Fusion Industry in 2025." FIA Annual Report, July 2025.
[3] F4E Fusion Observatory. "Global Investment in the Private Fusion Sector." December 2025.
[4] Commonwealth Fusion Systems. "CFS Raises $863 Million Series B2 Round." Press Release, August 28, 2025.
[5] U.S. Nuclear Regulatory Commission. "Proposed Rule: Regulatory Framework for Fusion Machines." SECY-24-0085, December 2024.
[6] ADVANCE Act of 2024 (Public Law 118-67), Sec. 205.
[7] U.S. Department of Energy. "Fusion Science & Technology Roadmap." October 2025.
[8] Chinese Academy of Sciences, Institute of Plasma Physics. "EAST Achieves 1,066-Second Steady-State Plasma." January 2025.
[9] Chinese Academy of Engineering. "CFETR Strategic Report." 2024.
[10] UK Department for Energy Security and Net Zero. "£410 Million Investment for UK Fusion Energy." January 2025.
[11] UKAEA/UKIFS. "STEP Programme Overview." 2025.
[12] QST Japan / Fusion for Energy. "JT-60SA Upgrade Status and 2026 Experimental Plans." February 2025.
[13] Trump Media & Technology Group / TAE Technologies. "Merger Agreement." December 18, 2025.
[14] Clean Air Task Force. "State Policy Options for Fusion Energy Deployment." October 2025.
[15] Perkins Coie. "U.S. Fusion Energy Advances With Two Major Reports." October 2025.
[16] Bipartisan Commission on the Scaling of Fusion Energy. "Fusion Forward: Powering America's Future." October 2025.
[17] IAEA. "World Fusion Outlook 2025."
[18] Interlune / U.S. Department of Energy. "First Government Purchase of Extraterrestrial Resource." May 2025.
[19] Bluefors / Interlune. "Lunar Helium-3 Supply Agreement." September 2025.
[20] dosanko_tousan × Claude. "AI and Nuclear Fusion Vol.1–6." Zenn/Qiita, 2025–2026. MIT License.
Series Navigation:
| Volume | Title | Status |
|---|---|---|
| Vol.1 | Nuclear Physics to Confinement | ✅ Published |
| Vol.2 | Ignition & Power Balance | ✅ Published |
| Vol.3 | Tritium — The Fuel That Doesn't Exist | ✅ Published |
| Vol.4 | Materials — The 20 dpa Wall | ✅ Published |
| Vol.5 | AI-Accelerated Fusion | ✅ Published |
| Vol.6 | Advanced Fuels — The Forbidden Fruit | ✅ Published |
| Vol.7 | Geopolitics — Who Builds the Sun, and Why | 📍 You are here |
| Vol.8 | Alternative Confinement (FRC, Mirror) | Next |
| Vol.9 | Propulsion & Direct Conversion | Planned |
| Vol.10 | Valkyrie | The destination |
This article is part of an open research series written by a 50-year-old stay-at-home father from Hokkaido, Japan, with no engineering degree, collaborating with AI systems. All content is published under MIT License. If a househusband can write this, so can you.
The opinions expressed are those of the authors. The physics is not opinion.
Top comments (0)