The quest for nuclear fusion – often called the “holy grail” of energy has consumed billions of dollars and decades of research. Yet, in early 2025, a 20 year old Canadian math student, Hudhayfa Nazoordeen, achieved a milestone most thought impossible.
He built a neutron-producing nuclear fusor in his kitchen for just $3000, guided entirely by AI chatbots like Claude 3.5 Sonnet and OpenAI’s o1 Pro.
The $22 billion ITER project has 35 nations collaborating to achieve fusion by 2035. Meanwhile, in a cramped Ontario apartment, he detected neutron emissions from a homemade reactor using salvaged transformers, a Hydrocar toy’s fuel cell, and AI chatbots as his co-pilots.
This is the story of how open-source knowledge, AI assistance, and sheer determination rewrote the rules of energy innovation.
Why Fusion Remains Science’s Everest?
Nuclear fusion requires replicating stellar conditions – temperatures over millions of °C and pressures exceeding Earth’s core. While labs use lasers or magnetic fields, Hudhayfa’s approach – inertial electrostatic confinement – trades brute force for clever engineering.
Nuclear fusion powers stars like our Sun by merging lightweight atomic nuclei (like deuterium and tritium) under extreme heat and pressure, releasing massive energy. Unlike fission (used in nuclear plants), fusion produces minimal radioactive waste and no greenhouse gases. However, replicating star-like conditions on Earth requires:
- Temperatures exceeding 39 million °C
- Precision vacuum systems (pressure as low as 3 millitorr)
- Complex electromagnetic confinement
Governments and labs like ITER have spent decades tackling these challenges. But Hudhayfa’s project focused on a simpler goal: detecting neutrons – a key fusion byproduct – using a tabletop device.
This allowed Hudhayfa to sidestep plasma instability issues plaguing tokamaks. His goal wasn’t infinite energy but proving fusion’s accessibility – a philosophy he calls “citizen science’s moon shot.”
How Claude and OpenAI Helped?
With zero prior hardware experience, Hudhayfa turned to AI to decode fusion engineering:
Documentation Decoding
He fed decades of forum threads, research papers, and technical manuals into Claude Projects, Anthropic’s AI tool. Claude synthesized step-by-step assembly instructions, identified critical components, and flagged safety risks.
The AI distilled them into bite-sized build steps while flagging lethal risks:
Claude caught a reversed polarity error that could’ve turned my vacuum chamber into a fragmentation grenade.
Electrical Wiring and Debugging
OpenAI’s o1 Pro assisted with circuit design for the 30kV electrostatic precipitator and rectifier setup.
When his neon sign transformer arced uncontrollably, OpenAI’s o1 Pro suggested adding a 50kΩ ballast resistor and pulsed DC modulation – stabilizing the 30kV output. The AI even designed a Faraday cage using aluminum mesh from a BBQ grill.
When the vacuum system failed, AI suggested troubleshooting steps like checking O-ring seals and pump calibration.
Cost Optimization
AI identified budget-friendly alternatives:
- A $32 Hydrocar toy’s PEM cell to electrolyze deuterium gas from heavy water ($80 for 50g)
- eBay-sourced vacuum pumps and transformers
I primarily relied on a giant claude project filled with documentation from forums, call transcripts from @meetgranola, email threads and more.
I use it extensively to debug, help me with safety, and follow otherwise unknown instructions that were outlined in the build process.… pic.twitter.com/qzhYQv3dUb
— HudZah ⁂ (@hud_zah) January 17, 2025Building the Fusor
The four-week sprint
- Week 1: AI-optimized shopping list prioritized critical path items
- Week 2: Claude guided vacuum system leak testing (-3 mTorr achieved)
- Week 3: o1 Pro debugged erratic plasma using real-time voltage analysis
- Week 4: Neutron bubbles confirmed D-D fusion at 10^5 reactions/sec
| Part | Function | Cost |
|---|---|---|
| Electrostatic Grid | Accelerates deuterium ions | $450 |
| Neon Sign Transformer | Generates 30,000V to ionize gas | $220 |
| High-Vacuum Chamber | Maintains 3 mTorr pressure | $900 |
| Bubble Counter | Detects neutron emissions | $150 |
The Breakthrough Moment
By electrolyzing deuterium gas and applying high voltage, the fusor produced a glowing plasma orb – visual proof of ionized particles colliding. The bubble counter later confirmed neutron emissions, marking successful fusion.
X-Rays and Explosions: Managing Risks in a Home Lab
Nuclear fusion experiments involve high voltages, radiation, and vacuum hazards.
Shielding strategies
- Borated polyethylene sheets (eBay-sourced) blocked 94% of neutrons
- Webcam repurposed as remote monitor behind lead-glass shield
- AI-enforced protocol: 15-minute cooldowns between 30-second runs
Near-disaster moments
A deuterium leak during Week 3 nearly caused deflagration. Claude’s emergency response protocol – immediate nitrogen purge and circuit lockout – prevented catastrophe.
Building fusion devices is 10% physics, 90% damage control. Building a fusor to do fusion is an order of magnitude more lethal. I relied on Claude to double-check every step.
Final Words: A New Era of Garage Inventors
Hudhayfa Nazoordeen’s $3,000 kitchen fusion reactor isn’t about solving energy crises. It’s a proof-of-concept that AI can democratize cutting-edge science. While his fusor doesn’t generate net energy, it challenges the notion that fusion is exclusive to billion-dollar labs.
By proving that fusion isn’t exclusive to billion-dollar consortia, he’s sparked a grassroots movement where hobbyists pressure governments to accelerate research.
As AI tools evolve, expect more basement innovators to rewrite the rules of physics – one plasma orb at a time.
I didn’t have a physics degree or a team. Just curiosity and Claude.
– Hudhayfa Nazoordeen
over a 36-hour livestream, I built a neutron-producing nuclear fusor in my kitchen using Claude.
successfully achieving nuclear fusion, entirely assisted by AI.
this was my first hardware project!—full story below pic.twitter.com/yuvjE5IHFb
— HudZah ⁂ (@hud_zah) January 17, 2025