Yang Zhao on Controlled Fusion: From Honghuang 70 to Commercial Engineering

Yang Zhao believes controlled fusion could substantially affect energy and civilization over coming decades, while warning that every stage of the complex engineering effort brings problems.

Host Zhang Xiaojun introduces Energy Singularity founder Yang Zhao and asks how far fusion is from commercial use and what abundant energy might change.

The discussion examines technical routes, costs and schedules largely through the guest’s assessment; future results still require experimental and engineering verification.

Yang begins with nuclear fission and fusion: both involve nuclear reactions, but fission splits heavier nuclei while fusion combines lighter ones.

He uses mass–energy equivalence to explain energy release and distinguishes controlled fusion from an uncontrolled, one-time release.

The goal of control is to obtain energy steadily in accordance with a design, rather than merely produce one reaction.

A tokamak is a magnetic-confinement device: magnetic fields confine hot plasma. Other approaches on Earth include inertial confinement.

Yang says the Sun uses strong gravity to create fusion conditions, while terrestrial devices need other means to reach suitable conditions.

He explains why magnets matter: field strength, device size, operating time and manufacturing difficulty all affect the design.

Earlier devices could use copper magnets relatively straightforwardly, but resistance and heating limited operating time and increased energy use.

Superconducting magnets require more complex engineering, including cryogenic conditions. Yang says the goal of longer operation motivates research into them.

High-temperature superconductors still require cooling; “high” is relative to another class of superconductors, not room-temperature operation.

They discuss ITER as an example of a large international collaboration and a more established design route.

Yang believes stronger fields may permit a more compact device for certain targets, but says specific devices still need validation and theoretical potential is not achieved cost.

On the gain factor Q, he explains its role in comparing fusion output with input; a power plant must also account for losses elsewhere in the system.

He uses the product of density, temperature and confinement time to explain the relationship among fusion conditions. Experiments and Q definitions must be interpreted within their stated boundaries.

For tokamaks, Yang describes empirical scaling: researchers use results from different devices to estimate relationships among field, size and performance.

He says plasma systems are complex and many relationships require empirical fitting; simple formulas alone cannot establish every engineering outcome.

The interview reviews fusion research history and returns to the civilian challenge of turning a brief reaction into an energy system that operates on demand.

Yang says the engineering goal is not merely to produce fusion but to operate at designed power while addressing cost, reliability and continuous operation.

He compares inertial and magnetic confinement and their experimental and commercialization challenges. The comparison is the guest’s technical assessment.

For magnetic confinement, tokamaks have accumulated substantial research and experimental experience, which he considers important when choosing a route.

They discuss why large devices in multiple countries use tokamaks and how international projects accumulate comparable experimental data.

Yang introduces the team’s Honghuang 70 experiment. Its purpose must be distinguished from the goals of a future electricity-generating device.

He explicitly says Honghuang 70 did not demonstrate a cost or size advantage at equivalent performance; the team hopes a later device will test that.

He reviews the path from laboratory discovery to larger-scale supply of high-temperature superconductors and says availability influenced the startup’s decision.

Yang describes the team’s 2021 view: fusion had accumulated decades of scientific and engineering work, while commercialization required addressing economics and implementation.

He considers the schedules and costs of large experiments major constraints, while actual progress must determine whether and when targets are achieved.

In his view, commercialization requires bringing generation costs to a competitive level. This is a company goal, not an achieved electricity price.

He says the company chose to control the device design and some key subsystems to iterate and manage costs.

Asked why he founded a company rather than remaining in academia, Yang says a business can drive engineering iteration and commercial feasibility while scientific research retains its role.

He recalls his interest in fundamental physics and complex systems, discussing why reducing a system to its smallest parts does not always directly explain its macroscopic behavior.

On AI and physics, he says the fields have borrowed ideas and tools, while AI development is mainly driven by computer science and related fields.

The team must select suitable materials and address structure and heat dissipation; Yang says this does not mean inventing every material from scratch.

He recalls becoming interested in fusion as an undergraduate and imagines abundant future energy affecting space exploration. Such long-term applications remain speculative.

He discusses his doctoral background in theoretical physics and later efforts to combine technical experience with entrepreneurship.

An earlier startup experience led him to reconsider whether fusion could be pursued at that time and what conditions would otherwise be missing.

Yang believes founders need deep commitment to the problem because major projects bring continuing difficulties and competing alternatives.

He describes assessing knowledge, talent and engineering conditions, and believes experience from one device can inform the next design.

He says the domestic supply chain was at an early stage and the construction of research devices did not fully match the capabilities a commercial company hoped to build.

“70” refers to a device dimension. The team grew from a few founders while building the first device over roughly two years.

He divides the work into physics, conceptual and engineering design, and says the device contains many interdependent subsystems.

A feasible concept is not yet a manufacturable device: engineering must specify practical requirements such as cooling, vacuum, power and interfaces.

Yang recalls overlapping design freezes, manufacturing and assembly, and corrects a year he initially misstated during the interview.

The team included theorists, but many members were engineers in structures, cryogenics, vacuum and magnet manufacturing.

He emphasizes that engineering problems emerge as designs become physical, and that early correction and late repair have different costs.

After assembly, the team must test systems individually and then verify that they operate together.

Yang describes first plasma as an important startup milestone: systems work together to create plasma for subsequent experiments.

He distinguishes starting the device from achieving commercial power-plant parameters: the former is an engineering milestone, while the latter requires substantial further work.

Yang says Honghuang 70 demonstrated that high-temperature superconducting material could be incorporated into a complete experimental device; it did not demonstrate commercial generation.

He discusses in-house development and local sourcing while acknowledging imported cryogenic equipment and sensors.

The roadmap includes later devices and magnet validation. Schedules and performance targets are team plans, not completed achievements.

He explains that a fusion magnet needs a larger opening for plasma and equipment, unlike some high-field magnets built only for small samples.

Stronger fields and currents create mechanical forces. Designers must balance structural support against space available for conductors.

Cooling channels also take space, so magnet design must account for mechanical, thermal and electrical constraints together.

Yang describes many small manufacturing issues that must be solved individually and says only a built device can test whether a process is reliable.

He says power, diagnostics and cryogenic systems also require validation, and the team aims to use physics designs supported by substantial prior evidence.

Even with established underlying physics, specific engineering parameters, reliability and integration can introduce new uncertainty.

Yang discusses later-device targets in relation to earlier experiments and emphasizes increasing capability step by step rather than skipping validation.

He sees potential benefits of higher-temperature superconductors in stronger fields and more compact devices, while identifying material processing and demanding manufacturing as risks.

He offers the team’s estimates for a future device’s cost and size. The figures are forecasts contingent on engineering and performance validation still to come.

Discussing a demonstration plant, he compares construction costs, fuel and waste; actual safety and waste impacts depend on the specific fusion design.

He warns that extrapolating from low-parameter experiments to high-parameter devices may encounter unobserved conditions, so scientific and engineering risks cannot be ignored.

Yang discusses how capital, talent and early experience can shape competition within one technical route. This is his assessment of the industry.

He predicts that substantial energy-cost reductions could lead to new demand; this is an economic forecast, not a guarantee.

On AI, Yang says machine learning may help plasma control and diagnostics, citing existing research as an example.

He says such applications remain early and that his team is exploring algorithms for part of the control work.

They discuss capital, supply chains and markets in Chinese and US fusion. Yang predicts local teams in both, while actual industry development remains uncertain.

He compares Energy Singularity’s superconducting tokamak approach with US company CFS and notes that others are exploring different routes.

Asked how difficult competitive fusion electricity would be, he acknowledges the challenge and gives the team’s target timeframe for a demonstration device.

He distinguishes an experiment pursued regardless of cost from stable, economical commercial operation; the two are not equivalent.

He analogizes to fission development and speculates that demonstration to scale could still take years; this is not a validated fusion deployment schedule.

Yang believes a device reaching key targets could change market confidence, but the outcome depends on experiments.

Asked how much scaling will cost, he admits insufficient visibility now and says a demonstration device and its remaining problems must be observed first.

On the organization, he wants the team to identify critical problems, remove unnecessary work and learn causes from each engineering issue.

He describes team size and hiring, saying fusion engineering requires a different talent mix from a typical internet company.

Yang says building a device involves extensive engineering; the team reads research papers but publishes selectively because of time and priorities.

He sees other teams’ public designs and research as useful references, while daily work centers on specific problems.

The company’s near-term commercial aim is to develop equipment capabilities for a future demonstration plant; current experiments are preparatory steps.

He discusses seeking capital patient with long research cycles and says the pool of potential investors is not unlimited.

Looking back on Honghuang 70, he says the team focused on solving problems one by one and only later saw the full difficulty.

He stresses that moving from the first experiment to higher-performance devices requires major growth in team capabilities and engineering demands.

Reflecting on management, he acknowledges that he may have invested too little in team morale and motivation.

They discuss possible failure and personal pressure; he identifies capital, talent and delivery as major risks.

Yang says insufficient funding could prevent the next stage from starting, while missed targets could affect later devices and business plans.

He believes fusion has a clear but difficult engineering route and says he cannot define a precise AGI standard. This is his belief, not a guarantee of success.

Asked what he would do if the project failed, he says he would regret it and hope to learn from the problems and keep trying.

A closing rapid-fire section returns to other technological opportunities, daily life and long-term goals.

Yang says the fusion project occupies much of his attention, and the interview ends by discussing continued exploration.