<!-- ARTICLE_BODY_START -->
The next fuel delivery is already inside the design
On August 6, 2026, Fusion for Energy reported progress on preliminary design reviews for two European test blanket concepts intended for ITER. These are components for investigating how a future fusion machine might generate tritium, one of its fuels, using neutrons from the fusion reaction. The announcement described review meetings and remaining work, with an ambition to close the reviews by the end of the year. It did not report a commercially operating, fuel-self-sufficient power station. Fusion for Energy's design-review update
That distinction opens a more interesting discussion than the familiar argument about whether fusion is always decades away. The immediate engineering question is concrete: after a machine uses some fuel, how does usable replacement fuel reach the place where the next operation needs it? A successful answer must survive delays, material interactions, maintenance and measurement uncertainty. A claim that enough tritium is produced somewhere inside a calculation is only part of that answer.
Deuterium–tritium fusion brings a peculiar supply-chain problem inside the power plant. A proposed breeder blanket would use the reaction's neutrons to create replacement tritium from lithium. The resulting tritium must then become available to the fuel system. Meanwhile, fuel that passes through without reacting must be handled and recirculated. The plant is therefore both an energy converter and a continuously operating fuel-processing system. These roles influence one another rather than occurring in a convenient sequence.
This article considers that relationship at a conceptual level. It is not a guide to acquiring, producing, separating or handling radioactive materials. The numerical examples use abstract inventory units and deliberately simplified balances, not a proposed reactor configuration. They are intended to explain why timing and boundaries change the meaning of a fusion claim. Specific scientific and institutional evidence is dated separately; hypothetical operating stories do not describe any named project's actual performance.
All sources were checked on October 31, 2026. The evidence includes current development plans and older design research, each with a different status. A milestone on a roadmap is a goal. A calculated design result is a prediction under assumptions. A component experiment establishes behavior within its tested conditions. Keeping those categories intact lets progress be recognized without confusing it with a completed power system.
Three meanings of gain
Fusion reporting often uses a ratio to summarize a difficult experiment. Ratios are valuable because they compress a comparison into a number. They are dangerous when the comparison disappears. Energy released relative to energy delivered to a target, energy generated relative to external plasma heating, and net electricity exported by a complete facility are different quantities. A favorable value for one does not automatically establish a favorable value for the others.
Lawrence Livermore National Laboratory's July–August 2025 Science & Technology Review explicitly distinguishes the National Ignition Facility from a power plant. Its discussion of inertial fusion energy identifies additional requirements involving drivers, repeated target delivery, energy conversion and fuel breeding. These are the laboratory's descriptions of a development challenge, not evidence that all of those requirements have been integrated successfully into commercial service. LLNL's Beyond Ignition issue
A purely hypothetical energy example clarifies the accounting. Suppose an experiment releases ten energy units after two units reach its fuel. Its target-level ratio is five. Suppose the apparatus needed twenty electrical units to deliver those two units, and the released energy could be converted into four electrical units. The complete electrical comparison would look very different. These invented efficiencies and amounts do not represent NIF or any proposed reactor. They demonstrate why the boundary must accompany the number.
Fuel self-sufficiency introduces another ratio, with another denominator. It concerns whether replacement tritium becomes available in sufficient quantity to support use, losses and inventory needs. A plant could have an impressive energy ratio while depending on continuing external fuel deliveries. Conversely, an experimental fuel-processing system could demonstrate useful recovery without generating electricity. These achievements can contribute to the same ultimate project while answering different questions.
The practical consequence is that a serious project description needs several statements of success. One concerns the plasma or target. Another concerns net power. Another concerns fuel circulation and replacement. Another concerns the availability of the combined installation over time. Listing them separately does not diminish a breakthrough. It identifies what the breakthrough has accomplished and where the next uncertainty lies.
The inventory has a clock
Tritium is radioactive hydrogen. The US Environmental Protection Agency gives its half-life as 12.3 years and describes its occurrence in gaseous forms and in water. That establishes two basic considerations for a fuel discussion: an inventory changes through radioactive decay, and a record of its quantity must be accompanied by information about its physical and chemical context. EPA's tritium background
Using that rounded half-life, a sealed hypothetical inventory with no additions or removals would retain a fraction equal to two raised to minus one divided by 12.3 after one year. The result is approximately 0.945. About 5.5 percent of the initial tritium would have decayed over that year. This is an explanatory decay calculation, not a description of losses from an operating facility. Radioactive decay and an environmental release are different events and should never be combined casually under one alarming word.
The calculation also shows why a distant delivery date matters. A stock set aside for a future machine is not static merely because it remains securely stored. If construction or commissioning takes longer than expected, the quantity remaining changes. A planning document that compares future requirements with today's inventory must put both on the same date. Otherwise it can promise the same starting stock twice: once before decay and once after an unexplained delay.
Availability adds a second clock. Material can exist within a site while being unavailable for the next operating step. It may be undergoing processing, retained in equipment or awaiting an authorized transfer. The distinction resembles the difference between money recorded as revenue and cash available to pay a bill, but the physical implications are more demanding. The material cannot be made available by changing an accounting category.
A useful inventory picture therefore includes quantity, location category, condition and expected return time, while protecting sensitive facility details. The public does not need a map of every component to understand the principle. It needs to know whether the stated fuel reserve means total material somewhere in the installation or material actually available to support the stated operating commitment. Those are different promises.
Circulation is larger than consumption
A fuel-cycle diagram can hide a surprisingly important distinction between material consumed and material circulated. If only some of the fuel entering a reaction region reacts during a pass, the remaining material still has to be managed. The processing burden is linked to that circulating flow, not just to the smaller amount that disappears through fusion. Improving how effectively fuel is used can therefore change requirements elsewhere in the plant.
Consider a toy system that circulates 100 inventory units in a defined interval and consumes five of them. The amount requiring continued management is much larger than the five units that need replacement because of consumption alone. If the system later circulates the same amount but consumes ten, its consumption and return-flow balance changes. Neither scenario establishes an actual plasma operating point. The example simply prevents a reader from treating consumption as a complete measure of processing throughput.
Fusion for Energy and EUROfusion's 2025 fuel-cycle technology mapping report describes a broad system encompassing pumping, purification, storage and related technologies. Its overview highlights the inventory tied up during fuel processing and the motivation for shortening treatment time. The final-publication version is dated July 4, 2025. It is a technology mapping exercise, not a demonstration that every listed approach is ready for a power plant. Fuel Cycle Technology Mapping Report
For an analyst, the next question is what happens when throughput changes. A component adequate for an isolated test may become a bottleneck when linked to continuous circulation. A specification that states only total capacity can miss a delay introduced by batch operation. A nominally fast subsystem can also be slowed by the need to wait for another subsystem's result. These are questions about the behavior of the connected arrangement, not accusations against any particular technology.
The distinction also affects how demonstrations should be described. Successfully processing a representative material establishes something useful. Demonstrating the required throughput over a representative duration establishes more. Showing that several components maintain compatible flows through changes in operating state establishes another layer. A public account should identify which layer has been tested rather than allowing the word demonstrated to expand silently.
A surplus is useful only after it arrives
A breeding ratio expresses replacement production relative to fuel consumption under a defined calculation. A value above one is attractive because replacement must cover more than the fuel consumed in the reaction. But a gross production ratio does not directly tell an operator how much usable fuel will be available tomorrow. Recovery, retention, decay and time delays intervene between the production event and the next use.
Suppose an invented calculation reports 110 units created for every 100 consumed. The gross ratio is 1.10. Now suppose only 104 of those new units become available within the accounting period, while other needs consume three units. The usable surplus over the 100-unit consumption is then one unit, not ten. The missing six units have not necessarily escaped into the environment; they may be delayed or retained. A correct interpretation requires identifying what happened to them.
This example is intentionally abstract and does not prescribe a breeding target. Its purpose is to separate a gross ratio from a timely usable balance. A different system could have the same gross ratio and a different recovery schedule. A different accounting period could also change the apparent balance by including material that had not yet returned at the earlier cutoff. Comparing ratios without their definitions can therefore produce a false disagreement or false agreement.
A 2024 paper by Michael Lord and colleagues describes an outline fuel-cycle and breeder design for the UK's STEP project. It explains the relationship between self-sufficiency, circulation and inventory, and distinguishes a practical minimum breeding objective from a higher early-design target intended to provide margin. These are design-specific calculations and objectives, not an experimentally established universal threshold for fusion plants. Lord and colleagues' STEP fuel-cycle paper
The transferable lesson is that margin has a job. It may accommodate uncertainty, design changes, delayed recovery or future inventory requirements. If those demands are omitted from an early estimate, they can reappear as a surprise later. A credible project explains what its margin is intended to cover, how uncertainty was estimated and which changes would consume it. The public can then understand why a seemingly small design alteration might matter to the fuel balance.
The start of the cycle is a different problem
A self-sustaining process still needs to start. Before a proposed plant can produce replacement tritium through its own fusion operation, it needs enough usable fuel to begin operating and populate the necessary parts of the cycle. Its startup need is therefore different from its eventual net annual consumption. A claim about steady operation cannot by itself answer the startup question.
Imagine a closed-loop demonstration with an external starting inventory of twenty abstract units. During commissioning, twelve become committed to the system and eight remain readily available. If an unexpected delay prevents material from returning when planned, the total inventory can remain twenty while the available reserve falls toward a limit that stops further testing. The failure would be one of timing and accessibility, not necessarily insufficient total ownership.
The same logic applies to a proposed second machine. Fuel that the first machine expects to export eventually cannot support the second machine's startup before it becomes available. A fleet schedule must connect those dates explicitly. Treating every plant's future surplus as immediately accessible can create an expansion plan in which several machines depend on fuel that none has yet produced.
This dependency does not show that expansion is impossible. It identifies a variable that expansion forecasts must represent. A slower rollout, lower startup requirement, more rapid usable recovery or a different supply arrangement could alter the result. Each option brings its own evidence requirements and constraints. The responsible comparison is between complete scenarios, not between an optimistic sentence and a pessimistic sentence.
The word self-sufficient can also conceal the time horizon. It might mean that imports are unnecessary during a selected steady operating period, that the full life cycle balances after an initial supply, or that a plant can support additional startups. These claims are progressively different. A reader should ask which one is intended and whether the calculation includes commissioning, outages and retirement rather than only the most favorable operating phase.
Delay creates an inventory requirement
An elementary relationship makes processing time tangible. In an ideal steady stream, the average material tied up in a process equals its throughput multiplied by its average residence time. If a fictional subsystem handles ten abstract units per hour and material spends an average of two hours inside it, the average amount within that subsystem is twenty units. Shortening the average residence time to one hour reduces that simplified amount to ten units at the same throughput.
This is not a tritium-system design formula sufficient for procurement. Real behavior can include distributed residence times, batch schedules, inaccessible retention and changing flows. It is an accounting relationship that reveals why speed and inventory are linked. A subsystem can meet a daily throughput target while holding more fuel than a project expected. That matters when startup stock and safety-related inventory constraints are important.
Average time also differs from the tail of the return-time distribution. Suppose most material returns quickly but a small fraction remains much longer. An average can describe that mixture while leaving the operator exposed to a persistent unavailable fraction. The appropriate model should represent the behavior relevant to the operating decision. Replacing a distribution with a single convenient delay should be a justified simplification, not an unnoticed one.
An integrated test can investigate this by following the system's response to a controlled change in its normal test program, without assuming that every measured response comes from the same mechanism. Some changes may reflect physical transport; others may reflect instrument delay or the scheduling of measurements. Separating those explanations is essential before claiming that a new component has shortened the real fuel turnaround time.
The economic implication is subtler than simply buying a faster machine. Reducing holdup may reduce the initial fuel commitment, improve responsiveness or make a reserve more effective. It may also require other equipment, greater complexity or additional maintenance. The appropriate comparison includes the whole operating consequence. A component's speed matters because of what it enables in the combined system, not because the fastest isolated number always wins.
Materials remember the operating history
A circulating material can interact with the surfaces and structures around it. Some interactions may temporarily remove it from the readily available flow; others may affect how components behave over time. For a proposed fusion plant, the relevant questions include how the operating environment changes those interactions and how much confidence can be placed in measurements made under less demanding conditions.
The US Department of Energy's Fusion Science and Technology Roadmap, published in June 2026, identifies unresolved fuel-cycle and accountancy challenges. It specifically calls for validated measurement approaches and integrated testing, and notes gaps involving material retention and nuclear conditions. The roadmap is a statement of development priorities, not a report that those gaps have already been closed; it also disclaims a commitment to specific funding levels. DOE's June 2026 roadmap
The original analytical consequence is that history belongs in the model. A new component and a previously exposed component may not be interchangeable experimental objects. If performance changes with prior operation, the record must preserve that prior operation well enough to interpret the change. Otherwise a later measurement may look like random disagreement when it actually reflects different material histories.
A small sample test can be especially valuable when its question is precise. It might investigate one interaction under stated conditions. Scaling that result to a plant requires identifying which aspects of the sample environment are representative and which are not. Geometry, exposure history, interfaces and temperature variation may all complicate transfer. The answer is neither to dismiss small experiments nor to let them stand in for an entire lifetime demonstration.
Maintenance introduces another boundary. Removing a component from service does not necessarily return all associated fuel immediately to the operating reserve. Material may remain with the component while it is characterized or managed through authorized processes. A lifetime balance therefore needs to connect the operating system with its maintenance and retirement pathways. Otherwise the apparent closed cycle can depend on ignoring material whenever it leaves the main process diagram.
Accountability is an engineering measurement
Fuel accountability is sometimes imagined as paperwork added after the real engineering. In a tightly coupled system, it is part of knowing whether the engineering works. An operator needs to distinguish material consumed, material transferred, material retained and measurement error. Without that distinction, a favorable balance may reflect an incorrect estimate just as readily as successful recovery.
Suppose a hypothetical experiment begins with 100 units, receives ten more and ends with 105 measured units after five units are known to have been consumed. The simple balance closes. Now suppose the final measurement is uncertain by several units and the incoming measurement has its own uncertainty. The arithmetic still closes at the central estimates, but the strength of the conclusion depends on those uncertainties. Equality between rounded numbers is not a demonstration of perfect knowledge.
The reverse is equally important. A mismatch does not immediately identify an environmental release or wrongdoing. It can arise from calibration differences, time boundaries, an omitted inventory category, retained material or an actual loss. Each possibility requires investigation appropriate to its consequences. Public reporting should avoid both dismissing discrepancies automatically and assigning a dramatic explanation before the evidence distinguishes causes.
Timing can create an apparent discrepancy even with accurate individual instruments. If one record includes a transfer before the reporting cutoff and another records its receipt after the cutoff, the totals will not reconcile for that interval. Shared time definitions and explicit in-transit categories make the discrepancy intelligible. This is a general accounting principle, but its practical implementation must fit the physical system and the governing safety arrangements.
The most informative performance report would therefore show how closure was assessed, what uncertainty remained and whether unexplained differences persisted across repeated periods. It would not need to disclose sensitive operational details to explain those methods. Independent review can examine confidential evidence under appropriate arrangements while the public receives a bounded account of what was demonstrated and what remains unresolved.
A test facility should answer an integration question
The UK Atomic Energy Authority's strategy for 2026–2030 identifies fuel-cycle research and facilities including H3AT and LIBRTI. Its timeline distinguishes development, commissioning and later use. Those entries are planned milestones, not evidence that every listed facility has already reached its final operating state. The strategy illustrates the institutional attention being given to fuel-cycle development alongside plasma work. UKAEA's 2026–2030 strategy
A good test facility is valuable because it can resolve a question that separate component tests cannot. For example, two devices may each satisfy their own specification while requiring incompatible schedules. One may deliver material in bursts while the next expects a steady input. Another may generate a measurement only after a delay that prevents the controller from responding as assumed. These are illustrative integration problems, not reports of failures at H3AT or LIBRTI.
A useful demonstration begins by identifying the uncertainty it is intended to reduce. Is the question about throughput, inventory, recovery time, measurement confidence or the response to an interruption? A long period of operation may answer one of those questions poorly if the relevant condition never occurs. Conversely, a shorter, carefully characterized experiment may expose an important incompatibility before a larger installation is committed to it.
The sequence of demonstrations should preserve what each stage actually established. Operation with nonradioactive substitutes can test some mechanical and control behavior. It cannot establish every isotope-specific or radiation-related effect. An experiment using relevant materials under limited conditions can add another layer. An integrated test under representative conditions adds more. The public story should identify the transitions rather than calling the entire sequence complete at its first successful stage.
There is also value in an inconclusive result. If an instrument cannot distinguish the expected effect from its uncertainty, that finding identifies a measurement problem worth solving. If a test does not reach the intended condition, it can still inform equipment development, but it should not be counted as evidence for the untested condition. A disciplined program gains credibility by explaining those limits before they become exaggerated claims elsewhere.
Outages change both sides of the balance
A plant that stops producing fusion energy also changes its fuel-production and consumption pattern. It may stop consuming fuel in the reaction while other parts of the system continue processing or remain unavailable. Some inventories may return during the outage; others may remain committed to equipment. The direction of the resulting change cannot be inferred merely from the word shutdown.
Consider a fictional maintenance period during which the main reaction is paused but previously circulated material continues returning to an available reserve. The reserve could increase even while breeding is absent. In a different maintenance event, a component containing retained material could be removed and the reserve could be constrained. Both situations are compatible with a plant having the same steady-state breeding ratio before the outage.
This is why availability and self-sufficiency must be evaluated together. An annual calculation that multiplies full-power performance by a percentage may capture some effects while missing the sequence of interruptions. Two plants with equal annual operating hours can have different startup and reserve requirements if one operates in long continuous campaigns and the other experiences many interruptions. The ordering of events matters when flows have memory and delay.
A credible operational scenario includes the recovery from an interruption. How long before the system again supplies fuel of the required condition? Which measurements establish readiness? What material remains unavailable during that transition? Those questions should be addressed through authorized engineering and safety processes, not improvised from a favorable annual average. A public assessment can ask whether they were addressed without publishing detailed procedures.
The financial implications also deserve separate treatment. An outage can reduce energy revenue while leaving staffing, capital and other obligations in place. If maintenance also ties up scarce fuel, the restart schedule may become dependent on a resource omitted from a conventional availability estimate. A commercial model should connect those dependencies rather than assigning every subsystem an optimistic independent availability and assuming the plant inherits them all.
A growing fleet cannot borrow from its future twice
Fusion's proposed role in energy systems is usually larger than one successful machine. A fleet introduces a collective inventory question. If each new plant needs a startup commitment and existing plants produce only a limited usable surplus, the timing of deployment depends on how those quantities evolve together. The result is a system-level constraint even if each individual design appears capable of eventual self-sufficiency.
A simple hypothetical illustrates the arithmetic. Suppose an operating plant can make two abstract units per year available beyond its own continuing needs, and a new plant requires ten units before starting. Ignoring all other sources, decay and delays, accumulating that commitment takes five years. If two new plants each count on those same ten units at the same date, the plan has double-counted the first plant's contribution.
Real scenarios would require more detail, including multiple suppliers, uncertain startup schedules, changing operating performance and authorized transfers. The toy example is not a forecast of fusion deployment. It shows why a list of company target dates cannot be summed into an energy forecast without checking shared dependencies. Different developers can make individually understandable plans that are collectively inconsistent if they rely on the same limited resource.
The appropriate response is to model alternatives transparently. A scenario might assume slower deployment, lower initial inventory per machine or different timing for usable surplus. Another might depend on an additional lawful supply source. Those assumptions should be visible and distinguished from demonstrated availability. A sensitivity analysis can then reveal which assumptions dominate the outcome instead of concealing them inside a single smooth growth curve.
There is a policy consequence as well. Research facilities and first-of-a-kind plants may compete for access to the same enabling capabilities even when their long-term goals align. Coordinating development can therefore involve more than funding individual devices. It can involve measurement infrastructure, qualified personnel and scheduling of shared test capacity. These are ordinary industrial-development problems made more consequential by a fuel cycle that must become established as the industry expands.
Two balances that can disagree without contradiction
A useful way to picture the cycle is to draw two boxes rather than one. The outer box contains all the material associated with the plant. The inner box contains only the material immediately available for the next planned operation. A transfer between those boxes can change operational readiness without changing the outer total. Consumption or a transfer beyond the outer boundary changes the total as well. Confusing these two kinds of change makes discussions of inventory needlessly mysterious.
Suppose an entirely hypothetical outer box contains 50 units and its available inner portion contains 15. Over a short interval, four units return from processing, six leave the available portion for a new operation, and two newly recovered units enter it. The available balance ends at 15 again: 15 plus four minus six plus two. That stable reading does not mean nothing happened. Twelve units of transfers crossed its boundary, and the processes responsible for those transfers may have changed the distribution elsewhere.
Now delay the four-unit return until the next interval while leaving the other events unchanged. The available portion ends at eleven. The delay alone has not established a loss from the outer box. It has changed readiness. If a report presents only the total, it can conceal that operational constraint. If it presents only the available amount, it can invite a mistaken inference that four units disappeared. Both balances are useful because they answer different questions.
The example also reveals why a closing inventory is insufficient to describe a run. Two experiments can begin and end at the same values while taking different paths between them. One might maintain a comfortable available reserve throughout. Another might approach an operating limit and recover just before the final measurement. A time series can distinguish those experiences; the final pair of numbers cannot. The minimum available amount and the duration of any constrained period can matter alongside the net change.
Control decisions add another layer. Suppose the system intentionally reduces throughput to preserve a reserve. A later report could show an acceptable inventory balance because production was curtailed. That can be the correct operational response, but it should not be interpreted as proof that the original full-throughput schedule was sustained. Performance statements need to connect the inventory record to the actual operating record. Otherwise the intervention that protected the system disappears from the explanation of why it appeared stable.
For public reporting, this need not become a disclosure of detailed equipment inventories. Aggregated and appropriately protected results can still distinguish total stock, available reserve, throughput and curtailed operation. The important intellectual move is to stop demanding that one number represent all four. A transparent boundary definition makes an inventory trend understandable and makes later changes easier to interpret. It also gives developers a fairer way to explain a successful protective response without disguising it as uninterrupted operation.
What a convincing announcement would contain
A useful announcement about a fuel-cycle advance would begin with the exact achievement. It might report a component measurement, a validated model, an integrated run or a design review. Those categories should not be interchangeable. The announcement should state the tested boundary, the duration, the relevant conditions and the comparison used to judge improvement. It should then identify the remaining gap to the intended use.
For a modeling result, the reader needs to know which inputs are measured and which remain assumptions. Agreement with one dataset can support a model within that tested domain without validating every extrapolation. A prediction about a future plant should retain that status even when the underlying physics is well established. Uncertainty often enters through geometry, material behavior, operating history and integration rather than through doubt that conservation laws apply.
For an experimental result, the reader needs more than the best observed efficiency. Repeatability, measurement uncertainty, excluded periods and the condition of the test article all matter. If a component was new, the report should not imply lifetime performance. If the experiment did not include representative nuclear conditions, the transfer of its result to those conditions remains a separate question. None of those qualifications makes the measurement unimportant.
For a review milestone, the report should identify whether the review was held, whether actions remain open and whether approval or closure has occurred. The August 2026 blanket announcement is useful precisely when read at that level. A review process organizes evidence and exposes issues; it is not itself a measurement of fuel self-sufficiency. Conflating the two would make genuine engineering progress harder to understand, not easier.
Finally, a convincing account should say what would change the conclusion. A new measurement could reveal higher retention, a changed component could reduce margin, or an integration test could show a longer return time. Naming those possibilities gives the audience a way to follow subsequent progress. It replaces the demand for unconditional optimism or pessimism with a set of observations that can actually settle narrower questions.
The achievement is a working circulation
The fuel-cycle challenge is not a rhetorical objection invented to move the goalposts after advances in plasma physics. It follows from the ambition to operate a deuterium–tritium power plant repeatedly, predictably and at useful scale. The fuel must exist at startup, return on time, remain accounted for and support the operating schedule. Producing replacement atoms is essential; turning them into a dependable circulation is the broader achievement.
Nor does acknowledging that challenge establish that fusion cannot contribute to future energy systems. The inspected primary sources show active design work, research planning and the development of facilities intended to investigate it. They also show why a careful assessment should avoid declaring the problem solved on the basis of a gross ratio, a single component result or a proposed milestone. Progress occurs at several levels, and those levels need their own evidence.
For readers, the most productive questions are now quite specific. What does the quoted gain measure? Does the fuel figure describe gross production or timely usable recovery? How much material is committed to circulation rather than available as reserve? Which operating histories have been tested? Does the expansion scenario count the same future surplus more than once? These questions expose a proposal's logic without requiring the reader to design a reactor.
A power plant earns confidence when its linked systems support the promised service under defined conditions. In fusion, that service depends on a loop that crosses the plasma boundary and returns through materials, processing, measurement and time. Following that loop does not take attention away from the scientific achievement. It explains what must join the achievement before a bright experiment can become dependable electricity.