Evidence checked October 31, 2026. Regulatory decisions, project estimates and achieved operating results are kept separate. Numerical teaching examples are hypothetical.
1. A permit is progress with a defined boundary
On September 28, 2026, the US Nuclear Regulatory Commission issued a construction permit for Tennessee Valley Authority's Clinch River Nuclear Unit 1. The Federal Register notice records an actual regulatory decision, not a developer's aspiration or an application awaiting review. It is an important milestone. It is also a construction permit, not a report of electricity already reaching customers. NRC construction-permit notice
The distinction matters because small modular reactors are increasingly discussed as answers to immediate electricity demand. A project can be strategically important while remaining several consequential steps away from dependable operation. Compressing those steps into the word “approved” makes the story easier to tell and harder to understand.
Canada provides another useful example. The Canadian Nuclear Safety Commission reports that Ontario Power Generation holds a licence to construct one BWRX-300 at Darlington. A first regulatory hold point was removed on March 30, 2026. OPG also applied in March for an operating licence; the regulator describes that application as subject to a later Commission decision. These are separate events with separate implications. CNSC Darlington project record
A reader does not need to decide whether nuclear power is universally good or bad before interpreting this evidence. The immediate task is narrower: establish what a named project has demonstrated, what remains conditional and which organisation bears each unresolved risk.
The engineering design is only one part of the answer. The project also needs a suitable site, a construction organisation, qualified suppliers, usable fuel, finance, operating staff and a path through testing and commissioning. Success in one category can reduce uncertainty without resolving the others.
This article follows that project chain. It does not offer an investment recommendation or a legal opinion about any licence. Its purpose is to make announcements comparable. A credible assessment should allow licensing progress to count as licensing progress, while reserving claims about delivered electricity and cost for the evidence that can actually support them.
2. The word approval covers different decisions
Regulatory vocabulary is not interchangeable across countries or licensing pathways. A design review, a site decision, a construction authorisation and permission to operate answer different questions. Their names may sound similar to a general reader, but the underlying scope determines what has been established.
The Canadian regulator makes one particularly important distinction explicit: its optional vendor design review neither issues a licence nor certifies a reactor design. It is an early review intended to identify potential issues. Describing completion of that process as permission to build a plant would overstate the decision. CNSC vendor design review explanation
The analytical lesson extends beyond Canada. Whenever an announcement uses “approved,” the next sentence should identify the approving body, the exact document, the design or facility covered, and the activities it authorises. If conditions remain, they belong in the description. If a decision applies to one unit, a proposed fleet should not inherit that status automatically.
A useful way to read the record is as a set of bounded claims. The design may have reached a particular review stage. The proposed location may have passed a separate assessment. The owner may have permission for specified construction activities. The assembled plant may still need tests and further decisions before operation. None of these distinctions diminishes the work already completed.
They do prevent false comparisons. A company with an early design review and another with a site-specific construction authorisation are not at the same point merely because both can cite a regulator. Conversely, an older design with extensive operating experience may still face substantial project-specific obligations at a new location.
The dates matter too. A press release can remain online long after the project changes. A current assessment should reconcile the announcement with the latest accessible regulator record rather than treating the first favourable milestone as a permanent description of project status.
A precise account therefore uses verbs carefully: submitted, accepted for review, reviewed, authorised, constructed, tested and operating. Each verb marks a different evidentiary threshold. The public debate becomes more useful when those thresholds remain visible instead of collapsing into a binary label of approved or unapproved.
3. Small and modular describe a strategy, not an outcome
A smaller unit can reduce the size of an individual investment compared with a larger plant, but that does not automatically reduce cost per unit of electricity. A modular construction strategy may shift work into factories, but it does not eliminate the need to assemble, connect and test the complete facility. These are economic and organisational hypotheses that must be demonstrated in projects.
The OECD Nuclear Energy Agency's SMR Dashboard deliberately looks beyond technical feasibility. Its readiness framework includes licensing, siting, financing, supply chain, engagement and fuel. That multidimensional approach is useful because a reactor concept can be advanced in one dimension and immature in another. A single ranking based only on design maturity would miss those differences. NEA SMR Dashboard framework
The word modular can refer to several arrangements. Components may be manufactured as transportable assemblies. A site may contain multiple reactor units. A standard design may be intended for repetition across many locations. Those features can overlap, but they are not the same claim. A project should explain which type of modularity is supposed to create its savings.
Factory manufacture can make work more repeatable and controlled, yet it introduces its own dependencies. The factory needs qualified processes, orders, workers, equipment and a reliable flow of materials. The modules need transport arrangements and interfaces that fit the site. If the design changes late, the factory may produce rework more consistently rather than producing savings.
Multiple units can spread some common costs, but their benefits depend on completing enough of the planned programme. A first unit carrying shared infrastructure may look expensive relative to later units. A later-unit estimate may look unusually attractive because the earlier investment is excluded. Both numbers can be correctly calculated and still mislead if their boundaries are not disclosed.
The right question is therefore not whether modularity is inherently real or imaginary. It is where the project expects repetition to reduce labour, schedule risk or material use, and what evidence supports that expectation. An actual production line, a qualified supplier and a completed installation provide different evidence from a rendering of a future factory.
Small reactors deserve assessment as projects with specific designs and supply chains. Treating every proposal as a member of a single economic category hides the very differences that will determine which approaches succeed.
4. Construction is a sequence of dependencies
A construction schedule is often presented as a date attached to the first concrete pour and another attached to commercial operation. The work between those dates is better understood as a network of dependencies. Some activities can proceed in parallel; others cannot start until a prior result is accepted, equipment arrives or an interface is resolved.
Darlington's 2025 construction authorisation provides a concrete example of a decision with conditions and scope. The Canadian regulator authorised one BWRX-300 and stated a licence validity period ending March 31, 2035. A licence's expiry date is not a promised completion date. It defines an authorisation period, while the project schedule remains a separate matter. CNSC April 4, 2025 decision announcement
For analytical purposes, imagine a project with civil works, major equipment delivery, electrical installation and commissioning. If a critical component arrives late, some other work may continue, but the final completion date can still move. Adding workers to an unrelated task does not necessarily recover time on the controlling sequence.
This is why percentage-complete figures need context. A project can finish a large amount of low-risk work while a small amount of unresolved work determines the remaining schedule. Expenditure is not a reliable substitute for physical completion, and physical completion is not a substitute for accepted performance.
The evidence should identify the schedule's critical dependencies and the basis for contingency. Does the timeline assume that equipment arrives on a contracted date or merely a supplier's preliminary estimate? Does it include time for correcting test findings? Are regulatory submissions sufficiently mature to support the assumed review sequence?
A credible schedule also records changes. When a date moves, the explanation should distinguish scope changes, unexpected conditions, procurement delays, productivity problems and revised assumptions. These causes have different implications for later projects. A delay caused by a one-time first-unit design issue may be less repeatable than a persistent shortage of qualified components.
Schedule transparency is not an argument for predicting failure. It is how a project demonstrates that its forecast rests on an organised delivery plan. The more ambitious the promised completion date, the more useful it becomes to show the dependencies that make that date plausible.
5. First-of-a-kind costs require careful denominators
OPG's current Darlington project page states a first-SMR cost of C$6.1 billion, plus C$1.6 billion for systems and services common to all four planned units. It gives an estimated total budget of C$20.9 billion for the four-unit programme and expects costs to decline with subsequent units. These are the owner's estimates and expectations, not final audited outturn costs. OPG Darlington SMR project information
Those figures illustrate why a cost comparison needs a boundary before it needs a headline. One reader may assign all shared infrastructure to the first unit. Another may allocate it across four. The resulting first-unit figures differ even when both use the same underlying expenditure. Neither method should be hidden.
A project comparison also needs consistent treatment of financing, inflation, contingency, owner costs, grid connection and decommissioning provisions. A quoted engineering cost excluding interest cannot be compared directly with a programme budget that includes financing and escalation. Currency and price year must be identified before apparent differences are interpreted as technological advantages.
Cost per kilowatt is useful for comparing installed capacity, but it does not describe the cost of delivered electricity by itself. The latter depends on operating performance, fuel, maintenance, financing terms, lifetime and other factors. A plant with a lower initial cost can still have a higher unit energy cost if it delivers substantially less electricity than assumed.
A simple hypothetical example makes the denominator visible. A 300-megawatt plant operating at a 90 percent annual capacity factor would generate 300 multiplied by 8,760 multiplied by 0.9 megawatt-hours, or 2,365,200 megawatt-hours. At 75 percent, it would generate 1,971,000. The same annual fixed cost would be spread over roughly 17 percent less output in the second case.
This calculation is not a forecast for a named reactor. It shows why a cost model should publish its operating assumptions. A low price based on mature-fleet availability is not evidence that a first unit has already achieved that availability.
The most informative cost statement separates observed expenditure, committed contracts, remaining estimates and contingency. That structure lets readers see which parts of the budget are anchored in transactions and which still depend on successful future execution.
6. Financing changes the price of waiting
A delayed project can become more expensive even if the physical design stays unchanged. Money committed during construction has a cost, and revenue arrives only when the relevant service begins. The interaction between schedule and finance is especially important for projects with large upfront expenditure.
Consider a deliberately simplified hypothetical case. If one billion currency units remained outstanding for an additional year at an illustrative annual financing cost of 8 percent, that extra year would add 80 million before considering compounding, fees, changing balances or other effects. The example is not a financing estimate for any reactor. It isolates one reason schedule certainty has economic value.
The ownership model changes who experiences that cost. A private developer, a public utility and a regulated utility can finance and recover expenditure differently. A government guarantee may lower borrowing costs while transferring some risk to the public. A long-term customer contract may improve revenue certainty while creating obligations for the buyer. Lower financing cost is a real benefit, but it should not be confused with the disappearance of risk.
A useful project account therefore asks who pays if construction takes longer, output begins later or additional capital is required. The answer may be shared among contractors, owners, customers and governments. A contract can allocate a risk on paper, but the allocation is only as strong as the counterparty's ability to perform.
The phrase fully funded also needs definition. Funding for development studies is different from committed financing for complete construction. A conditional financing agreement may depend on milestones that have not yet been reached. A public announcement of support may be significant without being equivalent to cash available for every remaining obligation.
These distinctions matter when comparing technology options. A project with unusually favourable finance may appear cheaper than another because of policy support or ownership structure, not solely because its engineering is more efficient. Both effects can be relevant to a decision, but they should be reported separately.
The analytical goal is not to strip finance out of the story. Finance is part of how infrastructure gets built. The goal is to explain whether an attractive electricity price depends on repeatable construction economics, a particular allocation of public risk, or a combination of the two. Readers should not have to infer that distinction from a promotional tariff.
7. Fuel is a chain, not a commodity label
Some advanced reactor designs require high-assay low-enriched uranium, while other proposed small reactors use different fuel specifications. It is a mistake to assume that every SMR has the same fuel requirements. A project's actual design determines the relevant supply chain and qualification work.
GAO's 2026 nuclear-fuel report describes a chain that includes mining and milling, conversion, enrichment, deconversion and fabrication. It also identifies weaknesses in planning and cost information for federal efforts to support uranium supply. That is evidence about the programme and its management, not proof that every reactor will face the same shortage or delay. GAO-26-107385 nuclear fuel report
For a project buyer, the important distinction is between material at one stage and usable fuel delivered in the required form. An announcement about enrichment capacity does not establish that fabrication, transport, licensing and delivery are all ready. Each step needs to match the reactor's specifications and schedule.
The unit of measurement can also confuse the discussion. A mass of uranium, a mass of a chemical compound containing uranium and the mass of finished fuel are not interchangeable quantities. A supply announcement should say what is being measured and at which stage. Without that information, comparing demand and supply can produce a false impression of sufficiency.
A supply contract is stronger evidence than an informal expression of interest, but its terms still matter. Is the quantity reserved? Are delivery dates firm or conditional? Does the agreement cover the initial core only, or subsequent operating needs as well? What happens if a facility needed elsewhere in the chain is late?
Fuel planning also has an operating dimension. A plant designed around a particular fuel form cannot casually substitute whatever material becomes available. Qualification and regulatory obligations constrain changes. The economics of fuel security therefore include inventories, alternative suppliers where feasible and the consequences of interruption.
None of this requires readers to become nuclear-fuel specialists. The practical question is whether the project can trace its fuel from credible production capability through all necessary processing steps to a usable, authorised product at the plant. A reactor design without that chain may be technically compelling and still unable to begin operation on its advertised date.
8. Supply-chain capacity must match the design
An SMR programme can announce many potential units before suppliers have demonstrated the capacity to deliver their components at the required rate. The resulting order pipeline is commercially interesting, but it is not identical to a production system. The distinction becomes sharper when the design contains components with demanding quality requirements or long manufacturing lead times.
For analytical purposes, divide the supply chain into three groups: items that are readily available, items that require qualified specialist production, and items that need new capability. This is a proposed reporting framework rather than an official classification. It helps identify where the project depends on existing capacity and where it depends on future investment.
A supplier agreement can fall anywhere along that spectrum. It may reserve a slot in an operating facility, fund a capacity expansion, support development of a new process or simply establish an intention to collaborate. Press releases often use similar language for these very different commitments.
The key evidence includes manufacturing readiness, quality assurance, delivery schedules and the treatment of design changes. A factory cannot optimise repeat production if the specification remains unsettled. Conversely, freezing a design too early can lock in issues that later require costly correction. The project needs a disciplined process for resolving that tension.
Interfaces deserve particular attention. A component can meet its own specification and still create problems when connected to another system. The responsibility for dimensional tolerances, controls, installation sequence and testing must be clear. Modularity makes those interfaces more important, not less.
A multi-project programme also creates competition for the same qualified resources. If several developers assume they will each obtain the earliest available manufacturing slots, the aggregate schedule may be impossible even though every individual plan looks plausible. Public assessments should distinguish announced demand from contracted capacity.
The strongest supply-chain claim is therefore specific: a named component or service, a qualified supplier, an agreed specification, a delivery commitment and a credible plan for verification. The number of companies listed in a partnership announcement is a much weaker indicator. A working project needs the right parts in the right sequence, not merely a long list of organisations willing to discuss them.
9. Commissioning is where the assembled plant answers back
A completed building does not establish that the plant works as intended. Commissioning tests the installed systems and their interactions, resolves deviations and prepares the facility for operation. Factory testing can provide valuable evidence, but the final arrangement at the site introduces interfaces and conditions that must also be addressed.
IAEA guidance on commissioning discusses the relationship between factory tests and site conditions, the validation of procedures, and updating documentation in response to test results. The broad lesson is that commissioning is an evidence-producing stage, not a ceremonial interval between construction and revenue. IAEA commissioning guidance
For a first-of-a-kind plant, this stage is particularly informative. The project may discover that equipment performs differently when integrated than it did in isolated tests. Procedures may need revision. Instrumentation may reveal behaviour that requires further analysis. Finding and resolving those issues is part of commissioning's purpose, not automatically evidence of incompetence.
The schedule should include realistic time for that work. A timeline that assumes every test passes on the first attempt is a different forecast from one that includes credible allowances for diagnosis and correction. If the contingency is omitted, the apparent speed may reflect optimism rather than improved engineering.
Public reporting should distinguish first criticality, first electricity, operation at a selected power level and commercial service. These milestones can all be meaningful without being equivalent. A brief generation event does not establish annual availability, and a successful test at one operating point does not establish the entire operating envelope.
The buyer's relevant question is when the plant can reliably meet the contracted service, under the applicable authorisations and operating procedures. That requires more than showing that the generator has turned. It requires evidence about sustained performance, operational readiness and the handling of foreseeable outages.
A good project narrative therefore continues after the first electricity announcement. It reports the progression through testing and the transition to routine operation. That longer account is where the promised benefits of a standardised design begin to become measurable, and where remaining first-unit lessons can be captured for later units.
10. Safety claims need their own evidence
SMR discussions often combine economic and safety claims in the same sentence. A design may incorporate features intended to simplify operation or improve safety, but those claims should be evaluated through the relevant technical and regulatory evidence. They do not automatically establish construction cost, delivery schedule or commercial availability.
The Canadian regulator's supplemental information for SMR proponents recognises the diversity of designs and the value of early identification of potential regulatory or technical issues. This supports a design-specific approach rather than assuming that every small reactor shares one safety case. CNSC supplemental information for SMR proponents
A public explanation should state what feature is being claimed, under which conditions it functions and how its performance is supported. Terms such as passive or inherent can be meaningful in engineering, but they are not substitutes for describing the systems and assumptions involved. A concise explanation can remain accessible without implying that a label removes every hazard.
The project also needs organisational capability. Training, maintenance, configuration control and the quality of operating procedures affect how the installed plant is managed. A design assessment cannot by itself establish that a future organisation will perform every task effectively over decades.
For readers, it is useful to keep three questions separate. What is the design intended to do? What has the regulator accepted for the specific application? What has the built facility demonstrated? Confusing these levels can either overstate certainty or unfairly dismiss completed technical work.
The same discipline applies to waste and decommissioning. A smaller unit still needs an appropriate plan for the materials and obligations it creates. Comparisons should use a defined output basis and a specified fuel and operating case rather than assuming that smaller nameplate capacity settles the issue.
This article does not attempt to substitute a general argument for the detailed safety assessment of any named reactor. Its point is evidentiary: safety, cost and schedule each need their own support. A favourable result in one category should not be used as a rhetorical shortcut around the others, and a difficulty in one category should not automatically invalidate every claim in the rest.
11. A fleet promise needs an order sequence
The economic case for standardisation often depends on building a series of similar units. Repetition can support learning, supplier investment and more predictable work. But a proposed fleet is not the same thing as a committed order sequence, and the difference matters to both factories and customers.
Imagine a hypothetical supplier that can justify a dedicated production line only if it receives ten orders over a defined period. Several developers express interest in ten units collectively, but each plans to wait for the first project's results before committing. The apparent market contains ten possibilities while the factory has only one firm order. The financing problem has not disappeared merely because the aggregate pipeline is large.
The reverse problem is also possible. A factory expands for expected demand that arrives later or in a different design configuration. Its costs then need to be recovered over fewer units, or the investment becomes underused. A learning-curve forecast should therefore state the volume and timing assumptions that support it.
A claimed percentage cost reduction per doubling of production is a model parameter, not an observed law of nature for every technology. It can be useful for scenario analysis, but the reader should ask what evidence supports the rate and which costs are expected to decline. Site-specific work may respond differently from repeat factory manufacture.
Design stability is another condition. If later units are substantially redesigned, some lessons and tooling may transfer while others do not. Calling every version the next unit in a sequence can exaggerate the amount of repeat production actually occurring.
The strongest fleet case connects three records: the technical design baseline, the committed customer sequence and the supplier capacity plan. If one changes, the economics should be updated. A programme can still be promising while those records are incomplete, but its forecast should reflect the uncertainty.
For public decision-makers, this creates a choice about supporting early units. The possible benefit is knowledge and capability that later projects can use. The corresponding obligation is to identify that public benefit explicitly rather than presenting the first project as though all mature-fleet savings have already been earned. Learning has value, but it is not free evidence of future low prices.
12. The electricity service must be specified
A reactor's nameplate capacity does not describe every aspect of the service a customer receives. Buyers may care about annual energy, dependable capacity, availability during particular periods, operational flexibility or a combination of these. A project should state which service its economics and contract are designed to provide.
A hypothetical 300-megawatt unit produces a different risk profile from a group of smaller units with the same aggregate capacity. The distribution of outages can differ, as can shared-system dependencies. Multiple units may reduce exposure to a single-unit outage, but a common cause or shared infrastructure problem can affect more than one. The relevant design details matter.
The annual capacity factor is useful but incomplete. Two plants can produce the same annual energy while having different outage timing. If one is unavailable during a period of high system stress, the consequences for a customer or grid can be greater than the annual average suggests.
Likewise, flexibility needs definition. A claim that a reactor can change output should identify the operating range, frequency, constraints and economic effects assumed. Technical capability does not mean that every dispatch pattern is commercially attractive or included in the licence and operating plan.
A serious comparison with other resources therefore starts with a common service requirement. If the task is to supply a specified load continuously, the analysis must include the complementary resources or arrangements needed during outages. If the task is annual low-carbon energy, the comparison may be different. Treating unlike services as identical produces misleading cost rankings.
Grid connection is part of that service. A plant can be physically complete while dependent network work remains unfinished. Conversely, a suitable connection does not resolve the plant's own construction or operating risks. The project schedule should reconcile both sides of the interface.
The practical conclusion is that a buyer should not purchase a technology adjective. It should evaluate a defined delivery obligation supported by a credible plant, fuel chain, operating organisation and network arrangement. The more clearly those components are specified, the easier it becomes to compare an SMR proposal with other ways of meeting the same need.
13. What the next evidence should look like
A useful SMR progress report would present a small number of clearly separated records rather than a composite impression of momentum. The licensing record would identify decisions and remaining applications. The construction record would identify accepted work and critical dependencies. The commercial record would separate committed finance and contracts from conditional support.
The technical record would include the status of design completion, supplier qualification and testing. The fuel record would trace the required product through its supply chain. The operating-readiness record would show staffing, procedures, commissioning progress and remaining prerequisites. Each record would have a date and an accountable source.
This proposed structure is an analytical tool, not a claim that a single regulator requires exactly this public dashboard. Its advantage is that it prevents one favourable announcement from filling every column. A new permit belongs in the licensing record. A manufactured component belongs in the supply-chain record. Neither should silently become a revised proof of cost or completion date.
Forecasts should be preserved alongside later outcomes. When a schedule or budget changes, readers should be able to understand the cause and its relevance to future units. This is especially important when the first project is expected to create knowledge for a wider programme.
A strong report also distinguishes uncertainty from non-performance. Some unknowns remain because a test has not yet occurred. Others remain because a test produced ambiguous results. Still others reflect unresolved commercial agreements. These situations require different responses and should not be merged into a generic risk label.
The evidence becomes more valuable when it includes unfavourable findings. A project that explains a problem and demonstrates a credible correction can teach more than an uninterrupted sequence of optimistic announcements. The goal is not to reward bad news; it is to make the record useful for decisions rather than merely supportive of a narrative.
For journalists and readers, the simplest discipline is to attach every major claim to its stage. A reactor can be technically advanced, licensed for a defined activity and commercially unfinished at the same time. Accurate reporting should be able to hold all three facts without forcing them into either celebration or dismissal.
14. Compare the same revision of the project
A less obvious source of confusion is configuration drift. A project may retain the same public name while changing the number of units, planned capacity, construction sequence or scope of common facilities. Comparing an early budget with a later budget without reconciling those changes can misstate both cost growth and technical progress.
A useful comparison begins with a short revision table. It identifies the design version, unit count, capacity basis, scope boundary, price year and target service date for each estimate. Only after those differences are visible should a reader calculate a percentage change. This is a proposed analytical practice, not an assertion about changes at a named project.
For example, imagine an early estimate covering one unit and a later estimate covering that unit plus infrastructure for three more. The larger total does not by itself measure an overrun on the original scope. Conversely, removing work from a revised estimate can make a budget appear stable while leaving the project with additional costs elsewhere. Both directions require reconciliation.
The same discipline applies to schedules. A date for first electricity should not be compared with a later date for full commercial acceptance as though they were identical milestones. A date conditional on a financing decision is not equivalent to an unconditional construction commitment.
Preserving these definitions makes scrutiny fairer. It prevents proponents from moving the target unnoticed, and it prevents critics from attributing every change to execution failure. A project record becomes useful when readers can follow the same object through time rather than comparing different versions hidden beneath one familiar name.
15. The milestone after the milestone
The September 2026 Clinch River permit and Darlington's documented licensing progress deserve to be treated as real developments. They move specific projects forward. They also make the next questions more concrete: how construction proceeds, how interfaces are resolved, how fuel arrives, how commissioning performs and how the final costs compare with the forecasts.
A working project will eventually provide a different kind of evidence from a design review. It will show actual construction productivity, installed-system performance, operating availability and maintenance requirements. A series of projects can then begin to test the broader claim that repetition improves economics and predictability.
Until that evidence exists, it is reasonable to use scenarios. It is not reasonable to present the most favourable scenario as an accomplished result. A first unit can justify investment because of its expected service and learning value while still carrying substantial uncertainty. The honest case for it should say so.
The same standard protects against unfair pessimism. A construction challenge at one project does not prove that every design or delivery model will fail. A fuel issue affecting one class of reactor does not automatically apply to all SMRs. A cost estimate from one jurisdiction cannot be transplanted elsewhere without adjusting its scope and assumptions.
The public interest lies in making these distinctions legible. Electricity systems need dependable infrastructure, and decisions made today can commit money, land, institutions and customers for a long time. The quality of the evidence should match the scale of that commitment.
Small modular reactors are best understood as a set of engineering and delivery propositions now being tested through specific projects. The regulator's decision is one essential part of that test. The factory, construction site, fuel chain, control room and electricity meter will answer the rest.
The most convincing announcement will eventually be less about what the design could do and more about what the completed plant has done, over a stated period, at a stated cost and under clearly described conditions. That is the milestone toward which every earlier milestone should be pointing.