# Is China's Linglong One Proving the SMR Cost Case Wrong?

The cleanest head-to-head test of small modular reactor economics in history is running at Changjiang, Hainan — and after [first criticality](https://smrintel.com/glossary/criticality) at the full-size reactor next door on July 10, the SMR sitting 216 days past its revised schedule has a cost-per-megawatt problem it cannot explain away.

Here is the core finding, grounded entirely in numbers reported by the project's own officials: Linglong One, [China National Nuclear Corporation](https://smrintel.com/companies/cnnc)'s 125 MWe (100 MWe net) SMR, cost 5 billion yuan to build. Changjiang Unit 3, a full-size Hualong One at 1,000 MWe net, cost 20 billion yuan. Divide cost by net capacity and Linglong One lands at 50 million yuan per megawatt. The Hualong One lands at 20 million yuan per megawatt. The SMR costs two and a half times more per megawatt than the large reactor built on the same island, by the same industry, paid for in the same currency.

The Hualong One also reached first criticality — the moment a controlled chain reaction is first sustained — at 18:00 local time on July 10, 2026, after beginning construction on March 31, 2021. Linglong One poured first concrete on July 13, 2021, 104 days later. As of this writing, CNNC has announced no equivalent milestone for the SMR, and the IAEA's reactor database shows the criticality, grid connection, and commercial operation fields for Linglong-1 as blank.

---

## The Race Nobody Planned to Have

Nobody designed Changjiang as a controlled experiment. The co-location happened because Hainan Nuclear Power Co. was already developing the site when CNNC selected it for Linglong One's commercial demonstration. The result is something the global SMR industry has never had: two reactors on a shared fence line, construction starts separated by 104 days, one of which is the machine SMR proponents say will make the other obsolete.

The big machine won.

Changjiang Unit 3's first criticality was announced by CNNC and reported by World Nuclear News on July 13, 2026. Grid connection and a staged power ascension follow, with commercial operation currently targeted for early 2027. The project followed its schedule publicly and without notable controversy.

Linglong One's last confirmed milestone was a non-nuclear turbine test passed on December 23, 2025 — a dress rehearsal for the conventional power-generation side of the plant, run on externally supplied steam before any fuel is loaded. That is the last thing the world knows for certain about the world's first commercial land-based SMR: that, as of late December, the turbine worked. Everything since has been silence from CNNC, which is notable because the company announced the main coolant pump installation, the containment dome, cold functional tests, and the core module's sea delivery by barge.

The design schedule called for 58 months from first concrete, placing commercial completion on May 13, 2026. That date passed nine weeks before this writing. The schedule had previously been revised to 53 months in July 2023, after Chen Jianxin, chief engineer at Hainan Nuclear Power Co., told the Global Times that work was running ahead. Fifty-three months from July 13, 2021 points to December 13, 2025. That date is 216 days behind today. A CNNC official told Reuters on December 11, 2025 that commercial operation would begin in the first half of 2026. The first half of 2026 ended June 30.

---

## Dissecting the Headline Numbers

CNNC's standard Linglong One fact sheet promises a billion kilowatt-hours annually, enough for 526,000 households, and 880,000 tonnes of CO₂ avoided. Each figure deserves scrutiny.

**The output claim.** The IAEA lists Linglong One's gross capacity at 125 MWe and net capacity — what reaches the grid after the plant powers itself — at 100 MWe. Running 125 MWe continuously for a full year yields 1.095 billion kilowatt-hours. Reaching the round figure of one billion kilowatt-hours on gross output requires an annual [capacity factor](https://smrintel.com/glossary/capacity-factor) of approximately 91.3%. China's mature large reactor fleet has achieved numbers in that range. A [first-of-a-kind (FOAK)](https://smrintel.com/glossary/foak) machine in its first operating years historically does not. The IAEA's net figure of 100 MWe running flat-out for a full year yields 876 million kilowatt-hours — meaningfully less than the promotional billion.

**The household count.** Five hundred twenty-six thousand households from one billion kilowatt-hours implies annual per-household consumption of approximately 1,901 kilowatt-hours. The U.S. Energy Information Administration puts the average American household at roughly 10,500 kilowatt-hours per year. Chinese residential consumption is genuinely lower, and the comparison is not a direct criticism of the engineering. It is a criticism of a marketing figure that translates poorly across borders — and much of the capital chasing SMR projects globally is American and European.

**The construction cost.** Chen gave the Global Times the figures used above: 5 billion yuan for Linglong One, roughly 20 billion yuan for each Hualong One. China Huaneng, which holds 51% of Changjiang's second phase, puts units 3 and 4 together at 40 billion yuan, confirming the per-unit figure. The 2.5x cost-per-megawatt premium for the SMR is not an estimate derived from third-party modelling. It is straight arithmetic on numbers the builder's own engineers provided to state media.

---

## The Industry's Standing Answer — and Its Limits

The SMR sector's response to FOAK cost overruns is consistent, and not unreasonable: the first unit of any manufactured product carries tooling costs, supply chain premiums, and learning-curve inefficiencies that disappear by unit ten or twenty. The argument is that you cannot judge a replication strategy on its first replication.

That is the same logic underpinning construction now at Darlington in Ontario, where a 953-tonne basemat went into a shaft in June 2026 for the first of four planned BWRX-300 units from [GE Vernova / GE Hitachi Nuclear Energy](https://smrintel.com/companies/ge-vernova). The four-unit sequence at a single site is specifically designed to capture learning-curve benefits across near-identical builds.

The argument is fair. It also cannot be empirically validated until multiple units exist. Linglong One is unit one globally for commercial land-based SMRs. There is no unit two yet. The supply chain premium Chen's figures capture is real, and it may compress substantially on subsequent builds — but subsequent builds require customers willing to pay FOAK prices while waiting for NOAK economics to arrive. That is the financing gap the entire Western SMR industry is trying to bridge with government loan guarantees, tax credits, and offtake agreements from data center operators.

What Changjiang adds to that picture is uncomfortable specificity. This is not a modelled cost comparison or a consultancy projection. It is a real SMR and a real large reactor, built concurrently by the same national nuclear industry, on the same site, with costs reported by the same company. The large reactor is cheaper per megawatt and reached first criticality first. The industry should engage with that data honestly rather than routing around it.

---

## What This Means for the Broader SMR Market

The Changjiang data point arrives at a moment when SMR developers from [Rolls-Royce SMR Ltd](https://smrintel.com/companies/rolls-royce-smr) to [NuScale Power](https://smrintel.com/companies/nuscale-power) to [Kairos Power](https://smrintel.com/companies/kairos-power) are all pitching some version of the factory-built, faster, cheaper narrative. None of them have a commercial unit operating. Linglong One was supposed to be the proof of concept — the machine that demonstrated the pitch worked at full scale. Its 216-day slip and its 2.5x cost-per-megawatt premium compared with the Hualong One next door will appear in every due diligence deck written against an SMR investment for the next several years.

That does not mean SMRs fail as a category. Different reactor designs serving different markets — remote industrial sites, data center campuses, grid firming in jurisdictions that cannot site large reactors — face different competitive baselines than a utility-scale grid reactor in China. The Hualong One is not a competitor for a 20 MWe microreactor serving a copper mine in Chile or a hyperscale data center in Virginia that cannot wait eight years for a gigawatt-class plant.

But the [baseload](https://smrintel.com/glossary/baseload) grid-replacement pitch — that SMRs can do what large reactors do, only cheaper and faster — took a measurable hit on July 10, 2026, when Changjiang Unit 3 achieved first criticality and Linglong One's fields in the IAEA database remained blank.

---

## Key Takeaways

- Changjiang Unit 3, a 1,000 MWe net Hualong One, reached first criticality on July 10, 2026 — beating Linglong One to the milestone despite starting construction 104 days later
- Linglong One, 125 MWe gross / 100 MWe net, is at minimum 216 days past its revised 53-month schedule from July 2023
- CNNC's own reported figures show Linglong One costs 50 million yuan per net megawatt versus 20 million yuan per net megawatt for the Hualong One — a 2.5x premium
- The last confirmed Linglong One milestone was a non-nuclear turbine test passed December 23, 2025; no criticality announcement has been made
- The FOAK cost premium argument is structurally valid but empirically untestable until multiple units are built — and no second Linglong-class unit is under construction
- Changjiang's co-location provides the most direct real-world SMR vs. large reactor cost and schedule comparison yet assembled anywhere

---

## Frequently Asked Questions

**Why did Changjiang Unit 3 reach first criticality before Linglong One if it started construction later?**
Construction on Changjiang Unit 3, a Hualong One, began March 31, 2021 — 104 days before Linglong One's July 13, 2021 first concrete. Despite the shorter head start, the Hualong One leveraged a mature supply chain and established construction methodology. Linglong One, as a first-of-a-kind design, has faced delays whose specific causes CNNC has not publicly disclosed. The Hualong One achieved first criticality July 10, 2026; as of July 20, 2026, Linglong One has not announced an equivalent milestone.

**How much does Linglong One cost compared to a Hualong One on a per-megawatt basis?**
According to figures provided by Chen Jianxin, chief engineer at Hainan Nuclear Power Co., to the Global Times, Linglong One cost 5 billion yuan for 100 MWe net capacity, equalling 50 million yuan per megawatt. Each Hualong One unit at Changjiang cost approximately 20 billion yuan for 1,000 MWe net, equalling 20 million yuan per megawatt. That is a 2.5x cost-per-megawatt premium for the SMR over the large reactor built on the same site.

**Does the Linglong One delay and cost premium mean SMRs are not viable?**
Not necessarily. SMR developers and their investors argue that FOAK units carry supply-chain and engineering premiums that compress significantly across a series of identical builds. The Linglong One comparison is specific to a single unit of a Chinese pressurized-water SMR competing against one of the world's most mature large reactor designs. SMRs targeting different market segments — remote industrial heat, data center power, or sites where large reactors cannot be permitted — face different economic baselines. The Changjiang data does, however, challenge the specific claim that SMRs can deliver cheaper per-megawatt grid power than large reactors in a mature nuclear market.

**What is the current status of Linglong One as of July 2026?**
The IAEA's reactor database lists Linglong One as under construction, with first criticality, first grid connection, and commercial operation date fields all blank. CNNC's last announced milestone was a successful non-nuclear turbine test on December 23, 2025. The project is past both its original 58-month schedule (May 2026 completion) and its revised 53-month schedule (December 2025). A CNNC official told Reuters in December 2025 that commercial operation would begin in the first half of 2026; that window closed June 30.

**What does this mean for SMR projects in other countries?**
The Changjiang comparison will inform due diligence across every SMR program globally. Projects like the BWRX-300 at Darlington are designed with multi-unit sequences specifically to capture learning-curve cost reductions — an implicit acknowledgment that unit-one economics are not the selling point. For investors, offtakers, and regulators, Linglong One's schedule and cost data provide the first empirical anchor for what FOAK SMR performance actually looks like at commercial scale, as opposed to modelled projections.