On September 29, Saskatchewan said it wants two large reactors in the northwest, more than 2,000 megawatts, the first of them online by 2042. It also kept a smaller plan near Estevan. At least 600 megawatts of small reactors, with the Rafferty site chosen for further study. The whole target is at least 2,600 megawatts of nuclear by 2050. The province has no nuclear plant today. Its record demand, set in December 2021, was 3,910 megawatts. This is not a side project. It is a new spine for the grid.
Investors should ignore the year 2042. The only decision that pays anyone is the one SaskPower says it will make in 2026 and 2027. Which machine. That choice is not about a logo. It is about who sells the fuel for the life of the plant. A CANDU burns uranium in the form the ground already offers. An AP1000 cannot. It needs enriched fuel, and Canada does not enrich uranium. Buy the second machine and Saskatchewan imports a bottleneck it does not have. Buy the first and the rock under its own feet stays the product. That is the whole trade.
Nothing here is a recommendation to buy or sell any security. A target for 2042 is not a contract. A refurbishment is not a new build. Names are on this page so the map is readable, not so a reader can skip the filings.
The short list is already in the room
SaskPower has not named the large design. The province's own release still did the introductions. John Gorman of Westinghouse Canada and Ian Edwards of AtkinsRéalis were both in the announcement. Westinghouse sells the AP1000, a reactor of about 1,100 megawatts that is already running in China and, at last, in Georgia. AtkinsRéalis sells CANDU, including a proposed Monark of about 1,000 megawatts that has not been built. Two vendors. Two fuel cycles. One Crown utility that has to pick.
The small end of the plan is not a mystery. Saskatchewan has already pointed at the GE Vernova Hitachi BWRX-300 for Estevan, about 300 megawatts a unit, with a goal of at least 600. That machine uses enriched fuel. So the dependency the large choice can still avoid has already been accepted for the small one. Investors who treat the whole 2,600 megawatts as one bet are mixing two businesses. The Estevan units are a Hitachi-GE story and an enrichment story. The northwest units are the open question, and they are four times the power.
Why the fuel is the asset
Natural uranium comes out of the ground at about 0.7 percent of the fissile isotope, U-235. A CANDU, because it is moderated with heavy water, can run on that. Fuel for it is mined, milled, refined, and made into oxide. No enrichment plant sits in the middle. Most of the rest of the world's large reactors want the uranium in the core enriched to something like 3 to 5 percent. The AP1000 is in that camp. The BWRX-300 is in that camp too.
Canada has no commercial enrichment plant. If Saskatchewan orders machines that need enriched fuel, someone else will own that step for as long as the cores run. The someone else is not a detail. Enrichment is a short list of firms, mostly outside this country, and it is a licensed, political, bottlenecked business. The world is trying to build a fleet of enriched-fuel reactors at the same time. Fuel that used to be a commodity line item becomes a queue. CANDU is the design that lets a uranium province step out of that queue.
Cameco sits on both sides of the table, which is why the ticker is a bad shortcut. Cameco, traded as CCO in Toronto and CCJ in New York, is the major uranium producer with its name on Saskatchewan rock. It also owns 49 percent of Westinghouse. Brookfield owns the other 51 percent. The deal closed in November 2023. A CANDU order feeds Cameco's mine, mill, and conversion business in the most direct way a reactor can. An AP1000 order feeds Cameco's equity in Westinghouse, and the reactor services and fuel fabrication that follow a Westinghouse core, while the enrichment margin goes somewhere else. One company can win either way and still not win the same way. An investor who buys Cameco "for Saskatchewan nuclear" has not asked which dollar.
AtkinsRéalis, traded as ATRL in Toronto, is the other name that changes with the pick. It owns the CANDU technology. A northwest order would be the first real commercial test of that franchise. Can it sell a new large reactor in its own country, or only keep old ones alive? No such order exists tonight. The Monark is a design, not a plant. Treat a hope as a backlog and you will overpay for it.
A Canadian owner is not a Canadian machine
Westinghouse wears a Canadian flag in its sales line because Cameco and Brookfield own it. The AP1000 was not drawn in Mississauga. It is an American design. US nuclear technology remains inside the American export-control system no matter whose pension fund holds the shares. Parts, data, and the right to build can still depend on a signature in Washington. That is a mild constraint with a friendly White House and a hard one when the White House is not friendly. Ownership does not repeal it.
There is a political story about an AP1000 order. It would be a gift Saskatchewan could hand a tariff fight. A US export win, dressed up as a local job. It is possible. It is not a plan, and it is not in the September 29 release. Investors should not pay for a theory about Donald Trump's mood. They should price the thing that lasts longer than a presidency. A reactor ordered in the late 2020s will still need fuel in the 2060s. The enrichment contract outlives the communique.
The build record, without the poster
The American construction record for this machine is ugly, and it should be said with numbers. Vogtle 3 and 4 in Georgia were sold as a project of about $14 billion. They finished above $30 billion and years late. Westinghouse filed for bankruptcy in 2017 in the middle of that build. The sister AP1000 project at VC Summer in South Carolina was cancelled after billions had already been spent. Later units in China were built on better schedules. The design can be finished. The design has also ruined an owner. Both sentences are true. A brochure that says "proven" is quoting the second country and hoping you forget the first.
The Canadian record people are waving is real, and it is narrower than the wave. Ontario finished the refurbishment of four CANDU units at Darlington against a $12.8 billion budget, about $150 million under it. The last unit, Unit 4, was cleared to return about four months early. In June 2026 Ontario said Bruce Power's Unit 3 refurbishment finished seven months early and about $150 million under its projection. These are repairs of machines Canada already knows how to run. They used Canadian shops, Canadian trades, and a regulator that has lived with CANDU for decades. They are the best evidence this country has that a nuclear job can come in close to the promise.
They are not evidence that a new Monark, on a river in northwest Saskatchewan, will do the same. A refurbishment replaces channels, feeders, and steam generators inside a building that already exists. A first plant pours concrete, licenses a site, trains a workforce that has never run a reactor, and builds a supply chain that has been thin. The CANDU new-build record in Canada is old. Work has resumed on CANDU units in Romania. That is not a Saskatchewan commissioning date. Anyone selling the Darlington underrun as a forecast for 2042 is selling a different job.
Hold the two records side by side and the investor point is still plain. One design has just been rebuilt, twice, in Canada, close to budget. The other design's only finished pair in the United States blew the budget and broke the vendor. If the only question were "who has recently done this work on this continent," CANDU wins. Ask instead who has a large reactor already licensed and operating in the West. The AP1000 has the operating units. CANDU has a design waiting for a first order. Saskatchewan is being asked to pick which risk it would rather own.
What to watch, and when
The cash does not arrive in 2042. The repricing arrives when the technology is chosen. The province's own timetable says the near term is 2026 and 2027. That is when large-reactor selection is supposed to be finished, and when planning for the two northwest units is supposed to get a direction. Estevan's site work continues on its own clock. A final site for the small reactors has been described as a this-year decision, with Rafferty now the place under further study. Those are the dates that move stocks. A groundbreaking in the 2030s will be a photograph.
Watch three things and ignore the rest.
First, the fuel sentence in the pick. If the release says CANDU or Monark, the natural-uranium chain is the asset. Cameco's mines and conversion plants sit closer to the cash flow. AtkinsRéalis has a home reference plant to sell to the next buyer. Enrichment stays someone else's problem. If the release says AP1000, Westinghouse's Canadian owners get a flagship order in a uranium province, and the fuel cycle grows a foreign step in the middle. Cameco can still sell uranium into that step. It will not own the step.
Second, the contract, not the target. More than 2,000 megawatts is a goal. A goal is not an engineering, procurement, and construction price. Vogtle was a goal with a price, and the price did not survive. Until there is a number, a schedule, and a party who eats the overrun, the 2042 date is a sentence. Do not capitalize it.
Third, the supply chain that can actually bid. A CANDU win means Canadian heavy industry has to show it can still make the parts. An AP1000 win means a queue. Other hopes, from Poland to a US framework that talks about six more units, have to have room for Saskatchewan. A full order book is not a gift to a late buyer. It is a wait.
The close
Saskatchewan has uranium, a grid that will need new firm power, and a premier who has now put a number on the ambition. It does not yet have a reactor. The two machines in the room do not do the same job for an investor. One keeps the fuel in the form the province already sells and leans on a refurbishment record Canada has just earned. The other imports enrichment, a US design, and a construction history that includes a bankruptcy and a cancelled plant. In exchange it offers a machine that is already running in Georgia and China, and a vendor half-owned by Cameco and Brookfield.
The mistake is to buy the dependency because the owner is local, or to buy the homegrown story because a refurbishment finished early. The decision that matters is dull and final. For the life of the plant, who sells the fuel, and can they be told no. Everything else is a press conference.
A note on sources and limits
The September 29, 2026 plan is from the Government of Saskatchewan and SaskPower. It sets a goal of at least 2,600 megawatts of nuclear by 2050. More than 2,000 megawatts would come from two large reactors in the northwest, the first expected online by 2042, plus at least 600 megawatts in the Estevan area. Rafferty was selected for further SMR study. The large design was not named. Westinghouse and AtkinsRéalis were both attached to the announcement. Saskatchewan had already identified the BWRX-300 as its SMR path. Cameco's 49 percent stake in Westinghouse, with Brookfield at 51 percent, closed in November 2023. Darlington's refurbishment was reported by Ontario in early 2026 as about $150 million under a $12.8 billion budget, with Unit 4 about four months early. Bruce Unit 3 was reported in June 2026 as seven months early and about $150 million under its projection. The rest of the Bruce life-extension program is not finished. Vogtle's original cost is widely cited near $14 billion and its final cost above $30 billion. VC Summer was cancelled. AP1000 units have been completed in China. CANDU fuel is natural uranium, about 0.7 percent U-235. Light-water reactors typically use uranium enriched to a few percent. Canada has no commercial enrichment plant. The Monark is a proposed CANDU, not an operating plant. This is not investment advice, not an offer, and not a forecast that any project will be built on the dates in a strategy paper.