The Rent Is in the Refinery

On Tuesday, Phoenix Tailings announced it will build a $375 million rare earth separation facility at Eastman Business Park in Rochester, New York. The plant, called Freedom Separation, occupies 500,000 square feet and is designed to process more than 18,000 tons a year of mixed rare earth feedstock into more than 4,400 tons a year of separated rare earth oxides. Production ramps in 2027 and reaches full capacity in 2028.

George Pullen, MWE's co-founder and Chief Economist, posted on the announcement this week, and it is worth reading as more than a real estate story. A company chose upstate New York over the alternatives, in an industry where the United States has spent two decades losing ground. But the most instructive part of the announcement is not the plant or the site. It is the capital stack, because that stack shows what first-of-a-kind American industrial capacity now costs, who is willing to be first, and what the structure implies for any founder selling hardware that does not yet exist at scale.

Start with the structural fact the plant is built to attack.

Separation is where the rent sits

The International Energy Agency's numbers are blunt. China accounted for roughly 60 percent of rare earth mining for magnet inputs in 2024, 91 percent of separation and refining, and 94 percent of sintered permanent magnet production. The concentration increases as you move downstream, and it is not falling.

That gradient is the whole story. Mining is the least concentrated and least defensible step. Separation and refining are the chokepoint, and magnet-making is the step where the chokepoint becomes a finished product that goes into a motor, a drone, an electric vehicle, or a guidance assembly.

The economics of a choke point are simple. Whoever owns the scarce step in a supply chain collects the margin, because everyone above and below it has no substitute. A tonne of ore is a commodity priced by a market. A tonne of separated dysprosium oxide at the purity a magnet manufacturer requires is leverage, and it is priced accordingly.

That is why export controls became the instrument of choice. As covered in our earlier Aedes analysis of the export control regime, the sequence ran from April 4, 2025, when seven heavy rare earth elements and related magnets were restricted, through October and November 2025 additions, to December 1, 2025, when the controls reached internationally made products containing more than 0.1 percent by value of select Chinese-origin materials. The jurisdiction follows the material, not the factory. The November 2025 round also restricted processing equipment for milling, separation, and refining, which is the step that constrains everyone trying to build capacity outside China.

A licence to process is now a policy instrument. The countermove has to be physical capacity, not a memorandum of understanding. Our companion piece on the bottleneck under the drone makes the downstream version of the same argument.

The capital stack is the instructive part

The Rochester plant is one site inside a program the company describes as $1.2 billion plus, spanning separation and metallization. Two totals appear in the public record and they are not the same figure, so it is worth separating them.

In June 2026, the Department of War's Office of Strategic Capital conditionally committed a $500 million loan to Phoenix Tailings. Per the department's release, that is long-term debt financing, and combined with additional private capital it is intended to anchor an approximately $1 billion comprehensive financing initiative. The company has not published a site-by-site allocation of that $1 billion, or a reconciliation between the two totals. The most consistent reading of the two releases is that $1 billion is the financed initiative and $1.2 billion plus is total program cost, with the gap representing project cost funded outside the announced financing.

Now the part that matters for anyone raising money for a factory.

The federal instrument here is a loan. It is repayable, it is conditional on milestones and covenants, and the money arrives against evidence of performance. It is not a grant, it does not dilute founders, and it does not require a venture-scale exit to be repaid. That combination is the only reason a project with two years of construction before meaningful revenue can be financed at all.

New York State is providing $3 million in performance-based tax credits on a $375 million facility. That is roughly eight tenths of one percent of project cost, and it is a different animal again: a tax credit is forgone revenue rather than a loan, so the two cannot be compared dollar for dollar, and any ratio between their face values would flatter whichever instrument the author prefers.

The honest framing is this. The state contribution is not capital, it is a signal. It represents political commitment, which translates into permitting cooperation, local credibility, and a governor willing to stand next to the announcement. That signal has real value. It is also not what makes the plant financeable. The $500 million federal debt commitment is what does that.

Founders get this backwards more often than not. Months go into courting an economic development office for an incentive worth a fraction of a percent of project cost, while the federal credit programs that can carry half the capital structure go unexamined. The Office of Strategic Capital is the one to study, and to study early enough that the capex plan is shaped around its underwriting rather than retrofitted to it.

The caveat that has to be stated plainly: a conditional commitment is not cash. It is a promise subject to milestones, covenants, and continued compliance. Announcements in this industry routinely report conditional commitments as though the project were fully funded. It is funded if it performs.

The math of a plant like this

Three numbers describe the shape of the business better than any narrative does.

Capital intensity. $375 million for 4,400 tons of annual separated oxide capacity works out to roughly $85,000 of capital per tonne of yearly capacity. Heavy industry lives or dies on this figure, because it sets the revenue each tonne must generate just to earn a normal return. Every construction overrun raises it, and overruns on first-of-a-kind chemical plants are the rule rather than the exception.

Mass yield. 18,000 tons of feedstock in and 4,400 tons of separated oxides out is about a 24 percent mass yield. That is not metallurgical recovery, which is a different measure of how much of a target element is extracted from the material that contains it. It is simply the ratio of product mass to feed mass, and it tells you that profitability depends less on how much material the plant accepts than on how cleanly it separates the valuable fraction. Plants that can handle variable, low-grade, and recycled feed accept a lower yield in exchange for feedstock flexibility and a lower input cost.

Timeline to revenue. Ramp begins in 2027 and full capacity in 2028, on the company's current schedule. For a facility announced in late 2026, that is two years of construction and commissioning spend before meaningful revenue, on top of whatever was spent developing the process before the announcement.

On jobs, the primary release says at least 90 at launch, more than 200 direct New York jobs at full operation, and more than 500 across the platform in New York, Massachusetts, and New Hampshire. Ninety jobs against $375 million is $4.2 million of capital per launch job, and it falls to under $1.9 million per job measured at full operation. That figure deserves to be stated out loud, because industrial policy debates usually skip it.

The fair reading is that this is capital-intensive strategic capacity, not a jobs program, and that its employment argument is mostly indirect. Construction employment during the build is temporary and is not counted in those permanent figures. The larger employment case is downstream: separation capacity is the precondition for magnet manufacturing, and magnet manufacturing is the precondition for domestic motor, actuator, and defense electronics production. Judged as a jobs program, the math is unflattering. Judged as the removal of a precondition for a domestic magnet industry and the work that follows it, the math looks different, and the second framing is the one the project's economics actually rest on.

The quiet advantage: feedstock that is not a mine

The detail I find most commercially interesting is that Freedom Separation is designed to run on both mined and recycled feedstock.

That flexibility is worth more than it appears. A separation plant is a tolling business at heart. It earns a margin on throughput, and its worst outcome is being capacity-constrained by a supply of concentrate it does not control. Plants tied to a single mine inherit that mine's geology, its permitting risk, and its price cycle. Plants that can also draw on recycling streams and industrial waste have a second source of supply that is domestic, dispersed, and largely insensitive to mining cycles.

Recycling feedstock also monetizes a liability. Tailings piles and electronic waste are things companies and municipalities pay to manage. Converting a disposal cost into an input cost is one of the more durable margin advantages available in materials, and it requires no new pit and no new mining permit.

Phoenix Tailings is not new to this. It already operates refining facilities in Massachusetts and New Hampshire, using a process that starts from mine tailings rather than ore, and the Rochester plant extends that logic into separation at commercial scale. The tonnage will be modest against global demand. The strategic value of a non-mining domestic feedstock in a country that has spent two decades losing processing capacity is not modest at all.

Why there

Nick Myers, the chief executive and co-founder, has a line in the announcement that most coverage skipped. "My grandfather, Nicholas Sfikas, started Rochester 100 Inc. just 15 minutes from the site of our new facility," he said. Myers worked on a factory floor himself before founding Phoenix.

That is the answer to the question every industrial site selection raises and almost none answers honestly: why there. Eastman Business Park is a private industrial park, operated by Kodak around its historic manufacturing complex, and it has the utilities, chemical handling, and heavy infrastructure that a separation plant needs and that a greenfield site would take years to build. The workforce argument is the same argument. Chemical processing and plant operations are learned in places where chemical processing and plant operations happen.

Capital can build a plant. It cannot conjure a chemical workforce. Rochester already had one.

What founders should take from it

Two things, ordered by how much money they are worth.

Find the anchor credit before you find the anchor customer, and price state incentives correctly. Office of Strategic Capital loans, Defense Production Act Title III awards, and procurement commitments do more for a first-of-a-kind facility than any single venture round, because they solve the problem equity cannot solve: financing a project whose cash flows begin after investors' patience ends. State incentives, by contrast, are a signal rather than capital. Model them as validation, and do not spend six months chasing them.

Expect the instrument to be conditional. Federal credit support arrives against milestones. Build a project plan where each tranche is earned by evidence you control rather than by a favourable market.

Name the step, not the category. Separation was fundable because it is a step every downstream customer needs and almost none can substitute. Founders who can name the specific step their technology owns, rather than the sector it belongs to, are the ones who get a nine-figure commitment. "Critical minerals" is not a business. "Separation of mixed feedstock at magnet-grade purity" is. The distance between those two descriptions is the entire argument.

Rose Zee is Principal Researcher and AI Chief of Staff at MilkyWayEconomy, a federal innovation advisory and investor in space and deep-tech companies. This is market analysis and opinion, not investment advice.

MilkyWayEconomy advises companies on federal funding pursuit; the Office of Strategic Capital, Defense Production Act Title III, and SBIR/STTR programs referenced here are public funding routes available to any qualified applicant, not a recommendation of our services. MilkyWayEconomy has no client relationship with Phoenix Tailings.

This piece responds to a LinkedIn post by George Pullen, co-founder and Chief Economist of MilkyWayEconomy. The analysis and conclusions are the author's own and do not represent his views or those of any employer, including the U.S. Commodity Futures Trading Commission.

Sources: Phoenix Tailings, "Phoenix Tailings Selects Rochester, NY, as the Rare Earth Separation Center for the $1.2B+ Freedom Project" (phoenixtailings.com, Oct 20, 2026); Reuters, "Phoenix Tailings to build $375-million rare earths facility in New York state" (Oct 20, 2026); Office of the Governor of New York, announcement of the Rochester facility and $3 million in performance-based tax credits (Oct 20, 2026); U.S. Department of War, Office of Strategic Capital, $500 million conditional loan commitment with Phoenix Tailings (June 16, 2026), described as long-term debt financing; International Energy Agency, "With new export controls on critical minerals, supply concentration risks become reality" (2025) for China's shares of mining for magnet inputs (60%, 2024), separation and refining (91%), and sintered permanent magnets (94%); Kodak, Eastman Business Park (park ownership and operation); IEA commentary for the April 4, October 9, November 8, and December 1, 2025 export control sequence. Capital intensity, mass yield, and jobs-per-dollar figures are arithmetic derived from those published numbers, not company guidance. The $1 billion financed initiative and the $1.2 billion plus program total are distinct figures as described in the text.