$2,050 per kilogram for terbium. $575 for dysprosium. $110 for neodymium and praseodymium. Those are not prices anyone paid. They are floors — numbers the U.S. government has promised to top up whenever the market clears below them, written into a 15-year contract for Serra Verde's rare earths out of Brazil.
Strip the press-release language off the $1.55 billion Pentagon deal announced August 24 and here is what it actually is: not a purchase, but a guarantee. A take-or-pay structure in which a government-backed vehicle, US SIIE, buys 100% of the neodymium, praseodymium, dysprosium, and terbium in Pela Ema's phase-one output, at a price that cannot fall below a fixed line — one that rises 2% a year, with Serra Verde still keeping 70% of any upside above it.
Washington has stopped trying to compete with China on price. It has replaced the price signal itself.
Washington has stopped trying to compete with China on price. It has replaced the price signal itself.
Last week we traced the germanium reshoring thread through ReElement's privately built processing line in Marion, Indiana. This week the story moves up a layer — to the federal offtake-and-subsidy architecture now underwriting those lines, and to a single question the rest of this piece answers: what does $1.5 billion in price floors actually buy when the metal doesn't ship until 2028?
Start with why the floor is the load-bearing instrument, not the cash.
The MP Materials deal a year earlier already ran the template. A 10-year contract-for-difference pinned NdPr at $110 per kilogram; DoD pays the shortfall below the floor and captures 30% of the upside above it. The mechanism is borrowed straight from how the U.K. finances offshore wind and nuclear — a tool built for exactly one problem.
That problem is the cost curve. NdPr has cycled between roughly $40 and $200 per kilogram, and the troughs run deep enough and long enough to bankrupt any Western producer carrying separation opex. The marginal Chinese tonne sets the price, and Chinese producers are vertically integrated and willing to run thin or negative spreads specifically to keep Western entrants from ever clearing project finance.
A grant doesn't fix this. You hand a project cash, it builds a plant, then China prices it into insolvency during the next trough. A decade of Defense Production Act awards learned this the hard way.
The floor fixes it because it doesn't subsidize the commodity — it removes price risk from the producer. It converts a Chinese-priced commodity into a regulated-utility-style cash flow, the kind a project-finance desk can underwrite without losing sleep. It doesn't require Beijing to behave; it just changes who absorbs the volatility when Beijing doesn't.
That single insight now propagates across four federal channels. The Pentagon runs offtake and equity — the $750 million IBAS investment in US SIIE, plus a Defense Logistics Agency commitment to purchase at least $300 million and a bank revolver of up to $500 million stacked on top. DOE runs grants — another $500 million across seven battery and mineral projects on August 20 alone, in what DOE describes as potential funding. The Office of Strategic Capital runs loans — $725 million to Energy Fuels, $500 million to Phoenix Tailings, and, per a White House fact sheet reported by Supply Chain Dive, $2.03 billion more in agreements signed with DOD. And Commerce runs CHIPS money — up to $1.6 billion, per Fastmarkets, for USA Rare Earth's Round Top and its magnet lines.
Different instruments, one logic: de-risk the price, and private capital shows up. USA Rare Earth layered $1.2 billion of its own money into a South Carolina magnet plant on top of the Commerce package. Phoenix Tailings' $500 million loan pulled in enough private capital to reach $1 billion total.
The mechanism works. That's the part that's genuinely new.
Now line up the delivery dates, and the whole thing turns cold.
Serra Verde's phase-one output reaches only about 4,000 metric tons of rare earth oxide equivalent by the end of 2026, averaging roughly 6,400 tonnes a year over the mine's life. First deliveries land early in the fourth quarter — of this year — but that's mixed carbonate, an intermediate that still has to be separated and metallized before a single magnet exists.
Where does the separation happen? The whole point is to do it outside China. And every one of those facilities lands in 2028.
Energy Fuels' new separation and metallization plant: build-out, not online. Phoenix Tailings' Freedom Facility: explicitly targeting 2028. USA Rare Earth's Round Top: 2028. The four processors the Army is letting build on military bases through long-term leases — REalloys reportedly on dysprosium and terbium at Tooele, among them — target initial operating capability in 2028, with construction maybe starting in 2027.
And the front end of the pipeline is worse. RAPTOR — the Texas A&M drilling and real-time sensing platform meant to compress the discovery-to-reserves runway — is a $3.5 million ARPA-E prototype still headed to a test facility. The runway it's trying to shorten currently runs more than a decade and costs upward of $100 million per prospect.
So here is what $1.5 billion in floors buys: a bankable supply chain whose first magnet-grade metal arrives in 2028. The capital solved the financing problem. It did not, and cannot, solve the clock. No price floor moves a delivery date forward.
Michael Cadenazzi, the Pentagon's industrial base policy chief, put the stakes plainly: "You can dream all day long about scaling [weapons production but] if you don't have germanium, gallium, and rare earths, it is a pipe dream." He's right about the constraint. But the floors he's signing don't relieve it until 2028 — and China still reportedly controls about 90% of separation today, per a US Congressional Research Service article cited by Fastmarkets.
That's not a hypothetical gap. It's already binding, just not on American primes yet.
Look at India's physical-AI startups. Beijing has been denying visas to Indian executives who need to travel to source actuators, sensors, camera modules, and lidar — components that, as one founder put it, are "available only in China." XP Robotics' CEO, whose visa was rejected multiple times in two months, reportedly says a prolonged disruption could have 'a business impact running into a couple of million dollars' and put contractual obligations at risk. These are companies losing business today, not in 2028.
That is the shape of the chokehold the U.S. floors are insuring against — and the reason the insurance matters even though it pays out late. Washington has correctly diagnosed that price risk, not capital scarcity, was what killed Western rare earths for two decades. It built the right instrument to fix it.
But an instrument that de-risks a 2028 plant does nothing for a 2026 shortage. For the next three years, the binding constraint isn't the price signal Washington just nationalized. It's the calendar it can't rewrite.