SITELIVE 
What happened this year
On 11 February 2026 Albemarle announced it would idle Train 1 at Kemerton in Western Australia and place it into care and maintenance immediately, following its 2024 decisions to mothball Train 2 and cease the Trains 3 and 4 expansion. A plant designed as four trains producing roughly 100,000 tonnes a year of lithium hydroxide, fed by spodumene from Greenbushes — one of the best hard-rock lithium deposits on earth — is now producing nothing; the ABC reported about 275 jobs affected. Albemarle's own language was candid: recent lithium price improvements “are not enough to offset the challenges facing Western hard-rock lithium conversion operations”. Down the road at Kwinana, the Tianqi–IGO refinery produced 897 tonnes of lithium hydroxide in the June 2026 quarter — 15% of nameplate — at a conversion cost of A$40,670 per tonne, on an asset IGO had already fully impaired at 30 June 2025.
Refining is a chemical plant problem, and feedstock is an input you don't control
A conversion refinery is not a bigger mill. It is a continuous chemical plant: kilns and calciners holding tight temperature profiles, off-gas handling, acid and reagent circuits, crystallisation, water treatment, and impurity control measured in parts per million, because the customer is a cathode maker with a specification rather than a smelter with a discount table. Kwinana's June quarter was consumed by a major planned shutdown to install two capital projects, with a further shutdown across July and August to commission a calciner off-gas treatment system that will suppress September quarter output too. IGO guides FY27 production of 9,000–11,000 tonnes at a conversion cost of A$16,000–18,000 per tonne — a plant still buying its way up the ramp curve years after first product, which is the normal and expensive shape of first-of-a-kind process commissioning.
Lynas shows the same physics from the rare-earths side, and shows how far upstream the cause can sit. It produced 1,857 tonnes of NdPr in the June 2026 quarter; the shortfall traced to ore variation at Mt Weld and the replacement of a movable crusher with a fixed one, which reduced granular discrimination of ore feeding the mill, degraded concentrate quality, and in turn cut cracking-and-leaching productivity at both Kalgoorlie and Kuantan. A crusher swap at the pit propagated through two plants on two continents. Kalgoorlie's earlier constraint was utilities rather than chemistry — significant power disruptions in November 2025 forced remediation by the network provider, and Lynas continues to develop an off-grid solution for energy stability. Its Malaysian heavy rare earths expansion has meanwhile risen from about A$180 million to A$294 million, partly because sourcing equipment outside China costs more.
The economics, honestly
Demand is not the problem. Lynas recorded a record average selling price of A$98.2/kg across all rare earth products in the June quarter on A$288.9 million of revenue, its highest since 2022, and the samarium oxide it first produced in March 2026 is already in customer qualification. The problem is that Western conversion capacity competes with incumbents whose plants are debottlenecked, whose reagents and power are cheaper, and whose learning curves were paid for a decade ago. Policy now aims squarely at that gap: the Critical Minerals Production Tax Incentive offers a 10% refundable offset on eligible Australian processing expenditure from 1 July 2027 to 2040, and the Critical Minerals Strategic Reserve — drawing $1 billion from the expanded $5 billion Critical Minerals Facility plus $150 million for selective stockpiling, with offtakes and contracts-for-difference enabled by 2026 legislation — targets price risk rather than capital cost. Neither closes a gap between A$40,670 and A$16,000 a tonne. Only the ramp does that.
What is still being built, and what it demands of operators
The counterweight to Kemerton is 300km north. Iluka's Eneabba refinery — Australia's first fully integrated plant for separated neodymium, praseodymium, dysprosium and terbium oxides — had $1,101 million of capital spent at 30 June 2026 against an unchanged $1.7–1.8 billion estimate, with engineering complete and construction nearing 60%: concrete contractors demobilising, mechanical equipment, pipe-rack modules and tanks going in, the roaster kiln delivered, Civmec holding the structural, mechanical, piping, electrical and instrumentation package, and Export Finance Australia confirming access to the full A$1.65 billion non-recourse government loan. Commissioning is scheduled for 2027 — which, on this year's evidence, is when the real project starts. Three lessons sit in these quarterlies for anyone delivering process plant: write shutdown windows, tie-in sequences and performance-test criteria into the contract with the same rigour as the build programme; treat a change at the crushing circuit as a formal change to every downstream plant's operating envelope, with test data attached; and give utilities and site power quality the same diligence as the process islands, because Kalgoorlie lost throughput to the grid before it lost any to chemistry.