Six US Battery Projects Measured Against Irreversible Milestones

Atlas Public Policy's Q2 2026 clean-energy manufacturing report recorded zero battery investment cancellation value for the quarter, covering cell, pack, component, and recycling investment but excluding critical minerals, from a snapshot pulled 2026-08-17. The figure is correct as measured. It also tells you nothing about whether the pipeline it tracks is moving.

A project that stays in a registry without crossing an irreversible milestone occupies the same line as one that has crossed several. Equipment commissioned, qualification sample delivered, commercial output demonstrated: these are gates in the sense used below, distinct from announcements, selections, groundbreakings, and conditional approvals. The distinction is what matters to anyone sourcing US-manufactured cells or materials against a delivery date.

Six current projects, assessed against that standard. Lansing Grand River, where LGES declared production on 2026-08-18 (confirmed by Argus), serves as a seventh reference point and the clearest recent case of a project crossing from construction into production. Even there, output volume, yield, and buyer-accessible allocation remain undisclosed.

Warren, Ohio — Ultium Cells restart

Pre-shutdown production is established. Ultium says the plant produced its 100-millionth cell by December 2024, with initial shipments in November 2022.

Reuters reported on 2026-08-12 that Ultium expected most temporarily laid-off employees back the week of August 17. Local reporting on 2026-08-17 confirmed the final group of hourly workers had returned, with UAW Local 1112's president saying production would ramp "soon" after training. By 2026-09-03, approximately 700 Local 1112 members had returned and the plant had begun ramping cell production, with employees still completing safety and job training before production could "fully resume."

That is more than a scheduled restart and less than demonstrated output. The public record establishes workforce recall and a reported ramp; it does not establish a dated post-shutdown production figure, a utilization rate, or a confirmed customer shipment. I classified Warren as "ramping (output unverified)" in the Issue #11 registry update, and the September 3 reporting does not move it.

Next gate: A dated post-shutdown production figure, recurring shipments, or disclosed utilization.

New Carlisle, Indiana — Samsung SDI (formerly Samsung-GM JV)

The exterior shell is erected. Substantial public infrastructure obligations were committed under a 2023 county development agreement setting a 2027-12-31 substantial-completion deadline and a $4.5 million annual infrastructure-payment stream.

Samsung SDI announced on 2026-08-11 that it had acquired GM's 49.99% stake in the joint venture and plans initially to use the plant for stationary storage batteries rather than EV cells. Reuters described the acquisition prospectively and reported that Samsung's revised investment plan had not been finalized. No revised capacity, chemistry, customer, start-of-production date, equipment plan, or investment figure has been disclosed since.

The IEDC Transparency Portal still lists the project under General Motors (project ID 423232), showing $3.497 billion of expected investment and 1,620 contract-end jobs, with $135 million in conditional grant support and $32 million in state payroll tax credits labeled "pending accepted offer." The St. Joseph County Redevelopment Commission's 2026-08-11 meeting was canceled, and searches of county records through the cutoff located no resolution extending the substantial-completion deadline or consenting to an assignment reflecting Samsung's sole ownership and storage pivot. The commission's next meeting was scheduled for 2026-09-08. This is a bounded public-record finding; it does not exclude a confidential or not-yet-posted instrument.

New Carlisle is the clearest case in this set of a failure mode the Issue #12 cohort study found to be the most severe and least tracked in the US pipeline: a facility that changes end market or chemistry mid-build. The site continues; the EV supply promise announced in 2023 no longer exists, and no tracker records that as a cancellation because the shell is still standing. Whether storage capacity replaces it depends on evidence that has not appeared publicly: an equipment order, a chemistry and capacity disclosure, a permit amendment, resumed construction.

Next gate: Any of the above, or a publicly documented incentive amendment reflecting revised terms.

NOVONIX Riverside, Tennessee — anode material

Panasonic-required mass-production equipment was installed and commissioned, and a mass-production qualification sample was delivered (NOVONIX SEC filing, 2026-06-11). Most announced materials projects never get this far.

The company's 2026-08-31 half-year filing then moved the expected date for full 20,000-tonne/year capacity from December 2027 to September 2031, citing capital conservation. Initial Panasonic mass production remains targeted for H2 2027. Whether that initial target is insulated from the same capital logic that pushed the full buildout out by four years depends on the pending qualification. If Panasonic's acceptance arrives on schedule, the initial ramp has a customer-driven rationale the full buildout currently lacks; if acceptance slips, NOVONIX's cash position gives it little reason to force the pace. Panasonic's formal assessment of the sample was pending as of the June filing, and commercial production depends on successful qualification by Panasonic and its downstream customers.

The interval between delivering a qualification sample and receiving formal acceptance is where anode programs routinely stall. Customer testing cycles, downstream OEM approval chains, and iterative spec adjustments can consume 12 to 24 months. The four-year push on full capacity indicates NOVONIX is staging the ramp against cash rather than demand, which is rational and extends volume availability accordingly.

Next gate: Panasonic's formal sample acceptance, followed by recurring battery-grade shipments.

Celgard Charlotte, North Carolina — separator coating

Trial operation of a new Hipore wet-process coating line began in March 2026 (Asahi Kasei management briefing, April 2026), meaning the line is running product under test conditions rather than at commercial rate. Physical line completion and an opening ceremony followed in August. Asahi Kasei says commercial production is scheduled for H2 FY2026, ending March 2027.

The expansion belongs to a three-country coating program across Charlotte, Hyuga, and Pyeongtaek with approximately JPY 40 billion of combined investment and approximately 700 million m²/year of combined added capacity. No Charlotte-specific allocation of either figure has been disclosed. Charlotte Business Journal reported the expansion added roughly 50% to the plant's manufacturing footprint and more than 100 jobs. The April briefing identified an agreement giving Toyota Tsusho's US subsidiary contracted rights to output from the new capacity.

Celgard is the most advanced project here: installed equipment already running product, a named customer holding rights to that output, and a commercial-start window inside two quarters.

Next gate: Declared commercial-production start and customer-accepted shipments.

DOE $500 million battery material processing selections

Seven projects were selected for award negotiations on 2026-08-20: Waterleaf/Lilac Solutions, Jervois, Nth Cycle, Princeton NuEnergy, Arcanum Ventures, Elevated Materials, and Coreshell Technologies.

The governing funding notice states that selection does not commit DOE to issue an award, does not guarantee funding, and does not authorize project performance before a grants officer executes an assistance agreement — the document that actually obligates federal money. Recipient communications through the cutoff, from Lilac, Nth Cycle, Princeton NuEnergy, and Coreshell, are consistent with the negotiation stage. No executed assistance agreement, obligated funds, or grants-officer action was located for any of the seven through 2026-09-04. Three had not disclosed specific facility locations.

These are pre-gate. I count the slate as one observation rather than seven, since none has crossed any irreversible threshold.

Next gate: An executed assistance agreement for any of the seven.

Sila Technologies — conditional federal loan commitment

The Office of Strategic Capital, within the Department of War, announced a conditional commitment of up to $1.4 billion on 2026-08-07. Financial close remains subject to financial, legal, technical, and diligence conditions per the official release. Site, capacity, closing date, collateral, equity contribution, construction schedule, and disbursement schedule are all undisclosed. Federal willingness to lend has been stated; the facility it would fund has no public location, size, or cost.

Next gate: Definitive documentation and financial close, followed by site disclosure and construction commencement.

What the accumulated evidence shows

Of the six: Celgard is in trial operation with a near-term commercial target. NOVONIX has delivered a qualification sample and awaits customer acceptance. Warren has recalled its workforce and reported a ramp without disclosing post-restart output. The remaining three — New Carlisle, the DOE selections, Sila — have crossed no gate involving irreversible capital, physical completion, or demonstrated output under their current plans. New Carlisle retains a shell with no equipment, chemistry, capacity, or customer attached to the storage conversion; the DOE selections remain in negotiation; Sila has no disclosed site.

Add Lansing and the count is two facilities with declared production, one in trial operation, one at qualification-sample stage, three at or before announcement. Neither of the two producing facilities has disclosed output volume, yield, or allocation available to unaffiliated buyers, a gap the Issue #9 addressable supply inventory found to be persistent across every producing US cell maker.

What the advancing projects have in common is a single corporate sponsor with an existing manufacturing operation and a named customer. Celgard has Asahi Kasei behind it and Toyota Tsusho in front. NOVONIX has Panasonic. Lansing is wholly owned by LGES. The stalled and pre-gate projects are a JV unwinding (New Carlisle), a startup with conditional financing (Sila), and government selections not yet attached to executed commercial relationships. The Issue #12 cohort study found the same sorting across twelve facilities, with sponsor type predicting outcomes better than geography, chemistry, or policy support.

For a procurement lead building a 2027 or 2028 plan around US-manufactured material, the projects closest to commercial supply in this set are separator coating and anode material, not cells. Cell output is either undisclosed in volume (Lansing, Warren) or years from a first unit (New Carlisle, Sila). Nothing here supports a planning assumption that new US cell capacity becomes contractable within the next four to six quarters. What would change that read is narrow and specific: a disclosed utilization or allocation figure from Lansing or Warren, or an equipment order at New Carlisle. Absent those, a zero-cancellation quarter records only that the announcements are still on the books.