QuantumScape Builds for AI Data Centers While the Stock Burns

A 100 MWh line and a new in-rack battery product arrive as QS sits 45% lower and the AI infrastructure trade cools.

QS · Earnings Call · 2026-10-21

The announcement versus the tape

QuantumScape's third-quarter call is a study in divergence. Management announced a 100 megawatt hour mass production line with Mühlbauer, launched the QS power block for AI data centers, and recorded first customer billings from both Honda and its QSAS advanced-applications unit. The stock, meanwhile, is down 45% over the last 90 trading days and 50.5% from its June peak, at a $2.46 billion market cap and roughly 97% below its 2020 high. The market is not disputing the news; it is refusing to pay for 2028 revenue today. That tension is the story. QS has spent years as a pre-revenue solid-state battery story whose commercial proof was always “next year.” This quarter, for the first time in this sequence, management put a start date, a capacity number and an equipment-cost ceiling on a line beyond the Eagle pilot. “We are targeting equipment cost to QS of less than a $100 million.” — Srinivasan Sivaram · 2026-10-21 It also said the line will be funded with nondilutive sources, including customers, government and ecosystem partners. That is a more concrete manufacturing plan than investors have had. It is still not revenue.

From pilot line to a real line — with a partner

The Eagle Line, inaugurated in February, has become the workhorse for customer sampling and is now generating billings. The next step is a 100 MWh mass production line using cell-assembly equipment from Mühlbauer, with start of production targeted for 2028. Mühlbauer joins Murata and Corning in what management calls an end-to-end ecosystem. The strategic nuance is that QS is not abandoning its capital-light licensing model for automotive. It is adding a direct manufacturing capability for markets that do not have their own cell capacity. “This line lets us serve high value markets directly while keeping our capital light licensing model for automotive.” — Srinivasan Sivaram · 2026-10-21 That is a hybrid model, not a pure licensing model. It matters because the new line is aimed primarily at aerospace, defense and data centers, not cars. Management explicitly said “QSDC and QSAS as maybe a faster time to market than the more rigorous QSEV application” — Srinivasan Sivaram · 2026-10-21. The company is therefore shifting its nearest-term commercialization proof away from the automotive program that defined its first six years as a public company. The start of production date is 2028, which is close enough to plan for but far enough that the market can still discount it.

The AI data-center pitch is differentiated — and early

The QS Power Block is an in-rack energy storage reference design for AI data centers. Management claims 4x the power density and 5x longer run time than the Open Compute Project reference specification, plus thermal stability well above 300 degrees Celsius versus roughly 180 degrees for conventional lithium-ion, including LFP. That safety angle is the real pitch to data center architects: they are packing high-value GPUs into racks and want to avoid thermal runaway risk. Siva’s analogy was unusually good:

In the past, batteries were basically like a spare tire in a car. You carry it along, but you hope you do not need it. As we enter the 800-volt DC transition, and as single rack power is headed towards 1 megawatt, the battery is becoming more like a suspension system.
Srinivasan Sivaram · 2026-10-21

The problem is the tape. The global AI data-center basket is down 5.6% over both the 90-day and 30-day windows, with 48 decliners, and data center capacity is down about 8% over 90 days. QS is launching a data-center product into a cooling trade. The company’s safety differentiation may carve out a niche, but the market is currently treating “AI data center” as a source of downside beta, not a magic ticket. This is not a new idea for QS either; on the April call management already said “The data center market is going into the 800-volt architecture.” — Siva Sivaram, CEO · 2026-04-22 Other reporters this week — from ABB to KeyCorp to OVH — are also talking about data centers, which confirms the theme is shared rather than uniquely QS’s. The energy storage angle inside the rack is more company-specific.

Automotive is still there, but the terms keep moving

The automotive story has not disappeared. QS recorded first billings from Honda, signed a new agreement with a leading global mobility manufacturer, and says it is engaged with two other top-10 OEMs beyond Honda and Volkswagen PowerCo. But the Volkswagen PowerCo relationship keeps being re-scoped rather than simply expanded. On the prior call, management cut the total possible payments under the agreement from about $131 million to about $75 million, eliminated MotoE milestones and said the net cash impact would be neutral. “The total possible payments under the agreement have reduced from approximately $131 million to approximately $75 million.” — Kevin Hettrich, Chief Financial Officer · 2026-07-22 This quarter, the update was another amendment: eliminate transition payments and add milestones tied to technology advancements. The $130 million royalty prepayment is unchanged, and the 2029 automotive production target is unchanged. “That is correct. We have not announced any change from our original plans.” — Siva Sivaram, Chief Technology Officer · 2026-07-22 But the pattern is clear: the automotive relationship is being managed for technology milestones, not near-term capacity payments. That makes the new 100 MWh line strategically important. It gives QS a commercial route that does not depend on Volkswagen’s timing or Honda’s product cycle. It also explains why the keyword mix has shifted so sharply toward QS power block, megawatt hour and data center architects. The automotive keywords are still present, but they are no longer the only story.

The balance sheet buys time; the line is a claim on it

Q3 customer billings were $9.8 million, down from $10.8 million in Q2, which management attributes to the lumpy nature of development work. GAAP net loss was $84.5 million and adjusted EBITDA loss was $60.7 million. Full-year adjusted EBITDA loss guidance is still $250 million to $275 million, and capex guidance is $27 million to $37 million. The company ended the quarter with $819.8 million of liquidity. On the latest filed numbers, that is about 15.1 quarters of cash runway, though the new line’s sub-$100 million equipment bill — even if partly customer- or government-funded — would consume a meaningful slice if QS has to carry it alone. The good news is that QS has been tightening the expense base. Operating loss was $106 million in the latest filed quarter, and Q3 GAAP opex came in at $92.4 million. R&D was $83 million in the latest filed quarter, down 18% year over year. The bad news is that the company is still pre-revenue in any GAAP sense, and the market is focused on the gap between customer billings and a real product ramp.

What to watch

The call did not resolve the central question: can QS convert technical validation into durable customer commitments before 2028? Management says it wants multiple customers signed up before start of production, and the Q4 call should bring a capital plan and site selection. Those are the next hard data points. The stock’s 45% drawdown says investors are not yet underwriting the plan; the new line says management is done waiting for the automotive market to set the pace. That is a real strategic shift, even if the revenue is still years away.