QuantumScape Brings Solid-State Lithium-Metal Batteries to AI Data Centers With QS PowerBlock

Introduction

The AI buildout is beginning to reshape the battery industry, and the change runs in an unexpected direction: the consumers of next-generation cells are no longer only automakers, but the operators of AI data centers. On October 8, 2026, QuantumScape Corporation announced the QS PowerBlock, a modular solid-state lithium-metal battery system designed for in-rack energy storage in AI data centers. The company claims the design delivers four times the power density and five times the runtime of current data center industry standards, and says it was conceived to support the 800 VDC architectures that next-generation AI facilities are standardizing on.

The announcement matters because it puts solid-state chemistry — long treated as a automotive technology that would arrive in passenger cars first — into the most power-hungry segment of the computing industry within days of two European battery makers pitching the same bet. Rimac Technology and ECOBLOX announced their own joint development agreement on October 7, and battery suppliers across Europe are now describing data center-grade power as a distinct product line rather than a spin-off of EV manufacturing. For a battery market that has spent three years absorbing falling lithium prices and a wave of factory freezes, the arrival of a buyer willing to pay a premium for density and safety is a meaningful shift.

The AI Data Center Battery Problem

Every layer of an AI datacenter is being redesigned around power. Racks have moved from roughly 10 kW to 20 kW, and the industry is now engineering toward 1 MW racks, a jump of two orders of magnitude in three generations of hardware. At that scale, the traditional arrangement — a room-level lead-acid or lithium-ion UPS array plus a diesel or turbine genster — becomes difficult to engineer and expensive to build.

The constraint is a power wall, not an energy shortage. A training workload must be served continuously, and every watt inside the rack has to be conditioned and backed up at a higher voltage than copper-heavy legacy designs allow. This is why the industry is migrating toward 400 VDC and 800 VDC bus architectures: they move the same power at lower current, which cuts resistive losses and shrinks the cable and switchgear footprint.

Batteries sit at the center of that redesign. Conventional lithium-ion cells store substantial energy, but their usable power is bounded by their internal resistance and their thermal behavior. Run them hard and they heat, and heating is where their history becomes a liability. Most conventional lithium-ion cells enter thermal runaway at around 180 °C, and in a facility worth hundreds of millions of dollars, that failure mode is a design constraint, not a footnote. The industry has spent a decade mitigating it with containment, water suppression, and spacing — all of which consume floor area, which is precisely what a 1 MW rack cannot spare.

An in-rack battery is the market's answer: put the storage as close as possible to the load, condition power at the voltage the compute hardware uses, and give the site a momentary buffer that smooths the fluctuations a grid connection sees when thousands of accelerators start a training step simultaneously. The economics work only if the battery is denser and safer than a conventional pack, which is where solid-state chemistry has become relevant to data center buyers for the first time.

What the QS PowerBlock Claims

QuantumScape's launch materials are unusually specific for a prototype-stage product. The company describes the QS PowerBlock as a modular, adaptable reference design that can fit a variety of module, shelf, and rack formats, and is designed for deployment in industry-standard power rack configurations. Rather than a finished product, it is positioned as a blueprint that customers can deploy with their own power system components.

The headline numbers come from a demonstration using an expected data center training workload, benchmarked against the Open Compute Project (OCP) Open Rack V3 specification. QuantumScape claims the design exceeded key OCP specifications on three dimensions:

• 4x the power density of current industry standards, which the company argues lets power-hungry AI systems integrate into compact and cost-effective data center designs.
• 5x longer runtime, giving operators more margin to smooth out power fluctuations and improve hardware productivity.
• Better safety than conventional lithium-ion, with QS battery cells having demonstrated thermal stability well over 300 °C in third-party and customer safety testing, against roughly 180 °C for the onset of thermal runaway in conventional cells.

The safety figure is the one to read closely. A cell that does not enter thermal runaway at 300 °C is a different asset class from a conventional lithium-ion cell, and in a facility where a thermal event could take out a billion dollars of hardware, that difference has a price attached to it. The mechanism is QuantumScape's proprietary solid ceramic separator, which replaces the combustible organic porous separator and the liquid anolyte of a conventional cell with a solid, non-flammable ceramic barrier.

QuantumScape has also announced that it joined the Open Compute Project as a member. That is a quieter but strategically significant move: OCP is the industry consortium whose open hardware specifications define how racks, power and cooling are standardized for large-scale deployments. A battery supplier joining OCP is telling the market it intends to compete on standards conformance, not proprietary design, which lowers the barrier for a hyperscaler or a large-scale operator to evaluate the cell as a component.

The company is putting the design in front of buyers at the 2026 OCP Global Summit, held October 12-15 at the San Jose Convention Center, at booth C19, hosted by Megmeet, an official partner and power system component provider for the NVIDIA MGX Ecosystem.

The Chemistry Question

QuantumScape's approach is an anodeless architecture paired with a solid ceramic separator. The company's own explainer material describes the logic plainly: eliminate the carbon or silicon host material that stores lithium in a conventional anode, form the lithium metal anode in situ on the first charge, and replace the liquid separator with a solid-state barrier that prevents dendrites.

The claimed benefits follow from those two eliminations. Removing the anode host material improves energy density. Removing the liquid electrolyte removes the combustible material. Removing the diffusion bottleneck at the anode interface is what the company credits for fast charge — its stated target is a 10 to 80 percent charge in under 15 minutes.

The caveats matter as much as the claims, and the company's own disclosures are the honest source for them. QuantumScape states that its 24-layer A0 prototype cell has completed more than 1,000 full charge-discharge cycle equivalents with more than 95 percent energy retention, and that its commercial target for volumetric energy density is 800-1,000 Wh/L. It also states that it has not completed development of its multilayer commercial battery cell and therefore has not yet conducted safety tests on commercial target batteries and packs. In other words, the QS PowerBlock's numbers are prototype numbers, benchmarked against a specification rather than a shipping product line.

That is not an unusual position for a supplier entering a new market at the component level. It does, however, set the bar for what would count as proof. On the automotive side, QuantumScape's public narrative rests on a capital-light licensing model in which its QSE-5 cells and Eagle Line blueprint flow into partners including PowerCo, Murata and Corning. The data center play is a second front on the same platform, and the risk profile is the one investors have already been told to weigh: QuantumScape currently records no revenue from cells and remains loss-making, and the AI data center opportunity is only worth something if a paid development program follows the launch.

Two Companies, One Week, Same Answer

The most telling detail in this past week of announcements is not QuantumScape's specification sheet — it is how many other companies reached the same conclusion independently.

On October 7, one day before the QuantumScape release, Rimac Technology and ECOBLOX announced a joint development agreement to deliver a modular data center solution. Rimac is a Tier 1 automotive technology supplier wholly owned by the Rimac Group, employing more than 1,000 people in Croatia with offices in the UK. It supplies battery systems, e-axles and electronic systems to named OEMs, and it brings more than a decade of high-voltage engineering. ECOBLOX, headquartered in Dubai, is an NVIDIA Preferred Partner and the Middle East and Africa region's "NVIDIA Star Performer" for 2026, delivering turnkey modular data centers in six to eight months.

The joint product integrates Rimac Technology UNI battery modules — automotive-grade, developed to strict automotive standards, designed for serviceability and longer continuous backup power — with ECOBLOX's power conversion, UPS, precision cooling and IT infrastructure inside a single modular footprint. The battery management system streams telemetry directly into the ECOBLOX data center infrastructure management platform for centralized visibility and analytics. The companies state they are targeting the same 400 VDC and 800 VDC architectures as QuantumScape, and a completed modular data center is on track for installation at the Rimac Campus near Zagreb in November 2026.

The convergence is what is significant. A Croatian drivetrain supplier and a Dubai-based AI infrastructure builder spent nine months building an automotive battery into a data center power system, while a California company that spent fifteen years on solid-state chemistry announced a comparable product two days later. The buyer-pull is not coming from one software vendor or one cloud operator; it is coming from a structural power constraint that any AI-scale deployment hits.

This is a market formation signal rather than a product story. Two independent launches in 48 hours, targeting the same voltage architecture and the same pain point, indicate that "data center battery" is becoming a recognized category with its own specifications — one whose buyers have the same requirements as automakers (safety, density, thermal management) but different ones on cycle life, serviceability and the tolerance for prototype-stage hardware.

The Honest Caveats

Both launches come with qualifications that the press releases do not emphasize.

First, the performance figures are prototype numbers, not commercial commitments. QuantumScape's release states this directly: results were achieved through prototype testing and modelled system configurations under controlled conditions, with comparisons benchmarked against OCP Open Rack V3, and final performance will depend on product design, system configuration and real-world operating conditions.

Second, the generalization risk is real. QuantumScape's 1,000-cycle result comes from a 24-layer A0 prototype. Commercial cells with higher layer counts in a target form factor may behave differently, and the company has said so explicitly in its own FAQ material. The industry has seen solid-state timelines slip before, and the gap between a working prototype and a qualified production cell is where most battery programs lose their schedules.

Third, neither announcement is a shipping deployment. QuantumScape is showing a reference design at a standards summit; Rimac and ECOBLOX are demonstrating a modular data center at a campus in November. Both are credible next steps, and neither is a data center serving production traffic.

Fourth, the buyer-pull cuts both ways. AI data centers are an unusually attractive customer because their power density demands justify a premium for better chemistry, but they are also an unusually demanding one on cost and supply. QuantumScape's own FAQ notes that if solid-state succeeds, near-term demand will likely outstrip any single vendor's ability to produce, which is precisely why the data center market and the automotive market will need separate manufacturing capacity.

What to Watch

Three signals will show whether this becomes a market.

The first is whether a hyperscale operator or a major cloud signs a paid development program. QuantumScape's investors page names the test plainly: a disclosed commercial agreement with a large data center operator or hardware partner under OCP, with contracted rack units, contracted watt hours, or dollar value of new AI data center billings. Any of those numbers appearing in a quarterly update is a stronger signal than the launch itself.

The second is standards adoption. QuantumScape has joined OCP, and a PowerBlock-derived design being reflected in an open power specification is the mechanism by which a single supplier's chemistry becomes an industry component. That path exists in the semiconductor and networking worlds, but it is unproven for batteries, and it is the fastest way for a supplier to legitimize a new device class with procurement teams.

The third is whether the automotive-side story holds. QuantumScape's licensing model rests on converting QSE-5 cells and the Eagle Line blueprint into partner programs. If the automotive timeline slips, the data center market is where the platform has to earn its keep, and the two narratives are not independent.

Conclusion

The announcement of the QS PowerBlock and the Rimac-ECOBLOX agreement in a single week mark the moment solid-state lithium-metal chemistry stopped being an automotive project and became an AI infrastructure component. The numbers QuantumScape published — four times the power density, five times the runtime, thermal stability above 300 °C — are exactly the attributes a 1 MW rack needs, and it is not accidental that they arrived in the week two European companies presented the same argument with different chemistry.

The applications that follow are the interesting part. A battery that is safe at 300 °C changes the containment engineering inside a data center; a battery with 4x power density changes the rack layout; and a battery that costs a premium but delivers a decade of reliability inside a facility worth hundreds of millions of dollars changes the procurement arithmetic. None of that is certain yet — QuantumScape's own disclosure is that the multilayer commercial cell is unfinished — but the direction of travel is clear. Data centers started this year as a battery market nobody priced, and by October 2026 they were the place where the industry's newest chemistry was being pitched to buyers first.

For IoT and connected-device designers the read-across is more indirect but not zero: the same 400 VDC and 800 VDC shift that is redrawing data center power architecture is the voltage frontier that edge sites and micro data centers will inherit next, and modular, automotive-grade storage is the most likely template for it. That transition is tracked in our Internet of Things coverage, and adjacent to it in our Semiconductors category.

Images

References

• QuantumScape Announces QS PowerBlock for In-Rack Energy Storage in AI Data Centers
• QuantumScape Announces QS PowerBlock for In-Rack Energy Storage in AI Data Centers (full release text)
• Rimac Technology and ECOBLOX Partner to Bring Rimac Battery and Power Electronics Technology to AI Data Centers
• Rimac Technology partners with Ecoblox on solutions for AI data centers
• Solid State Battery Technology | QuantumScape
• Croatia's Rimac Technology partners with Ecoblox on AI data centre power systems