XPeng's announcement that solid-state battery cells for its IRON humanoid robot have entered small-batch production is a notable milestone, but its significance depends on a detail the company has conspicuously withheld. Gu Jie, XPeng vice president, said on October 28 at the SAE-China Congress and Exhibition that the IRON robot's solid-state battery is now in small-batch production, while the robot itself remains in the pre-SOP (start of production) phase1,2. The batteries are rolling off the line ahead of the complete robot, which XPeng plans to mass-produce by the end of 2026, with deployment in stores and commercial settings starting in 20274. What makes this disclosure consequential is what it reveals about XPeng's engineering priorities: the company has placed energy density at the center of its humanoid production timeline, ahead of software or actuator readiness. That bet is debatable; other humanoid programs treat data infrastructure or actuator cost as the binding limit. But XPeng is the first major player to stake a production timeline on a solid-state answer, and the chemistry behind that answer remains conspicuously unclear.
The numbers XPeng has disclosed underscore why it went this route. The company said in February that the solid-state battery pack achieves 78% volume utilization, 23 percentage points higher than conventional liquid battery packs, while cutting weight by 35%. For a humanoid form factor with very little room for a battery pack, those margins are not incremental improvements; they are the difference between a robot that works a shift and one that doesn't. Gu described the design as high-energy-density, built to fit IRON's limited internal space. A 23-percentage-point gain in volume utilization and a 35% weight reduction, taken together, represent a meaningful gain in the energy a humanoid can carry without compromising mobility—though the magnitude of that gain depends on whether the cells are truly all-solid-state or semi-solid-state, a distinction XPeng has not settled publicly.
But the most telling detail is the one XPeng has not provided. When the company unveiled the new-generation IRON at its 2025 Tech Day, it said the robot would be the first in the industry to use an all-solid-state battery. In September, some English-language reports said the production version would switch to a semi-solid-state battery. Gu's October 28 remarks used only the general Chinese term for solid-state, without specifying whether the cells are all-solid-state or semi-solid-state. Pandaily noted that XPeng has not publicly clarified which type the small-batch cells are.
That ambiguity matters. All-solid-state cells eliminate liquid electrolyte entirely, promising higher energy density and better safety but facing severe manufacturing challenges at scale. Semi-solid-state cells retain some liquid electrolyte, making them easier to produce but less of a leap. The distinction is not academic for a company that has said it will make all its batteries in-house starting in 2026 and that has invested in three battery cell companies. If XPeng has quietly downgraded from all-solid-state to semi-solid-state, the production timeline becomes more credible but the technical claim shrinks: semi-solid-state cells are an incremental improvement over conventional lithium-ion, not a category shift, and the 78% volume-utilization and 35% weight-reduction figures would need to be re-evaluated against the best conventional packs rather than treated as proof of a new chemistry regime. If it has held to all-solid-state, the manufacturing risk at scale is substantial. Either way, the disclosure forces the question into the open: XPeng has committed capital and a CEO-led org chart to the proposition that its humanoid cannot ship on today's battery technology, and the industry now has to respond to that framing whether or not the underlying chemistry justifies it.
XPeng is not building this capability in a vacuum. The company closed an undisclosed round from IDG and Gaorong Ventures on August 24, and on July 9 CEO He Xiaopeng took direct leadership of the robotics business. XPeng Robotics held its first supply chain partner conference in September, completing factory audits and designating core components. On September 8, the company commissioned its robot production line, where an IRON unit completed automated final assembly and walked off the line autonomously, with more than 80% of core processes automated. The sequence—funding close (August 24), CEO-led robotics unit (July 9), production-line commissioning (September 8), battery small-batch output (disclosed October 28)—spans roughly three months of disclosed milestones. The pace suggests XPeng is treating the year-end mass-production target as a hard deadline rather than an aspiration.
LG Electronics, on the same day, disclosed its own robotics acceleration during its third-quarter earnings call—but its disclosed bottlenecks are data volume and actuator readiness, not battery chemistry. The company said it plans to enter the home humanoid robot market by 2030, following its push into manufacturing robots3,5. Lee Jae-mo, head of robotics strategy at LG Electronics, said the company will first focus on academia, public institutions, and manufacturing, then expand to home humanoids once the market takes shape. LG's approach is to build a track record in industrial settings and leverage its home appliance manufacturing capabilities.
LG's near-term bet is on data infrastructure. The company raised its Data Factory collection target fivefold, from 100,000 hours to 500,000 hours of cumulative data by year-end, after accumulation ran faster than expected. The Data Factory is a 10,000-square-meter robot training facility at LG's Yangjae R&D campus in Seoul, built as a hub for physical AI cooperation with Nvidia. CFO Kim Chang-tae said LG plans to begin full-scale operation of the facility in the fourth quarter6. The company is also expanding mass-production preparations for actuators, a core robot component.
ANALYSIS LG's 2030 home-humanoid target and XPeng's 2027 commercial deployment target are not directly comparable; LG is building from components and data infrastructure upward, while XPeng is driving toward a finished product. The contrast is instructive precisely because the two companies have identified different binding constraints. XPeng has bet that energy density must be solved first; LG has bet that data volume and actuator manufacturing must be solved first. Neither company has published runtime benchmarks, duty-cycle data, or weight budgets that would let an outside observer adjudicate the question.
Lee Jae-mo's comment that securing a price point consumers can actually buy at is important for home robot commercialization points to the cost side of the solid-state equation—and to a constraint LG considers at least as binding as energy density. XPeng's decision to bring battery production in-house is a bet that vertical integration can compress that cost curve fast enough to hit commercial price points by 2027. Whether it can is an open question.
Vector Wire previously reported on XPeng's $6.3 billion robotics valuation failing to shield its stock from an EV earnings miss, and on the company courting global tech-licensing deals to fund robotics and robotaxi bets. The solid-state battery milestone adds a tangible hardware proof point to what has, until now, been largely a valuation and licensing story. But the unresolved chemistry question—all-solid-state versus semi-solid-state—is a gap that will define whether this is a genuine technical lead or a marketing position. If the cells are semi-solid-state, the headline achievement is vertical integration speed, not a chemistry breakthrough.
Three markers to watch. First, whether XPeng clarifies the battery chemistry before year-end mass production; the distinction between all-solid-state and semi-solid-state will determine whether the industry treats this as a chemistry breakthrough or a vertical-integration play. Second, whether XPeng hits its end-of-2026 mass-production target for IRON itself—not just the battery cells—with published runtime and duty-cycle data that substantiate the energy-density thesis. Third, whether LG's Data Factory hits its 500,000-hour target by December, which would validate the fivefold acceleration and sharpen the contrast between LG's data-first and XPeng's battery-first approaches to humanoid commercialization.