EVs cool while battery cells climb
In 2026 electric vehicles cooled. Tesla -16.6%, BYD -25.3%. Yet in the same year battery cells rose. Panasonic Energy +116.1% (after +36.2% in 2025), Samsung SDI +16.7% (after +75.7% in 2025), CATL +10.6% (a second straight positive year after its +73.7% rebound in 2024). That is the opposite of the assumption that automakers and cell makers move together. The reason cells rose while EV demand softened lies elsewhere. The equation of battery equals EV cannot read this cycle. What drove the 2025-26 battery rebound was ESS and AI-datacenter power storage.
Batteries look unrelated to AI at first glance. But line up the returns of the five cell makers this wiki tracks and the root of their rebound reaches back, again, to the AI datacenter. This is the test of whether, on a chain outside AI, infrastructure re-rates before the application (the same grammar inside AI was confirmed in The demand of techbio).
The ones who pivoted win
The deciding variable was a single thing: how fast each pivoted to ESS. Samsung SDI secured a Tesla ESS-LFP order (about $2.1B) and L&F LFP cathode to become a core node of the non-Chinese ESS chain. It carried an FY2025 operating loss of ₩1.72tn on ₩13.27tn of revenue, yet its stock rose +75.7%, which looks like the market paying for the pivot more than for the earnings. Panasonic turned its weakness, dependence on Tesla, into strength through Nevada and Kansas localization and datacenter ESS. CATL, with 464.7GWh installed in 2025 and 39.2% share, takes top-of-chain demand first and in the largest volume.
Those that pivoted late with EV-heavy customers were left out of the rebound. In the same rebound, a customer book tied to EVs collects little even as cell prices rise. SK Innovation (SK On), its customers concentrated in Ford, Hyundai and Nissan, fell -10.6% in 2025 and -20.1% in 2026, negative almost every year. LG Energy Solution, with its top five automaker customers at about 67% of revenue and locked into GM, Honda and Stellantis EV joint ventures, fell -16.7% in 2026, with the Tesla Megapack-3 LFP deal left as its ESS-entry option.
The root of ESS demand is again the AI datacenter
The pattern connects precisely to the wiki's AI and datacenter-power chain. AI compute lifts datacenter power demand, and the storage demand that governs when and where that power is used moves into grid and datacenter ESS, and then into battery cells. The true root of ESS demand is the AI power bottleneck. The power shortage in power is AI's next bottleneck creates storage demand, and that demand follows the propagation path drawn in did the AI cycle propagate along the value chain down to battery cells. The battery cell is a terminal node of that semiconductor and power chain.
The same grammar: infrastructure before application
Here the grammar of each frontier re-rates infrastructure before application shows up in batteries too. The application of the battery frontier is the EV; its infrastructure is grid and datacenter storage. While the application (EV) softened, the infrastructure (ESS) re-rated, the same order in which AI re-rated infrastructure (power, substrate) before application (drugs, chatbots). And because the root of that battery-infrastructure demand is the infrastructure of yet another frontier, the AI datacenter, the two chains interlock.
Verdict
Case two lands on the same conclusion. What drove the battery cycle was ESS, and the root of that ESS demand is again AI-datacenter power. Infrastructure re-rates before application, and the cells that pivoted to ESS won.
Yet climb a single step above the cell and the cycle breaks. As the cathode paradox shows, cells turning does not mean materials follow at once. Whether LFP and ESS-side materials turn before EV-side ones is that piece's question. And now the cases have gathered: AI semiconductors, space, techbio and batteries. What remains is to put all these chains on one screen. Many clocks overlays those clocks on a single map.