The cheapest layer is solved. The assembly is not.最便宜的那一层已经解决，装配没有。
A companion piece which found storage at roughly 2.6% of transition capital and called it "technology ready · capital thin." This station peels the object open to ask a harder question: thin where, exactly?
Each layer carries a FUNCTION 功能, a COST share 成本, and a CAPITAL note 资本 — who funds it, who makes it, what's missing. Scroll to peel the drawing apart, or click a number in the index.
Storage took $71B of the record $2.3T in energy-transition investment in 2025 — about 3.1%, up from $54B and 2.6% in 2024. Still a rounding error against the rest of the build.
A thin dollar figure is not automatically a thin buildout. In 2025 the world installed a record 112 GW / 307 GWh of storage, excluding pumped hydro, and BNEF forecasts 158 GW / 459 GWh for 2026 — a further 41% rise, achieved while solar and wind additions slowed. Ten-year additions grow about 23% annually to 2 TW / 7.3 TWh.
Storage is the one line where a small number is partly a triumph — priced down faster than it can be capitalised up.在这一行里，小数字本身就是一种成就——降价的速度超过了资本化的速度。
Co-located solar-plus-storage added 87 GW in 2025 at an average $57/MWh, while gas CCGT levelised cost rose 16% to an all-time-high $102/MWh: firmed renewables now beat new gas on price.
Then turn it: deflation in the cell layer does nothing for the layers that are not deflating. Transformers, interconnection studies, fire marshals, commissioning crews and separator plants do not fall 45% a year. The peel shows where the cost curve stops working.
Anode share of cell cost, against China's share of anode active material. Cheapness is why the dependency persists.
Balance of plant and soft costs, against a quarter to a third for the cell. The industry negotiates the wrong number.
Storage's share of transition capital, against its forecast deployment growth this year. Either capital is about to follow, or the price collapse is doing work capital should be doing.
The four-hour container is standard because four hours is what today's markets pay for, not what a winter week needs. Multi-day storage has no comparable cost curve and no standard product.
Separator, electrolyte and anode plants outside China at qualification-grade quality.
Sodium-ion arriving in commercial volume, signalled by a 60 GWh multi-year CATL–HyperStrong agreement — it removes graphite and lithium exposure entirely.
Battery-grade lithium carbonate rebounded over 90% from a June trough near $8,100/tonne, on storage demand, low Chinese inventory and the CATL Jianxiawo suspension. Metal shocks didn't reach 2025 pack prices — absorbed by LFP adoption, contracts and hedging — but that absorption is finite.
The queue no cost curve touches.
The cell is solved, and getting cheaper faster than anyone forecast. The assembly is not. Capital is not too thin for batteries in aggregate — it is misallocated within the object. Attention and money concentrate on the layer that is already efficiently priced and structurally sound, and thin out precisely as you move toward the layers that are cheap, invisible, geographically concentrated, queue-bound, and capable of stopping everything.
An exploded drawing is how engineers find the part that will fail. This is that drawing, for an asset class.装配图的用途，是找出会失效的那一个零件。