Grid-storage debates are usually framed as a contest among technologies. Lithium-ion batteries compete with flow batteries, compressed air, liquid air, gravity systems, iron-air and whatever new long-duration acronym has arrived with a funding announcement. That framing skips the more important system question: how much dedicated electrical storage is actually required after transmission, flexible demand, thermal storage and strategic reserve have done the jobs they perform more efficiently? My updated grid-storage scenario through 2100 lands at roughly 108.5 TWh of dedicated electrical storage in a mostly electrified world. That is a scenario anchor, not a prediction or a number copied from another model. It is also substantially lower than my earlier estimate because the boundary has improved. The new model no longer asks electrical storage to solve every seasonal, geographic and rare-event reliability problem. The technology hierarchy has changed as well. In my 2021 assessment, pumped hydro was the clear first choice for bulk storage, lithium-ion owned the shorter-duration market and redox flow looked like the strongest candidate to expand as durations increased. Pumped hydro’s role survived, but batteries scaled faster and became cheaper than I expected, while flow batteries failed to build comparable manufacturing scale or bankable deployment. Global battery-storage deployment reached 108 GW in 2025, about 40% more than in 2024, with installed capacity roughly eleven times its 2021 level. Stationary battery-pack prices fell to about $70/kWh, according to the data used in the pathway assessment, a 45% reduction in a year. Complete installations still require inverters, containers, civil works, fire protection, grid connections and permits, but the central manufactured component has become cheap enough to keep pushing into longer durations. The old shorthand that batteries are suitable only for two- and four-hour storage is becoming increasingly difficult to defend. The United Kingdom’s first long-duration storage support portfolio includes lithium-ion projects with proposed durations from eight to eighteen hours. These projects are not yet a global operating fleet, and final financing and delivery still matter, but longer-duration batteries have moved from theoretical extrapolation into formal utility procurement. China provides the larger-scale signal. Its battery-storage fleet reached approximately 106.9 GW and 240.3 GWh by May 2025. Average duration remained a little above two hours, so this does not prove that ten-hour batteries are already normal. It does demonstrate that batteries are scaling faster in power-capacity terms than any other dedicated storage class and have become central infrastructure rather than a demonstration technology. Pumped hydro remains enormously important because it already performs bulk gravity storage at grid scale. It combines mature turbines and generators with long asset lives and useful round-trip efficiency. China had 58.69 GW operating by the end of 2024 and, according to the evidence assembled in the Briefing pathway, roughly 218 GW under construction. The constraint is that pumped hydro is civil infrastructure rather than a manufactured product. Every project needs suitable geography, water, transmission, revenue certainty, permitting, patient capital and political tolerance. Batteries need factories, materials, inverters and interconnection, but each installation does not require a mountain, two reservoirs and a decade of project development. The 2100 pathway therefore keeps pumped hydro large without pretending that it can be replicated everywhere at the speed of a containerized product. Flow batteries retain a secondary role. Their stationary architecture, separation of power from energy capacity and potentially long cycle lives remain useful characteristics. The commercial problem is that grid-storage markets are won through delivered cost, financeable warranties, repeat procurement and operating fleets. Lithium-ion’s cost decline and manufacturing scale have narrowed the duration band that flow batteries once expected to inherit. The remaining long-duration category should be treated more skeptically. Compressed air, liquid air, gravity blocks, iron-air and thermal-to-electric concepts may all find specific applications, but a category label is not deployment evidence. A base-case pathway should not allocate tens of terawatt-hours to technologies merely because they promise one hundred hours of storage. They have to demonstrate repeatable projects, bankable warranties and operating fleets before they receive a large share of a century-scale model. The updated pathway is therefore batteries-led, pumped-hydro anchored and cautious about the rest. Dedicated electrical storage reaches approximately 108.5 TWh by 2100, with batteries taking the largest share, pumped hydro supplying the long-life bulk layer and flow batteries remaining material but secondary. Other electrical long-duration storage stays small until operating evidence requires an upgrade. The 2100 scenario is batteries-led and pumped-hydro anchored, with flow batteries secondary and other electrical LDES remaining small. That total makes sense only inside the correct system boundary. Storage shifts electricity across time; it does not produce net energy. A reliable clean grid also uses long-distance transmission to move electricity across weather systems and time zones, flexible demand to move consumption into periods of abundant generation and thermal storage to shift heating and cooling without converting the energy back into electricity. Seasonal thermal storage is especially important in cold climates. Heat pumps remain the right technology for electrifying buildings, but winter peaks become unnecessarily expensive when every unit of heat has to be supplied directly from the grid during the coldest evening. Aquifer storage, borehole fields, district-energy systems and large hot-water stores can shift heat across days or seasons while reducing the electrical generation, storage and transmission required for peak conditions. Flexible demand is another subtraction that should be counted before planners buy more storage. Electric-vehicle charging, water heating, refrigeration, cooling, pumping, desalination, data-centre workloads and parts of industrial production can move across hours or days. During rare stress events, paying a large industrial user to stop temporarily may be cheaper than building infrastructure that operates only a few hours each decade. Strategic reserve also belongs outside the daily storage stack. A system may retain small quantities of dispatchable biofuel, likely biomethane, for severe multi-day or multi-week events without turning that reserve into a normal cycling resource. Forcing rare-event resilience into the electrical-storage category inflates the storage requirement and creates an artificial market for technologies that are poorly suited to everyday grid operation. The implications for capital allocation are fairly direct. Battery infrastructure, power electronics, controls, thermal management, grid software, recycling and interconnection sit at the centre of the scaling market. Pumped hydro can be attractive where the site, permitting and revenue model work, but it is a long-duration infrastructure and project-finance business rather than a venture-scale manufacturing story. Flow batteries remain a diligence-heavy minority position, and other long-duration technologies still need fleets rather than demonstrations. Utilities procure services rather than hype. Response time, duration, cycling frequency, location, deliverability and availability are more useful requirements than a mandate for a particular technology. A genuinely technology-neutral procurement process will currently buy mostly batteries, some pumped hydro, selective flow projects and relatively little from the speculative storage basket. Policymakers should be equally careful not to turn storage targets into a substitute for system design. Transmission, demand response, seasonal thermal storage, industrial flexibility and strategic reserve are reliability resources. Ignoring them will cause some grids to overbuy batteries, others to overpay for immature technologies and many to underinvest in measures that reduce the storage requirement altogether. A mostly electrified world will need a great deal of dedicated electrical storage, plausibly on the order of one hundred terawatt-hours by 2100. Batteries have earned the leading position through cost decline, deployment speed and manufacturing scale, while pumped hydro remains the durable bulk anchor. The rest of the market must compete not only with those technologies, but with the transmission, flexibility and thermal systems that prevent storage from having to solve every grid problem. The full TFIE Strategy Briefing pathway review includes the decadal model, 108.5 TWh and 9.2 TW scenario stack, comparator analysis, update triggers, scorecard and downloadable workbook.