South Korea’s secondary-battery stocks just had their best month in years, and the reason has nothing to do with electric vehicles. According to a report published by JKN on October 8, 2026 and updated a day later, eight of the ten best-performing exchange-traded funds on the Korea Exchange this month were tied to the battery sector, not EVs. The driver: investors betting that AI data centers will need vastly more energy storage system (ESS) capacity than anyone budgeted for just two years ago.
The shift is reshaping how battery makers talk about themselves. Companies that spent the last decade pitching EV range and fast-charging specs are now pitching rack-level power density and millisecond response times, because that is what hyperscale AI campuses are asking for. JKN’s reporting frames the rotation as a direct consequence of AI infrastructure spending outrunning the electrical grid’s ability to keep up with it.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Happened: Battery ETFs Lead Korea’s October Rally
Data compiled by the Korea Exchange and cited in the JKN report shows that eight of Korea’s ten best-performing listed ETFs during the month tracked the secondary-battery industry. That is an unusually lopsided result for a market that normally splits gains across semiconductors, shipbuilding, defense and biotech alongside batteries. The Korea JoongAng Daily separately confirmed the ETF concentration, tying it to expectations that ESS orders tied to AI data centers will start showing up in battery makers’ earnings.
What makes this rally notable is what it is not about. It is not about a breakthrough EV battery chemistry, a new Tesla contract, or a subsidy announcement. It is about expectations — specifically, the expectation that AI data centers represent a new, structurally larger demand bucket for grid-scale and rack-level batteries than the EV market alone ever created. No confirmed order volumes or completed deployments have been published yet; what moved the ETFs was forecasting and investor positioning ahead of anticipated contracts.
Why AI Data Centers Suddenly Need So Many Batteries
AI training and inference clusters do not draw power the way a server farm running web traffic does. Thousands of GPUs executing synchronized training steps can swing a facility’s power draw by tens of megawatts in a fraction of a second, then drop just as fast when a batch finishes. Traditional grid infrastructure and diesel-backup architecture were never designed for that kind of oscillation, which is why battery system suppliers are being pulled into a conversation that used to belong almost entirely to utility engineers.
The same instantaneous-response problem has already shown up in the cooling side of data-center infrastructure, where Trane Technologies’ stock run on its AI cooling bet reflects a similar scramble by legacy industrial suppliers to retool for AI-specific demand curves. Scott Blalock, General Manager of Software Operational Technology at Wärtsilä Energy Storage, described how AI data-center developers now come to battery suppliers without even knowing their own numbers. “With an AI data centre developer, they don’t know that target, and they’re coming to us saying, ‘Tell me how much battery I need,'” Blalock said. He added a blunter assessment of why grid-scale batteries have become non-negotiable for these projects: “It’s basically impossible to have that without a BESS.”
Frank DeLattre, President of Americas at Musashi Energy Solutions, framed the same problem from the facility side in a conversation with Data Center Frontier. “The facility might have enough average power available. The problem is responding to the instantaneous changes,” DeLattre said, describing why average capacity numbers on a spec sheet stop being useful once GPU clusters start training. He went further, arguing the category itself is being redefined: “We’re moving from energy storage as a backup infrastructure to really now energy storage … as part of an active power infrastructure.”
TrendForce Forecast: ESS Capacity Set to Jump From 15.7 GWh to 216.8 GWh
The number driving most of the investor enthusiasm comes from TrendForce. The research firm projects that global installed ESS capacity serving AI data centers will grow from 15.7 GWh in 2024 to 216.8 GWh in 2030, a compound annual growth rate of roughly 46.1%. That is close to a 14-fold increase in six years, in a category that barely existed as a distinct market segment before AI training clusters started scaling past the hundred-megawatt mark. The 15.7 GWh figure is a 2024 baseline rather than current installed capacity, and the 216.8 GWh figure is a 2030 forecast, not a completed deployment or a 2026 market total. A separate forecast from DataM Intelligence, published in August 2026, points the same direction in dollar terms: it valued AI-data-center BESS at $1.65B in 2025 and projects $31.97B by 2035, a 34.5% CAGR over 2026 to 2035.
TrendForce also describes the architecture shift behind that number: rather than relying on one centralized backup system, operators are increasingly layering large-scale battery energy storage systems (BESS), rack-level battery backup units, and floor-level supercapacitors. Each layer handles a different time scale of power fluctuation, from sub-second transients to hours of grid outage coverage. That tiered approach is part of why the total addressable market for battery suppliers is growing faster than raw gigawatt-hour figures alone would suggest — data centers need batteries at multiple points in the power chain, not just one.
Samsung SDI and LG Energy Solution Pivot Toward the Rack
Samsung SDI has staked out a position in its own public materials, describing a data-center battery portfolio spanning battery backup units (BBU), uninterruptible power supply (UPS) systems, and full-scale ESS. One flagship product in that lineup, the U8A1, is rated at 263 kW of output per rack and uses an LMO cathode chemistry that Samsung SDI says delivers low internal resistance — a property that matters directly for handling the rapid charge-discharge cycling that AI workloads create. Separate Korean industry coverage from Edaily describes Samsung SDI as one of several battery companies moving beyond utility-scale ESS toward equipment positioned physically closer to server racks. At its September 2026 showing, the BBU lineup included high-power, high-capacity 21700 nickel-based cylindrical cells with a bottom-vent safety design, and the company unveiled a cylindrical LFP battery for the first time at that event. It also reportedly presented prismatic LFP batteries for Samsung Battery Box (SBB) 2.0, an integrated ESS combining batteries, power-conversion systems and related components.
LG Energy Solution’s shift is even more visible in its production targets. Industry research cited by KB Securities indicates the company raised its U.S. ESS production capacity target from 30 GWh to 50 GWh by the end of 2026, a jump the research firm links directly to AI-data-center demand. The same analysis forecasts that LG Energy Solution’s ESS revenue could roughly triple between 2025 and 2026, and that U.S. ESS shipments could exceed the company’s EV-battery shipments starting in 2026 — a startling reversal for a company whose identity has been built almost entirely around EV supply contracts for the past decade.
Why the EV-to-ESS Shift Matters for Battery Makers’ Balance Sheets
EV battery contracts are notoriously margin-thin and tied to automaker production schedules that have proven unpredictable over the past two years. ESS contracts tied to AI data centers come from a different buyer entirely: hyperscale cloud operators and specialized data-center developers with capital budgets measured in the tens of billions of dollars and an urgent timeline, because every month a facility sits without adequate power backup is a month of idle GPU capacity. That buyer profile is part of why Korean battery makers are reallocating manufacturing lines that once served EV customers toward ESS-specific cell formats.
QuantumScape Enters the Race With QS PowerBlock
QuantumScape, a company that built its public profile almost entirely around solid-state batteries for electric vehicles, announced on October 8, 2026 that it was entering the data-center market with a new product called QS PowerBlock. According to Axios, the company is positioning QS PowerBlock as an in-rack energy-storage system built specifically for AI data centers, rather than a repurposed EV cell format.
QuantumScape states that QS PowerBlock is designed to provide four times the power density and five times the runtime of current data-center industry standards, built around an 800VDC architecture and the company’s QSE-5 solid-state lithium-metal cell. The design is intended to support next-generation AI data-center racks of approximately 1 MW, and QuantumScape’s third-quarter shareholder materials, released on October 21, 2026, describe it as an “in-rack energy storage reference design” rather than a finalized mass-market product, with no purchase price published. Those are the company’s own figures rather than independently verified third-party benchmarks, but the pivot itself is significant: a company that spent years positioning solid-state chemistry as an automotive differentiator is now betting that AI infrastructure, not EVs, is the nearer-term commercial opportunity for the same underlying technology. The company also reportedly announced on October 21, 2026 plans for a 100 MWh solid-state battery production line aimed partly at AI data centers, aerospace and defense, with Muehlbauer Group providing cell-assembly equipment; that is a production-line plan and partnership, not a confirmed system launch.
Electrovaya and the Containerized Battery Model
Not every company chasing this market is building rack-level hardware. Electrovaya has taken a containerized approach with its ElvaPulse 1500 system, rated for up to 7 MW of continuous power with 2.88 MWh of nominal energy capacity, built on a 1,500 VDC architecture inside a standard 20-foot shipping container. The containerized format targets a different part of the data-center power stack than Samsung SDI’s rack-mounted U8A1 or QuantumScape’s in-rack PowerBlock — it is designed to sit at the facility level, smoothing larger swings in grid draw rather than handling the millisecond-scale fluctuations closest to the GPU itself.
That split between facility-level containerized storage and rack-level in-server storage is becoming one of the clearest architectural dividing lines in this emerging market, and it suggests there is unlikely to be one dominant battery format the way lithium-ion cylindrical cells dominated consumer electronics. Data centers appear to need both layers simultaneously, which is good news for suppliers pursuing different niches rather than head-to-head competition on a single spec sheet.
CBAK Energy Bets on High-Rate LFP Cells for Backup Power
CBAK Energy is targeting the backup-power and UPS segment specifically, rather than the larger grid-scale ESS category. The company’s 26650 HP V2.0 cell is specified at a 40C discharge rate, while its 26650 PFS2 V2.0 cell is specified at 38C, both using lithium iron phosphate (LFP) chemistry. According to a GlobeNewswire release, CBAK Energy states both cells are designed to support pulse discharge rates of up to 100C under specified test conditions — a figure aimed squarely at the instantaneous-response problem Frank DeLattre described, rather than at long-duration energy storage.
High discharge-rate LFP cells occupy a specific niche: they are not meant to store large amounts of energy for hours, they are meant to dump a lot of power very quickly and then recharge, which is exactly the profile a battery backup unit or UPS needs when a GPU cluster’s power draw spikes before a generator or grid adjustment can respond. It is a narrower, more specialized play than Samsung SDI’s or LG Energy Solution’s broader ESS portfolios, but it targets a problem every AI data center operator has to solve regardless of scale.
Battery Maker Comparison: Who Is Targeting What
| Company | Product / Focus | Key Specification | Target Use Case |
|---|---|---|---|
| Samsung SDI | U8A1 rack battery | 263 kW output per rack, LMO cathode | Rack-level BBU, UPS, ESS |
| LG Energy Solution | U.S. ESS capacity expansion | Target raised from 30 GWh to 50 GWh by end-2026 | Grid-scale ESS for data centers |
| QuantumScape | QS PowerBlock | 800VDC architecture; claims 4x power density, 5x runtime vs. industry standard | In-rack energy storage |
| Electrovaya | ElvaPulse 1500 | 7 MW continuous power, 2.88 MWh capacity, 1,500 VDC, 20-ft container | Facility-level containerized ESS |
| CBAK Energy | 26650 HP V2.0 / PFS2 V2.0 | 40C and 38C discharge rate; up to 100C pulse discharge claimed | Backup power units, UPS |
The Numbers Behind the Forecast
| Metric | 2024 / 2025 | 2026 / 2030 Forecast | Source |
|---|---|---|---|
| AI data-center ESS installed capacity | 15.7 GWh (2024) | 216.8 GWh (2030 forecast) | TrendForce |
| Implied CAGR, AI data-center ESS | — | 46.1% (2024-2030) | TrendForce |
| Korea’s top-10 ETFs tied to batteries | — | 8 of 10 (October 2026) | Korea Exchange, via JKN / Korea JoongAng Daily |
| LG Energy Solution U.S. ESS capacity target | 30 GWh | 50 GWh by end-2026 | KB Securities research |
| North American LG Energy Solution factories making ESS batteries | — | 5 of 8 factories by end-2026 | LG Energy Solution |
Market Impact: What the ETF Surge Actually Signals
It is worth being precise about what this rally does and does not represent. The ETF performance reflects investor positioning around anticipated ESS demand, not confirmed purchase orders flowing through to battery makers’ income statements. No public figures confirm how many gigawatt-hours of battery capacity AI data-center operators have actually contracted for delivery in 2026 or 2027. What is confirmed is the direction companies are moving: capacity expansion targets, new product lines, and public statements from Samsung SDI, LG Energy Solution, QuantumScape, Electrovaya and CBAK Energy all point the same way, toward data centers as a genuine new demand category rather than a marketing talking point. Tesla’s own Q3 2026 Update adds to the picture of strong storage demand in general: the company says its energy storage business recorded its highest-ever order rate and backlog during the quarter, with deployments of 13.7 GWh, up about 10% from 12.5 GWh a year earlier, and CEO Elon Musk said on the earnings call that Tesla is almost sold out of its Megapack battery systems.
That distinction matters because battery stocks have a history of running ahead of realized demand, particularly in Korea, where retail investor enthusiasm for thematic ETFs can move prices well before revenue catches up. The same dynamic played out with EV battery stocks between 2020 and 2022, when anticipated EV adoption curves drove valuations that took years to be justified by actual vehicle production volumes. Whether the AI-data-center ESS story follows the same pattern or resolves faster, given how much capital hyperscalers are already deploying into physical infrastructure, is the open question investors are currently pricing.
Historical Context: From Grid Backup to Active Power Infrastructure
Batteries in data centers are not new. Every data center has run some form of backup power for decades, typically lead-acid UPS batteries sized to bridge the gap between a power failure and generator startup, measured in minutes. What has changed is the role batteries play when the facility is operating normally. DeLattre’s description of a shift “from energy storage as a backup infrastructure to really now energy storage … as part of an active power infrastructure” captures the structural change: batteries are no longer a contingency plan, they are a standing component of how the facility manages its everyday power draw.
That evolution tracks a broader pattern in how the AI infrastructure buildout has repeatedly outpaced categories that were previously treated as mature and commoditized. Memory pricing, high-bandwidth networking, and now grid-interconnection capacity have all gone through similar demand shocks as AI data-center construction accelerated through 2025 and into 2026. Readers who have followed Microsoft, Meta and Google’s combined $725 billion AI capex commitments have already seen this dynamic play out in chip spending; batteries appear to be the next line item following the same curve.
Battery chemistry itself has also been part of this story outside the data-center context. ProLogium’s 381 Wh/kg solid-state battery, for instance, was positioned by its maker as a bet on AI hardware demand rather than EVs specifically, underscoring how battery makers across the industry are increasingly hedging toward AI-linked revenue streams instead of relying solely on automotive contracts.
Competitive Comparison: Grid-Scale ESS vs. Rack-Level Battery Backup
The companies profiled here are not all competing for the same contracts, and understanding the layering helps explain why so many different players can grow simultaneously. LG Energy Solution and Samsung SDI are positioned to win large, multi-year ESS supply agreements with hyperscale operators and utilities, the kind of contract that moves tens of gigawatt-hours at a time. Electrovaya’s containerized format competes in a similar facility-level tier but at a smaller unit scale, suited to operators who want modular deployment rather than a single mega-ESS installation.
QuantumScape and CBAK Energy are playing closer to the server itself. QuantumScape’s PowerBlock targets in-rack storage where solid-state chemistry’s safety and density advantages matter most, while CBAK Energy’s high-discharge-rate LFP cells target the backup-power and UPS tier that has to react in milliseconds, not minutes. None of these companies are positioned as direct substitutes for each other in the way, say, two rival smartphone chipsets compete for the same design win. That is part of why TrendForce describes the ESS architecture for AI data centers as layered rather than monolithic — facility-level BESS, rack-level BBU, and floor-level supercapacitors each have their own supplier ecosystem.
Risks and What Remains Unconfirmed
Several claims circulating around this story deserve a caveat. There is no confirmed, independently verified figure for how much realized revenue AI-data-center battery orders have generated for any of the companies named here in 2026. The TrendForce 216.8 GWh figure is a forecast through 2030, not a confirmed installed base today. QuantumScape’s “four times the power density, five times the runtime” claim for QS PowerBlock is the company’s own stated comparison against unspecified “current data-center industry standards,” not a benchmark run by an independent lab. Readers should treat these as directional indicators of where the market is heading rather than settled facts about current deployment scale.
There is also a broader risk familiar from other AI infrastructure categories: demand forecasts built on planned data-center capacity can shift if AI capital spending slows, if financing conditions tighten, or if power-grid interconnection delays push project timelines further out than currently planned. The same hyperscale capex that is driving battery demand is also the single biggest variable that could soften it, a dependency investors in this sector cannot avoid. The ongoing scrutiny around community and regulatory pushback against new data-center projects, as seen with Google’s India facility, is one example of a friction point that could slow the pace at which planned capacity turns into realized battery orders.
What’s Next: Five Predictions for the AI Battery Market
- Expect Korean battery makers to report ESS-specific revenue as a separate line item in quarterly earnings disclosures by early 2027, as investors push for proof that the ETF rally reflects realized orders rather than sentiment.
- LG Energy Solution’s claim that U.S. ESS shipments could exceed EV-battery shipments in 2026 will likely be tested publicly as the company reports year-end production figures, and could become a template other battery makers point to when repositioning away from automotive customers.
- More EV-focused battery and materials companies, following QuantumScape’s lead, are likely to announce data-center-specific product lines over the next two to three quarters, as the addressable market for AI-adjacent ESS continues to look larger and faster-growing than EV adoption curves.
- Watch for consolidation or partnership announcements between rack-level battery suppliers (like CBAK Energy) and larger data-center infrastructure integrators, since few companies currently offer a full stack from facility-level BESS down to in-rack backup power.
- Expect continued volatility in Korean battery ETFs tied to this theme, since the sector’s valuation currently rests more on TrendForce-style forecasts than on confirmed contract backlogs, making it sensitive to any single disappointing earnings report from Samsung SDI or LG Energy Solution.
Why This Matters Beyond Korea’s Stock Market
The battery story is really a power-infrastructure story, and power infrastructure has become one of the hardest constraints on how fast AI companies can deploy new compute. Every gigawatt-hour of ESS capacity added to a data-center campus is, in effect, a gigawatt-hour of flexibility that lets an operator run GPU clusters harder without tripping grid-stability limits or waiting years for a new substation. That is why battery suppliers, who used to sit several layers removed from AI infrastructure conversations, are now showing up in the same planning meetings as chip vendors and cloud architects.
It also explains why companies like AM Intelligence, which recently ordered 20,000 Nvidia Rubin GPUs for an India buildout, and other large GPU deployments are increasingly discussed alongside power and battery planning rather than as a pure compute procurement story. The GPUs cannot run at full utilization without the power-delivery and backup architecture to support them, and that dependency is exactly what is pulling battery makers into the center of the AI infrastructure conversation rather than keeping them on its periphery.
Firm Power Limits Behind the Battery Demand
Battery demand is easier to understand against the wider power shortage. The International Energy Agency estimates data centers used 485 TWh of electricity worldwide in 2025, up 17% in a single year, and projects roughly 950 TWh by 2030. AI-focused facilities are on track to triple over the same period. In Texas, ERCOT’s large-load queue has swelled to roughly 474 GW, about 90% of it data centers, far more than will ever be built, which is one reason forecasts built on planned capacity deserve caution.
Regulators are also changing the terms of getting connected. AEP Ohio’s tariff requires new data centers above 25 MW to pay for at least 85% of subscribed demand for up to 12 years. In Texas, the PUC replaced a proposed $50,000-per-MW fee with a flat $100,000 study fee, while ERCOT audits its queue. Rules like these make speculative requests more expensive and favor developers with projects they can actually energize.
For operators, the practical lesson is flexibility. One analysis of the firm-power gap advises treating flexibility as an asset: a campus that can curtail 1% of hours may connect years sooner, so training schedulers and contracts should be designed for it from day one. Its broader advice is to build a portfolio, because no single path covers 2026 to 2035, pairing near-term flexibility and storage with a restart or geothermal contract and an option on advanced nuclear. That is where on-site batteries fit: as the bridge while firm clean power, most of which arrives after 2030, is still being built.
Frequently Asked Questions
What caused the October 2026 battery ETF surge in Korea?
According to JKN and the Korea Exchange, eight of Korea’s ten best-performing ETFs in October 2026 were tied to the secondary-battery industry, driven by investor expectations that AI data centers will significantly increase demand for energy storage systems (ESS).
How much is AI-data-center battery demand expected to grow?
TrendForce forecasts that global ESS installed capacity for AI data centers will grow from 15.7 GWh in 2024 to 216.8 GWh in 2030, implying a compound annual growth rate of about 46.1%.
Is this battery demand confirmed by actual orders, or is it speculative?
The TrendForce figures and LG Energy Solution’s capacity targets are forecasts and company-stated plans, not confirmed order backlogs. No public source in this report confirms exact realized battery-order revenue tied specifically to AI data centers for 2026.
What is QuantumScape’s QS PowerBlock?
QS PowerBlock, announced October 8, 2026, is QuantumScape’s first product aimed at data centers. It is an in-rack energy-storage system built on an 800VDC architecture, which the company says offers four times the power density and five times the runtime of current data-center industry standards.
Why do AI data centers need different batteries than traditional data centers?
AI training workloads create rapid, large power-draw fluctuations as GPU clusters execute synchronized computing steps. Traditional backup batteries designed for occasional outages are not built to respond to these millisecond-scale swings, which is driving demand for active, rack-level, and facility-level battery systems.
Which companies are leading the AI-data-center battery push?
Samsung SDI, LG Energy Solution, QuantumScape, Electrovaya, and CBAK Energy have all made public moves into this market, each targeting a different layer of the power-storage stack, from grid-scale ESS to in-rack backup power.
Is LG Energy Solution moving away from EV batteries?
Not entirely, but its stated U.S. ESS capacity target rose from 30 GWh to 50 GWh by the end of 2026, and KB Securities research forecasts its U.S. ESS shipments could exceed its EV-battery shipments starting in 2026, reflecting a significant reallocation toward data-center demand.
What are the risks to this battery demand story?
The forecasts depend on continued AI data-center capital spending. Any slowdown in hyperscaler capex, grid-interconnection delays, or project cancellations — including regulatory pushback seen in some markets — could soften the pace at which planned ESS capacity turns into realized battery-maker revenue.
