How AI Data Centers Redrew the Battery Battlefield — The Three Korean Battery Makers’ ESS Gambit

What’s the story: second-quarter results that put ESS, not EVs, on the table

In the second-quarter results that LG Energy Solution and Samsung SDI both released on July 30, 2026, the center of growth was energy storage systems (ESS), not EVs, and one pillar of that demand is AI data centers. However, according to SNE Research, the two companies’ combined share of the global lithium-ion ESS market in the first half of 2026 was only 4.0%, so the three Korean makers’ shift to ESS should be read as a ‘catch-up race’. This article explains why AI data centers call for batteries, what cards the three companies hold, and where the axis of competition has moved in a market where the original LFP (lithium iron phosphate) patents have expired.

Starting with the industry picture, LG Energy Solution returned to profit after two quarters, with second-quarter revenue of 7.5602 trillion won and operating profit of 113.3 billion won. According to the company, first-half ESS revenue rose to 4.6 times the same period a year earlier, and its share of total revenue climbed to the high 20% range. In the first half alone it secured more than 3 trillion won in new ESS orders, including hyperscaler AI data center projects, and said it would expand North American ESS production capacity to more than 50GWh by year-end.

Samsung SDI posted a profit for the first time in seven quarters, with second-quarter revenue of 3.77 trillion won and operating profit of 204 billion won. On its conference call, the company projected that the US ESS orders it has secured fill a substantial portion of its production capacity through 2029, and that including likely projects in the second half, demand would exceed capacity from 2028. SK On signed a contract in September 2025 with US-based Flatiron Energy Development to supply up to 7.2GWh of LFP ESS through 2030, with shipments scheduled to begin in the second half of 2026.

The tension lies in the gap at the starting line. The three Korean makers are entering late into a market where Chinese companies completed the value chain first, from LFP materials to cells and packs. This article answers three questions: what batteries, and how many, does a data center call for; how do the three companies’ strategies differ; and what are the weapons in this catch-up race from a patent perspective.

Basic principle: three layers of battery demand created by a single data center

Training and inference for generative AI consume power in large and irregular amounts. As grid expansion fails to keep pace with the growth of data centers, grid-connection delays have become routine, and load fluctuations have grown. Batteries fill this gap.

By analogy, batteries in a data center resemble a building’s water supply system. The large ESS that stores and supplies its own power behind the meter is the rooftop water tank, the uninterruptible power supply (UPS) that bridges the moment of an outage is the emergency tank on each floor, and the battery backup unit (BBU) at the server-rack level is the water bottle on the desk. For the tank, capacity and cost matter; for the bottle, instantaneous output and volume matter. Because the requirements differ, the chemistries that go in differ too.

According to the International Energy Agency (IEA) Global Energy Review 2026, 108GW of battery storage was newly installed worldwide in 2025, up 40% from the previous year, and about 90% of new installations were LFP. The same source counted that new capacity of battery-based UPS, used mainly in data centers, rose 30% to 45GW. In large ESS, safety, lifespan and economics take priority over weight and energy density, which is why the three Korean makers, which built their production networks around ternary (NCM) chemistry, are all rushing to shift to LFP.

Here we add one insight not in the original. The equation ‘battery for AI data centers = LFP’ holds only at the water-tank layer. Looking just at the products Samsung SDI unveiled at InterBattery 2026, its UPS battery for AI data centers (U8A1) is a prismatic cell with an LMO (lithium manganese oxide) cathode, and its cylindrical battery for BBUs uses a high-nickel NCA cathode. In other words, for the three Korean makers the data center is a dual structure: a market where they compete head-on with China in LFP overlaps with a market where their existing ternary and high-power technologies still work.

Comparison: the three companies’ ESS strategies and market positions

Item	LG Energy Solution	Samsung SDI	SK On
Key card	North American local production + integrated solutions	Prismatic LFP and high-power (UPS/BBU) technology	US LFP ESS production using part of its EV lines
Recent results	H1 ESS revenue 4.6x, new orders of more than 3 trillion won	Substantial share of capacity through 2029 secured with US ESS orders	Flatiron contract for up to 7.2GWh (through 2030)
Production setup	Target of more than 50GWh North American ESS capacity by year-end, prismatic lines added in 2027	Preparing prismatic LFP mass production in the US	SKBA plant in Georgia, US, shipments starting H2 2026
H1 2026 ESS share	2.6% (12.0GWh)	1.4% (6.4GWh)	Tally not confirmed
Next generation	Sodium-ion sample supply planned for H1 2027	Developing sodium-ion products	Developing sodium-ion (per the original)

What this table says: All three companies put local US production at the center of their ESS strategies, but first-half shares were only 2.6% for LG Energy Solution and 1.4% for Samsung SDI. Order backlogs are building quickly, while their presence in terms of shipments is still small.

Item	Shipments (GWh)	Share (%)	Year on year
Global lithium-ion ESS total	461.3	100	+71%
CATL	125.0	27.1	–
Top three Chinese makers (CATL, EVE Energy, Hithium) combined	–	47.5	–
LG Energy Solution	12.0	2.6	+357%
Samsung SDI	6.4	1.4	+20%

What this table says: The combined share of the two Korean makers (4.0%) is about one-tenth of the top three Chinese makers (47.5%). However, LG Energy Solution’s shipments grew 357% year on year, closing the gap at a pace far above market growth (71%).

Patent perspective: with the original LFP patents expired, where has the axis of competition moved?

The original article had no patent discussion, but the rights landscape of the LFP market is key to understanding the three Korean makers’ strategies. US5910382A (title “Cathode materials for secondary (rechargeable) lithium batteries”, applicant University of Texas System, current owner Hydro-Québec), regarded as the original patent for LFP cathode material, has a priority date of April 23, 1996, a filing date of April 21, 1997 and a grant date of June 8, 1999, and is shown on Google Patents as having expired at the end of its term on April 21, 2017 (Expired – Lifetime). Claim 1 claims a battery including a cathode of LiMPO₄ composition (where M is one or more first-row transition metals).

US6855273B2 (“Electrode materials with high surface conductivity”, owners including Hydro-Québec, CNRS and Université de Montréal), a representative patent for the carbon-coating technology that compensates for LFP’s low electronic conductivity, has a priority date of April 30, 1999 and a grant date of February 15, 2005, and is shown on Google Patents as expired on May 4, 2020. Claim 1 claims an electrode material in which a conductive carbon material is uniformly deposited on the surface of complex oxide particles.

• Meaning of the scope of rights: With the expiry of these two original patents, the risk of infringement for the basic LFP composition and carbon coating itself has fallen sharply. The licensing barrier of the original patents to the three Korean makers’ shift to LFP has effectively disappeared.
• Insight: As the entry barrier has fallen, differentiation has also become harder. The axis of competition can be seen to have moved from composition patents to improvement patents on cell form factor (prismatic, pouch), cell-to-pack structure, thermal-propagation prevention and fire-response design, and manufacturing processes, and to a non-patent barrier of ‘origin’ created by US non-China supply-chain rules (PFE, prohibited foreign entity).
• FTO points: The expiry of the original patents does not by itself mean freedom to operate. Follow-on patents on improved compositions such as particle size, doping and coating thickness, prismatic LFP cell structures, and fire-suppression designs for ESS containers are subjects that each company must separately put through FTO (freedom-to-operate) review.
• Design-around scenario (counterargument): Conversely, Chinese makers may hold large numbers of improvement patents, so if the three Korean makers’ US LFP products fall within the scope of a particular improvement patent, the advantage gained through ‘origin’ may be offset by patent-dispute risk. In that case a realistic design-around would vary process and structural limitations rather than composition.

This article was unable to verify individual patent numbers corresponding to each of the three companies’ LFP cell structures for ESS or their UPS and BBU products for data centers. As they cannot be confirmed from published patent information, the next step is to track published applications in the three companies’ names using keywords such as prismatic LFP cells, ESS container fire-suppression structures and high-power cylindrical BBU cells.

Limits and outlook

Amid the EV chasm (the demand stall just before mass adoption), ESS is becoming a growth axis beyond a buffer for the three companies’ earnings. The optimistic outlook is that AI data centers and US non-China supply-chain rules combine to fill the three Korean makers’ North American LFP lines quickly, converting order backlogs into shipment share.

There is also an opposite scenario. Because non-China demand is demand created by regulation, it is sensitive to policy change, and if Chinese makers bypass the rules through third-country production or cut prices further, the Korean makers’ LFP may be left with only a cost disadvantage. It is also hard to rule out a situation in which the conversion investment that turned EV lines to ESS becomes a bottleneck again when EV demand recovers.

The strategic alternative is to separate and target the three layers of data center demand. Large ESS inevitably involves LFP cost competition, but layers such as UPS and BBU, where power density and volume matter, are areas where technologies and improvement patents Korean makers already hold, such as ternary, LMO and high-power cylindrical cells, still work. Whether the sodium-ion batteries the three companies are developing can become LFP’s next card in the UPS market is also likely to be decided at this layer.

Practical action: Staff planning ESS and UPS businesses for data centers should divide their products into the three layers of large ESS, UPS and BBU and tabulate the chemistry applied to each, and for LFP products should first draw up an FTO review list of improvement patents on particles, coatings and cell structures, separately from the expiry of the original patents.

Items requiring verification

• The primary source for the ’20GWh of additional discussions’ between SK On and Flatiron
• Official company materials on the status of SK On’s sodium-ion battery development
• The mass-production timing of LG Energy Solution’s sodium-ion battery for UPS (the earnings materials confirm only a plan to supply samples in H1 2027)
• SK On’s global ESS shipments and share in H1 2026
• The status of granted patents in the three Korean makers’ names on prismatic LFP and ESS fire-suppression structures

Sources

1. LG Energy Solution Battery Inside, LG Energy Solution Q2 2026 earnings release (2026-07-30)
2. Money Today, LG Energy Solution Q2 operating profit 113.3 billion won… “ESS up 4.6 times year on year” (2026-07-30)
3. Samsung SDI Newsroom, Samsung SDI announces Q2 2026 results (2026-07-30)
4. Dailian, [Conference call] Samsung SDI “US ESS orders expected to exceed capacity from 2028” (2026-07-30)
5. Samsung SDI Newsroom, Samsung SDI Unveils All-Solid-State Battery for Physical AI (InterBattery 2026 exhibit introduction, 2026-03)
6. Energy-Storage.News, SK On signs 7.2GWh deal to supply US-made LFP batteries to Flatiron Energy Development (2025-09-05)
7. IEA, Global Energy Review 2026 – Technology: Battery storage
8. Autodaily, Global ESS grows 71%, LG Energy Solution shipments up 357% (SNE Research data, 2026-08-04)
9. Google Patents, US5910382A – Cathode materials for secondary (rechargeable) lithium batteries
10. Google Patents, US6855273B2 – Electrode materials with high surface conductivity