Korean Supercapacitor Showdown: Vitzrocell (082920) vs Vinatech (126340) vs LS Materials (417200)

The like-for-like set. Same country, same device family, same AI-datacenter narrative pulling the tape. Prices as of 2026-07-01 for all three (live yfinance): Vinatech and LS Materials from their .KQ listings, and Vitzrocell from 082920.KQ, whose 40,550 close gives a clean live print (the .KS suffix is a yfinance fund proxy, see caveats). Built from the three source dossiers, the Vitzrocell deep-dive, and the Musashi head-to-head it supersedes as the real peer comparison.

The one-paragraph answer

There is no clean way to own Korean supercapacitors for AI datacenters, and the reason is that these three are not the same device, not aimed at the same part of the datacenter power stack, and not the same kind of risk. Vinatech (126340.KQ) is the only true supercap pure-play and the only one of the three that has actually booked a datacenter-adjacent order. You pay for that purity with a loss-making, levered, single-customer balance sheet that probably needs another raise before its 2028 capacity build lands. Vitzrocell is the best business by a wide distance – roughly 28% operating margins, net cash near 8% of its market value, compounding around 21% a year – but the supercapacitor is about 5% of it and the datacenter-relevant product is still "in preparation," so you are buying a specialty-battery compounder with an AI option stapled on, not a supercap bet. LS Materials (417200.KQ) is three-quarters commodity aluminum with a minority ultracapacitor slice, its datacenter angle is an MOU that has not converted, and its flagship rack product is really a grid-stabilization device, yet it trades on a full growth multiple. So "best" is a question about what you actually want. If you want the thesis, it is Vinatech. If you want the business, it is Vitzrocell. LS Materials offers a worse version of both, and for this specific theme it is the one to leave on the shelf.

They are not even the same device

The lazy version of this trade treats "Korean supercap for AI" as one bucket. It is three different chemistries.

Vitzrocell ships an electric double-layer capacitor (EDLC) – both electrodes activated carbon, all charge held electrostatically, no bulk reaction. Huge power, half-a-million-plus cycles, extreme cold tolerance, but low energy per litre. Its lithium-ion capacitor, the energy-denser device the datacenter job actually wants, is still on the roadmap; the company's own page lists it as being prepared.

Vinatech is the opposite. It is a supercapacitor pure-play branded Hy-Cap – supercapacitors (EDLC, hybrid, and its flagship 3.8V LIC/VPC line) are more than 90% of revenue, and the LIC is the device the datacenter rack job actually wants. One activated-carbon electrode plus one pre-lithiated faradaic electrode buys the LIC its higher cell voltage and the three-to-five-times energy density that a space-constrained rack rewards. Vinatech is the only one of the three whose flagship is the LIC, and the only one with a booked datacenter-adjacent order.

LS Materials is a carbon ultracapacitor (EDLC-class) maker, same device family as Vitzrocell, wrapped around a much larger aluminum-materials business. Its hybrid LIC exists only as a KEPCO development prototype. So on the device that wins the AI-rack socket, LS Materials is in the same "wrong tool today, right tool in the lab" position as Vitzrocell, with the added complication that the supercap is only about a quarter of what it sells.

Read the energy-density row against the device row and the picture sharpens. Volumetric energy per litre is the metric that binds in a rack, because rack space is scarce and revenue-bearing. Vinatech's LIC sits at roughly 45–61 Wh/L usable; the two EDLC makers sit in the single digits to low teens. On the physics of the canonical rack-buffering job, Vinatech has the right device and the other two have the device they hope to replace.

And they are not aiming at the same socket

This is the point the peer framing usually misses, and it matters more than the device row. Even inside "AI datacenter," the three are selling into different layers of the power stack.

Vinatech's Bloom Energy order is for supercapacitor systems that go into Bloom's solid-oxide fuel-cell units – the on-site generation and backup layer that powers data centers, sitting downstream of Bloom's own deals with Oracle/OCI, Equinix, and AEP. Real datacenter demand, booked and dated, but at the fuel-cell layer rather than inside the server rack. Vinatech rides Bloom's beta, and Bloom is itself a volatile AI-power thematic name.

Vitzrocell's Delta talk is aimed at the actual rack-buffer job – LIC on the roughly 48V rail inside an 800V-fed rack, absorbing the sub-second GPU load swings that threaten voltage and frequency stability. That is the canonical supercap-for-AI thesis, the same socket Musashi's Hybrid SuperCapacitor targets through the Flex CESS tray. It is also the socket Vitzrocell is least equipped to fill today, because its shipping product is the EDLC and the LIC that would win it is unbuilt.

LS Materials' Vertiv MOU points at datacenter UPS, and its flagship LS UltraGrid is a MW-level, sub-0.1-second power-bridging system whose primary market is grid and renewable stabilization, with datacenter UPS an explicit secondary. "Rack" there is a form factor, not evidence of server-rack peak-shaving. For that grid-and-UPS duty, where power density and cycle life dominate and energy density is secondary, the EDLC is genuinely well-suited – but it is a different job from the one the theme is priced on.

So the honest map is three companies aiming at three different sockets. Vinatech sits at the fuel-cell power layer, and it is booked. Vitzrocell is chasing the rack buffer, on talks and with the wrong current device. LS Materials is at grid and UPS stabilization, on an MOU and an EDLC built for a grid-first product. Only Vitzrocell is targeting the exact job the AI-supercap narrative describes, and it is the least ready to do it.

The purity-versus-quality axis

Line the three up and they trace a single trade-off. Purity of supercap exposure runs one way; quality of the underlying business runs the other; and no name gives you both.

Vinatech is the purest exposure available in this set, full stop. Ninety-plus percent supercapacitor, the right device, a booked order that equals half of last year's revenue, and a credible up-the-value-chain move from cells to full systems. If the supercap-for-datacenters thesis is right, this is the cleanest way to express it. The price of that clean expression is a genuinely stretched financial profile. Vinatech lost money in FY25, swung back to an operating loss in Q1'26 even as revenue grew 61%, burned about KRW 25B of free cash, carries debt-to-equity around 131%, and has a roughly KRW 41B convertible sitting as a dilution overhang. Heavy Hung Yen (2028) and India capex is still ahead of it, so another raise is the base case. And the whole thesis rests on one named customer. The Bloom concentration is simultaneously the reason to own it and the reason it can halve. On EV/sales it is also the most expensive of the three at roughly 8–10x, though that multiple has compressed hard from the 14x it carried in late May, because the stock fell faster than the story before the order arrived.

Vitzrocell is the mirror image. It is the highest-quality business on this page by a distance – a lithium-primary, defense, and oil-and-gas franchise compounding revenue around 21% a year with operating margins near 28.5%, roughly KRW 137B of net cash, and strong free cash flow. It is the only one of the three earning a genuine operating profit – Vinatech ran a FY25 net loss, and LS Materials' slim net profit rests on a non-operating credit sitting over an operating loss. But it is the least supercap-pure name here. The EDLC-plus-other bucket is about 5% of revenue, pure EDLC is lower, the datacenter-relevant LIC is not built, and the Delta relationship is a conversation rather than a contract. Owning Vitzrocell for the supercap story means paying compounder-plus-option prices (around 32x trailing, roughly 5.5x book, after a 2026 triple that has since given back roughly a third) for a low-single-digit slice of the business and a call on a product that does not yet ship. The business is worth owning; the supercap is upside, not the thesis.

LS Materials is the awkward middle that turns out to be the weakest corner for this specific theme. On paper it splits the difference – 26% ultracapacitor is purer than Vitzrocell's 5% but nowhere near Vinatech's 90%, and its balance sheet is clean (roughly net cash, low leverage), which puts it ahead of Vinatech on financial risk. But look at what the multiple is actually paying for. This is a business that is about 74% commodity aluminum, runs a 7.9% consolidated gross margin, is currently posting an operating loss, ships an EDLC rather than the LIC the rack wants, has a datacenter relationship that never got past MOU, and whose flagship product is really a grid-stabilization system. For all that, it trades at roughly 6.3x EV/sales and 106x EV/EBITDA – a full growth multiple. You are paying a growth-story price for the smallest, least-datacenter-proven, wrong-device supercap exposure of the three, diluted inside a commodity-materials shell. The one thing it has over Vinatech is that it will not need to dilute you to survive. The one thing it has over Vitzrocell is a slightly less token supercap share. Neither is enough to make it the right vehicle when the pure-play and the quality name both exist beside it.

Datacenter readiness, told straight

Strip the narrative and rank the three by what is actually booked versus what is hoped.

Vinatech is the only one with real, dated, disclosed datacenter-adjacent revenue: a KRW 41.2B Bloom Energy system order, about half of FY25 revenue, recognized from Q4'26 into 2027. High readiness, at the fuel-cell layer, with single-customer concentration as the standing risk.

Vitzrocell has the most thesis-aligned target and the least product. The Delta talk aims squarely at the rack-buffer job, the CEO has confirmed it and dated a decision "within 2026," and the company has the strongest balance sheet of the three to fund the LIC build. But there is no contract, no disclosed capacity, and the shipping product is the wrong device. Medium readiness on ambition, low on delivery.

LS Materials has the weakest datacenter position of the three despite being the name that has carried the "AI-datacenter ultracapacitor" label the longest in Korean press. The Vertiv MOU has produced no order, the UltraGrid is grid-first, and the datacenter piece is optionality layered on a grid-and-industrial UC business. Its most concrete non-legacy wins are a KEPCO grid project and a US nuclear-fusion demonstration supply – credibility markers, not datacenter revenue. Lowest readiness relative to a full growth multiple.

The honest ranking, and who each is for

For the specific question – the best risk-adjusted way to own the Korean supercap-for-AI theme – the answer refuses to collapse to one name, because two of them are genuinely different propositions and the third is dominated.

If you are buying the thesis, it is Vinatech. It is the only pure-play, it ships the right device, and it now has the only booked datacenter order in the set. Own it if you want actual supercap-for-datacenters exposure and can stomach a loss-making, levered, single-customer name that will likely raise capital again. This is the high-beta, high-conviction expression: right if the theme is right, painful if Bloom stumbles or the raise comes at a bad price.

If you are buying the business, it is Vitzrocell. It is the best company on the page and the only one earning a real operating profit, with real downside protection in the net cash and the defense/oil cash engine. Own it if you want quality with a cheap-ish embedded call on AI power, and if you are honest with yourself that the supercap is 5% of the story, not the story. The knock is that after the 2026 triple the option is already substantially in the price, so the disciplined entry is a deeper pullback that hands the option back closer to free.

LS Materials is the pass for this theme. Not because the company is bad – the balance sheet is fine and the UC has genuine grid and industrial demand – but because for this specific bet it is beaten on both axes. It is less pure than Vinatech, lower quality and less profitable than Vitzrocell, ships the wrong device for the rack, has the weakest datacenter traction, and costs a growth multiple anyway. The only book where it earns a place is one that specifically wants grid and renewable UC stabilization exposure, where the aluminum majority and the datacenter narrative are both beside the point.

If the goal is diversified exposure to the theme rather than a single bet, the cleaner pair is Vinatech (thesis-pure, high-beta) alongside Vitzrocell (quality, downside-protected). They express the same tailwind at opposite ends of the risk spectrum and do not overlap. Adding LS Materials buys a worse, more diluted version of the ultracapacitor exposure Vinatech already gives you cleaner.

Where Musashi fits, and why it is not in this table

The earlier head-to-head pitted Vitzrocell against Musashi (7220.T), and that comparison is a narrative cross-reference, not a peer set. Musashi is a Japanese auto-driveline maker – roughly 99% Honda-tied transmission and gear business, flat, thin-margined, BEV-exposed – with a small Hybrid SuperCapacitor subsidiary bolted on. It belongs in the technology conversation because its HSC is a lithium-ion capacitor, the same device class as Vinatech's Hy-Cap, and because it is furthest along on the exact rack socket Vitzrocell is chasing, shipping into the Flex CESS tray with UL certification. But it fails the like-for-like test on every other axis. Different country, different core business, and a supercap that is under 2% of revenue inside an unrelated auto shell.

Musashi is also the cautionary tape. Its 2026 sum-of-parts re-rating tripled the stock on a subsidiary that is pre-scale, undisclosed, and behind on a repeatedly delayed cell ramp, then round-tripped almost entirely. That mechanism – surfacing a trapped growth stub inside a low-multiple shell – is the closest structural analogue to LS Materials, where a UC call-option sits inside a commodity-aluminum wrapper. It does not map to Vinatech (no shell to separate; it is the pure-play) or to Vitzrocell (already a fully-rated whole-company growth stock). The useful read-across is the warning: a sub-scale supercap line asked to carry a whole-company re-rating is exactly the setup that round-trips. Two of these three names – LS Materials most of all, Vitzrocell to a lesser degree – carry a version of that risk. Vinatech does not, but it swaps it for the more direct risks of losing money, carrying debt, and leaning on one customer.

Confidence and caveats

• Prices are live 2026-07-01 yfinance pulls for all three, Vitzrocell now included via 082920.KQ (40,550 close). The earlier reliance on the 2026-06-26 figure is resolved; a clean live print is available, so the market cap and multiples are current rather than a few sessions stale. Re-pull before acting.
• USD conversions use ~1,542 KRW/USD, the dossier rate as of late June. Live KRW=X on 2026-07-01 is 1,557, about 1% weaker, so every USD figure here runs roughly 1% high. It changes no conclusion, and it does not rescue the old LS Materials US$0.79B print, which only reconciled at an implausible ~1,382 rate.
• Vinatech's 107,700 is a locked +29.9% limit-up print on thin volume, not a settled close. Valuation is shown at both the limit-up and the 30 Jun close (82,900) for that reason. Its valuation multiples were computed from statements because yfinance did not return EV/PE this run; net debt is FY25 year-end, pre the March-2026 convertible, so current net debt is likely modestly higher.
• LS Materials' revenue split is FY2023, the latest clean segment breakdown; the UC share is probably rising with the AIDC push but is unconfirmed for 2024–25. Its EV/EBITDA of 106x is directional and sensitive to a depressed EBITDA base, and its exact FY25 profit figures are secondary-press aggregations of the DART filing.
• Ticker discipline is load-bearing here. Vinatech is 126340.KQ; 126340.KS returns a mutual fund and 006440.KS is a different company (Hanil E&C). LS Materials trades on 417200.KQ. Vitzrocell trades as 082920.KQ: the .KS suffix is a yfinance fund-proxy (quoteType MUTUALFUND, null marketCap, stale regularMarketPrice 19,180), while 082920.KQ returns clean EQUITY data (marketCap, shares, live price). The phantom 2:1 split still corrupts pre-2025 per-share fields on both suffixes, so keep using the income statement, not the per-share fields.
• The energy-density comparison is not perfectly like-for-like. Vinatech publishes datasheet Wh/L; Vitzrocell publishes no farad or Wh figures (EDLC-class ranges are used); LS Materials' ~5.7 Wh/kg is a single flagship cell. The direction – LIC well above EDLC on volumetric energy – is solid; the exact spreads are not.
• SemiAnalysis validates the rack-level supercap/LIC use case under the 800V DC architecture but does not name Vinatech, and surfaces no Korean supercap winner (no dossier reports it naming Vitzrocell or LS Materials either). That silence is itself signal: the structural demand is real, the equity winners are unsettled, and being the priced-in "Korean supercap AI play" is not the same as being a named beneficiary.