Vitzrocell (082920) vs Musashi (7220) – EDLC vs Hybrid SuperCapacitor

One line: Both are optionality names, not pure-plays, where a sub-5%-of-revenue supercapacitor line is being asked to carry an AI-datacenter re-rating. On the technology, Musashi's hybrid cell is the better fit for the datacenter buffering job, and Musashi already ships a datacenter product where Vitzrocell does not. On the stock, the sum-of-parts re-rating that briefly tripled Musashi does not transfer cleanly to Vitzrocell, because the mechanism that drove it is absent here.

Built 2026-06-29 from a supercapacitor technology primer, the Vitzrocell deep-dive and Musashi grounding notes, and the Collyer Bridge / Made in Japan Substack sweep. Companion to 082920/082920, 7220/7220, and the frequency-domain framing in _compare/6981-vs-7220-pdn-stack.

The two devices are different chemistries, not two versions of the same thing

Vitzrocell sells an electric double-layer capacitor (EDLC). Both electrodes are activated carbon and all the charge is held electrostatically on the carbon surface, with no bulk chemical reaction. Musashi sells a lithium-ion capacitor it brands Hybrid SuperCapacitor (HSC). It is asymmetric: one activated-carbon double-layer electrode plus one pre-lithiated carbon electrode that stores charge faradaically, like a lithium-ion anode. That single design choice, making one electrode battery-like, is the whole story. It buys Musashi roughly three to five times the energy density and a higher cell voltage, at the cost of some power, weaker cold tolerance, and a swelling problem the EDLC does not have.

The "up to 10x energy density of standard EDLC" line on Musashi's site is a best-case volumetric figure against a low-end EDLC baseline. The defensible like-for-like multiple is three to five times, and the one Musashi-specific volumetric number anyone has surfaced is roughly 20 Wh/L, only modestly above Skeleton's 16 Wh/L (single secondary source, and Skeleton is primarily an EDLC/graphene maker, so the figure may not be a clean like-for-like LIC comparator). Treat the energy-density edge as real but incremental, not the order-of-magnitude lead the marketing implies.

What the differences mean for AI-datacenter buffering

The datacenter job is rack-level peak shaving. GPU clusters swing power on a millisecond timescale, and a supercapacitor sits on the intermediate rail and acts as a low-pass filter between chaotic GPU demand and a smooth upstream draw. Stated precisely, the Flex and Musashi Capacitive Energy Storage System (CESS) buffers the roughly 48V rail (45.5–51.5V, about 16kW, plus or minus 320A per tray) inside an 800VDC-fed rack. It is not an 800V device; 800VDC is NVIDIA's upstream distribution architecture, and conflating the two overstates the spec.

The constraints that bind in that job favour the hybrid cell. Rack space is scarce and revenue-bearing, so volumetric energy density, not weight, is the limiting metric, and the HSC packs more energy per litre. Its higher 3.8V cell voltage means fewer cells in series for a given bus. Cycle life and no-thermal-runaway safety matter to both devices and both clear the bar. The one place the EDLC clearly wins, operation down to −40°C, delivers no value in a climate-controlled hall the CESS itself rates to only +45°C. That is why Musashi, not a pure-EDLC vendor, landed the Flex design win.

Here is the wrinkle that the headline "EDLC vs HSC" framing buries, and it cuts against Vitzrocell. The product Vitzrocell ships today is the EDLC, which is the weaker device for this specific job. Its datacenter ambition rests on a lithium-ion capacitor it has not yet commercialized; the company's own LIC page says only "being prepared." So in the datacenter ring the honest matchup is Musashi's HSC, in delayed production, against Vitzrocell's LIC, still in development. On the device that actually wins datacenter sockets, Vitzrocell is behind, not ahead.

There is one technical point that runs the other way, and it is worth flagging with its low confidence. LIC anodes are pre-doped with lithium and swell over repeated cycling; a cylindrical can distributes that pressure evenly, while a prismatic cell forms uneven pockets at the corners. Musashi uses prismatic cells, chosen for packing density, and was reportedly asked to supply cylindrical samples, which the Collyer chat reads as corroboration of a swelling concern. Vitzrocell's EDLC heritage is cylindrical, so if it builds a cylindrical LIC it could hold a reliability advantage in form factor. This is a subscriber hypothesis, not a proven edge, and it depends entirely on a Vitzrocell LIC that does not yet exist.

Where each device actually sells today

Vitzrocell has no datacenter revenue. Its EDLC is roughly 5% of company revenue, and that bucket is shared with other new products, so the pure EDLC line is likely low single digits. The EDLC sells into smart metering and AMI backup, industrial IoT, automotive event recorders and stop-start, grid and renewables ride-through, and defense. The business engine underneath it is something else entirely: lithium thionyl chloride primary cells at roughly 60% of revenue, oil-and-gas high-temperature downhole batteries near 20%, and defense ampoule and thermal batteries near 15%. That core is genuinely strong, with a 21% revenue and 41% operating-income four-year CAGR, a 28.5% operating margin, net cash near 137 billion KRW, and strong free cash flow. The AI angle is a confirmed but discussion-stage talk with Taiwan's Delta to supply LIC for US datacenters, with a decision expected within 2026, zero booked datacenter revenue, and a position behind Korean peers LS Materials (Vertiv MOU, September 2024) and Vinatech (Bloom Energy, May 2025).

Musashi's HSC is the entire investment narrative, and it sells into exactly the target market, but at negligible volume. The HSC is less than 1–2% of consolidated revenue and is not separately disclosed; the other roughly 99% is a flat, low-margin, Honda-concentrated, BEV-exposed auto-parts business. The commercialization scaffolding is real: the Flex CESS has been in production since H1 2025, there is co-development with Mitsubishi Electric on power quality and an adoption by Mitsubishi Heavy Industries for its Prismo railway system, plus UL1973 and UL certification. No hyperscaler design win is named, and the Delta-as-Musashi-customer link, unlike Vitzrocell's confirmed Delta talks, is asserted in one analyst note and could not be independently verified. The cell ramp is the central problem. Production runs at roughly 200,000 cells in 2026 against a 6.5 million nameplate, and the new Minami-Alps plant start-up has slipped to next year, with the Hokuto plant running 24 hours to bridge. The FY03/26 guidance was cut to less than half the originally expected HSC revenue and a slight loss, a roughly 4 billion yen drag on group profit, even as the subsidiary's tangible fixed assets jumped from 1.9 to 13.8 billion yen, so heavy capex is already committed against demand that has not yet shown up.

So the two have mirror-image profiles. Vitzrocell has a real, profitable, diversified business and an embryonic datacenter option. Musashi has a weak core and a more advanced but still pre-volume and repeatedly delayed datacenter product. On the supercap-for-datacenters axis specifically, Musashi is ahead; on business quality, Vitzrocell is ahead.

The re-rating, and whether it transfers

What lifted Musashi was a sum-of-parts argument. On blended math the stock cannot clear about 12 times, because a high-growth HSC stub is buried inside a low-multiple auto-parts shell. Nomura's response, as relayed through the Collyer chat at medium confidence, was to switch to sum-of-parts and lift the fair multiple from 12 to 22.5 times, for a roughly 7,700 yen target, a figure live consensus median now matches. Made in Japan's version put the HSC segment at around 255 billion yen at full ramp, against a then-186 billion yen total market cap. The stock ran about +172% from Made in Japan's 4 April publication base, around 2,838 yen implied by the then-186 billion yen market cap, to roughly 7,700 yen by late May, then on to a 9,740 yen close and 10,550 yen intraday peak on 3 June, a larger move of +240% to +270% off that base, before round-tripping to 3,990 yen by 29 June, back to roughly its April close and still about 80% above the 2,210 yen trough. The re-rating priced a subsidiary that is under 2% of revenue, is not separately disclosed, and is behind on its ramp, as if it were already a standalone growth company. The tape then rejected it.

Can Vitzrocell re-rate the same way? Two conditions have to hold for the transfer: the supercap line must be material or credibly about to be, and it must be exposed to the same datacenter demand. Neither holds today. The EDLC is immaterial, the LIC is not yet shipping, and the datacenter exposure is a discussion-stage talk that trails two domestic peers. On those two tests Vitzrocell scores no better than Musashi, and on product readiness it scores worse.

The deeper reason the thesis does not transfer is mechanical. Musashi's re-rating worked because sum-of-parts surfaces value when a good stub is trapped inside a bad shell; separating them is the entire trade. Vitzrocell has no bad shell to separate. Its core is excellent and the market already prices the whole company at roughly 29 times trailing earnings and five times book, having already tripled in 2026 on the AI narrative as a single re-rating of the whole business. There is no low-multiple wrapper left to carve away. The Musashi playbook is a sum-of-parts separation; the Vitzrocell situation is already a fully-rated growth company carrying an AI option that the run has substantially priced in, now sitting about 42% below its May high after profit-taking.

The market rotation captured in the chat reads this half right. The Collyer view turned cautious on Musashi, calling it a bad setup and noting it is down from 10,000 yen, and is now eyeing Vitzrocell as the preferred vehicle while waiting for a deeper pullback. Made in Japan, the original Musashi bull, never bought and stayed on the sidelines, citing unset pricing, an unsettled depreciation schedule, the slipped ramp, and possible advance capex. The instinct that Vitzrocell carries less execution risk is correct about the core business and backwards about the datacenter thesis. On the AI play specifically, Vitzrocell has more execution risk, not less, because Musashi at least ships an HSC and has Flex CESS in production, while Vitzrocell's relevant product is still in preparation and its datacenter relationship is a conversation, not a contract.

For Vitzrocell to re-rate further on the AI angle rather than on its already-strong core, several things would have to become true together. The Delta discussion would have to convert into a signed contract with disclosed size. The "in preparation" LIC would have to ship and qualify on energy density against incumbents, ideally in the cylindrical form factor where its heritage might help. Vitzrocell would have to disclose a datacenter capacity build, and in time book revenue. Until those land, the datacenter piece is a free option already substantially in the price after a triple, not a fresh catalyst, and the cleaner supercap exposure to the same demand wave still runs through the LIC class generally rather than through Vitzrocell's EDLC.

Bottom line

The supercapacitor-for-datacenters thesis is intact at the technology and architecture level, reinforced by the OpenAI, Microsoft and NVIDIA power-stabilization work and, per Made in Japan, NVIDIA's Vera Rubin capacitor adoption, and the hybrid cell is the right device for the job. But neither stock is a clean way to own it. Musashi has the better datacenter product and a far worse business, and its narrative-driven re-rating has already round-tripped on a delayed ramp the bulls themselves will not underwrite. Vitzrocell has the better business and a weaker, less-advanced datacenter position, and the sum-of-parts mechanism that lifted Musashi does not apply to it because there is no low-multiple shell to separate out. Owning Vitzrocell is owning a high-quality lithium-primary, defense and oil-and-gas compounder with a richly-priced AI option attached, not a supercapacitor pure-play. Owning Musashi is owning an auto-parts shell with a real but pre-scale HSC lottery ticket. The two names are not the same trade, and the Musashi re-rating is not a template for Vitzrocell.

Confidence and caveats

• Vitzrocell trades as 082920.KQ (the .KS suffix is a yfinance fund proxy, mislabeled a mutual fund); a spurious 2:1 split corrupts pre-2025 per-share data on both suffixes, so use the income statement. Last price 36,950 KRW (2026-06-26 close, the basis of this note); live 082920.KQ prints ~40,550 KRW on 2026-07-01, market cap about 1.82 trillion KRW or US$1.18B. Full like-for-like peer set: _compare/082920-vs-126340-vs-417200-kr-supercap.
• Musashi last close 3,990 yen (2026-06-29), market cap about 261.5 billion yen. Forward P/E 22.46x happens to land on Nomura's 22.5x blended fair multiple, a coincidence struck on a rebounding EPS base rather than the same construct; trailing P/E of about 208x is distorted by the depressed earnings base.
• The Nomura 12x-to-22.5x and 7,700 yen figures come from a subscriber chat repost, not a primary Nomura document; the target is corroborated by live consensus median, the multiple by live forward P/E.
• Musashi's ~20 Wh/L volumetric figure rests on a single secondary source; the company publishes no absolute Wh/L or Wh/kg. The Skeleton 16 Wh/L comparator likewise traces to a single secondary source, and Skeleton is primarily an EDLC/graphene maker, so it may not be a clean like-for-like LIC benchmark. The "10x EDLC" claim is best-case marketing.
• The cylindrical-versus-prismatic swelling advantage for a future Vitzrocell LIC is a subscriber hypothesis from the Collyer chat, not a proven or shipping edge.
• Delta appears on both sides with different evidentiary weight: Vitzrocell's Delta talks are CEO-confirmed and dated; Musashi's Delta-as-customer link is analyst-asserted and unverified.