Valuation on Korea Invest Insights

SK Telecom Re-Rating: From Dividend Stock to AI Infrastructure

Sat, 25 Apr 2026 22:30:00 +0900

TL;DR

The substance of the SK Telecom re-rating is not “dividend normalization.” It is the recombination of telco core cash flow + Anthropic stake + AIDC/GPUaaS + Edge AI / AI-RAN / Network API options.
At spot ₩100,000, dividend normalization and the AI narrative are largely priced in. Further upside requires AIDC revenue growth, an Anthropic mark-up, and monetization evidence from AI-RAN / Edge AI.
Factor read: price support, dividend normalization, and AIDC growth confirmation matter more than a simple buy/hold label. Below ₩90,000 the dividend-yield floor improves; after 1Q / 2Q confirmation, the AI infrastructure option becomes easier to underwrite.

1. Core Investment Call

Verdict: Wait / Buy on pullback.

SK Telecom is no longer a simple telco dividend name. After 2025 earnings and dividends were impaired by a hacking incident and one-off costs, 2026 layered five things at once: core-business normalization, dividend recovery, AIDC growth, the Anthropic equity option, and the AI-RAN / 6G narrative. 2025 consolidated revenue was ₩17.0992T, operating profit ₩1.0732T, net income ₩375.1B — down -4.7% / -41.1% / -73.0% YoY — but AIDC revenue grew ~35%.

Spot is ₩100,000 (April 24, 2026 close). The dislocation off the early-/mid-March lows is largely closed; the move into ₩100,000s during April means the market has already pre-loaded part of the dividend recovery and AI equity option.

So the operative question is no longer “Is SK Telecom a good company?” — it is whether the current ~₩21.48T market cap is supported by enough cash flow or asset value to justify another 20–30% re-rating from here.

2. Thesis Architecture

One-line thesis: SK Telecom is a re-rating candidate that puts telco core dividend recovery as the floor, and tries to redefine the network as an AI inference distribution network through Anthropic, AIDC, AI-RAN, Edge AI, and Network API.

Re-rating path

Step 1: Normalize impaired 2025 earnings & dividend
 ↓
Step 2: Confirm AIDC / GPUaaS revenue growth
 ↓
Step 3: Mark up AI equity stakes (Anthropic)
 ↓
Step 4: Commercialize Edge AI + AI-RAN + Network API
 ↓
Step 5: Telco EV/EBITDA multiple expands to upper band

Current location: early Step 2. AIDC / GPUaaS is real. Edge AI, AI-RAN, Network API are still options.

3. Fact Check & Structural Decomposition

Item	Confirmed	Investment Read
Spot	₩100,000 (2026-04-24)	Already at 52w-high range. Chase risk real.
Shares out	214,790,053	Market cap ~₩21.48T at spot.
2025 results	Rev ₩17.0992T / OP ₩1.0732T / NI ₩375.1B	Lowered base from hack + one-offs. Room for 2026 normalization.
2025 DPS	₩1,660	Sharp drop from ₩3,540 in 2024. Dividend recovery = 1st trigger.
AI Infrastructure Superhighway	AIDC / GPUaaS / Edge AI three-axis	SKT officially positions as AI infra operator.
AIDC	2025 revenue ₩519.9B, +34.9% YoY	AI business that can be underwritten by cash flow.
Haein GPU cluster	1,000+ NVIDIA B200 GPUs in single cluster	Real substance of GPUaaS / Sovereign AI infra.
Anthropic	$100M additional investment in 2023	Stake / dilution undisclosed, but SOTP option.
AI-RAN / 6G	RAN service + AI workload run in parallel on xPU/COTS	Long-term option to convert network into edge AI infra.
Governance	SK Inc. owns 30.57% of SKT	SKT re-rating directly accretive to SK Inc. NAV and owner collateral capacity.

4. Why SK Telecom — Ontology Decomposition

4-1. Capital flow direction

AI infrastructure capex flows in this sequence:

GPU / AI accelerator
→ Power / cooling / data center
→ Network / edge deployment
→ Inference services
→ B2B SLA / Network API

The early beneficiaries were NVIDIA, HBM, server, power, and cooling. SK Telecom sits in the trailing segment — i.e., it benefits when AI workload migrates from central GPU clusters to actual customer use, inference, and SLA monetization.

4-2. Value-chain position

Value chain	Primary beneficiary	SKT position
GPU / accelerator	NVIDIA, AMD, Rebellions, Arm ecosystem	Not a maker. Buyer / operator
AIDC	Cloud, DC, power, cooling	Operator + developer
GPUaaS	Cloud providers	Domestic sovereign AI infra supplier
Edge AI	Telcos, cloud, manufacturing SI	Owns network, central offices, base-station footprint
AI-RAN	Equipment vendors, telcos, semis	Standardization, PoC, operator
Network API	Telcos	Can monetize location, identity, QoS, SLA
B2B SLA	Telcos, SI	Owns existing enterprise base

The key: SK Telecom is not selling AI hardware. Its share is AI infra operation, edge location, network control, and B2B SLA monetization.

5. Core Re-Rating Drivers

5-1. Dividend normalization — floor support

2025 DPS collapsed to ₩1,660. 2024 paid 3 quarterly tranches of ₩830 + a year-end ₩1,050 for ₩3,540 total.

Dividend yield sensitivity

Dividend yield = DPS / price

Price	DPS ₩3,200	DPS ₩3,540	DPS ₩3,600
₩90,000	3.56%	3.93%	4.00%
₩100,000	3.20%	3.54%	3.60%
₩110,000	2.91%	3.22%	3.27%

Read: Below ₩90,000, the dividend-stock case re-engages. At ₩100,000, even a full DPS recovery to ₩3,540–3,600 yields only 3.5–3.6%. Dividend recovery alone is not enough; AIDC and AI-infrastructure confirmation are still needed.

5-2. Anthropic stake — unlisted AI call option

SKT additionally invested $100M into Anthropic in 2023, alongside a partnership for telco-specialized LLM development.

Pros:

The Anthropic stake can be a meaningful share of SKT market cap.
Higher AI-model valuations improve SKT’s SOTP.
Anthropic’s Korea / Asia expansion deepens strategic ties with SKT.

Cons:

Exact stake percentage, post-dilution ownership, and exit conditions are not disclosed.
Unlisted valuations have wide gaps between primary-round prices and secondary marks.
Cash realization timing for SKT shareholders is unclear.

Conclusion: Anthropic is not a near-term EPS driver for SKT. It is a SOTP mark-up option. Partly priced in already. An IPO or large new round would be a re-rating trigger.

5-3. AIDC / GPUaaS — most substantive AI line today

SKT’s AI infra strategy has three pillars: AIDC, GPUaaS, Edge AI. The company laid this out as the “AI Infrastructure Superhighway” at SK AI Summit 2024 — connecting metro GPUaaS, regional AIDCs, and Edge AI.

2025 AIDC revenue was ₩519.9B (+34.9% YoY) — already booked revenue, not narrative.

The Haein cluster is real: 1,000+ NVIDIA B200 GPUs in a single cluster, provided to government-led Sovereign AI Foundation Model project participants.

AIDC valuation approach — EBITDA margin, capex, contract tenor, and power costs aren’t sufficiently disclosed for a precise DCF. Revenue-multiple sensitivity is appropriate.

AIDC EV = AIDC Revenue × Revenue Multiple

Base revenue	Revenue multiple	AIDC EV
₩519.9B	2.0×	₩1.04T
₩519.9B	3.0×	₩1.56T
₩519.9B	5.0×	₩2.60T

Read: AIDC alone can already justify a SOTP block of ₩1–3T. If GPUaaS gets exposed to commodity-cloud price competition, the multiple compresses. Bundled with Edge AI / SLA / Sovereign AI, the multiple expands.

5-4. Edge AI + AI-RAN + Network API — long-term multiple expansion

SKT’s 6G whitepaper describes AI-native RAN running RAN service and AI workload in parallel on xPU-based COTS hardware, with AI compute pushed forward to the edge for low-latency inference and high-security AI services. It explicitly names Edge AI services and Network API delivery as the new business models.

Central AIDC / Haein GPU cluster
 ↓
Regional AIDC / central offices
 ↓
AI-native RAN / Edge UPF
 ↓
Low-latency inference / secure AI / Network API / Enterprise SLA

Component	Technical role	Business meaning
AIDC	Large GPU cluster	GPUaaS, AI inference, sovereign AI
Edge AI	Inference placed close to customer	Low-latency / secure inference billing
AI-RAN	Telecom + AI workload in parallel	RAN cost reduction + new AI services
Edge UPF	User-data path edge-localized	Latency / backhaul cost down
Network API	Location, identity, QoS as APIs	API call, SLA, premium-network billing
Enterprise SLA	Quality-guaranteed contracts	B2B recurring revenue

Core read: AI-RAN is not a 2026 EPS driver. But if it works, it changes SKT’s terminal multiple — the network gets re-classified from “GB-billing asset” to “distributed AI inference infrastructure.”

5-5. SK Inc. / owner incentives — structural push for re-rating

SK Inc. owns 30.57% / 65,668,397 shares of SKT.

SK Inc.'s stake value at ₩100,000
= 65,668,397 × ₩100,000
= ₩6.567T

Per ₩10,000 of SKT upside, SK Inc. NAV moves:

65,668,397 × ₩10,000 = ₩656.7B

Chairman Chey Tae-won’s stake in SK Inc. is reported by Reuters at 17.7%, with reporting that SK Inc. share-collateral lending has been discussed in the divorce-case financing context.

Per ₩10,000 of SKT upside, the chairman’s economic attribution is roughly:

₩656.7B × 17.7% ≈ ₩116.2B

Per SK Inc. official disclosure, largest shareholder + related parties = 25.4% / 18,430,379 shares.

Total related-party economic effect per ₩10,000 of SKT upside:

₩656.7B × 25.4% ≈ ₩166.8B

Read: SKT re-rating is not a direct cash inflow to the owner, but is favorable to SK Inc. NAV, collateral capacity, control stability, and the group AI narrative. AI-business optics + dividend recovery sit in a window where owner incentives partially align with minority shareholders.

6. Valuation Bridge

Anchors

Spot: ₩100,000
Shares out: 214,790,053
Current cap: ₩21.479T

Cap = price × shares
= ₩100,000 × 214,790,053
= ₩21.479T

Price	Cap	Δ vs spot
₩90,000	₩19.331T	-₩2.148T
₩100,000	₩21.479T	base
₩118,000	₩25.345T	+₩3.866T
₩130,000	₩27.923T	+₩6.444T

To get from ₩100,000 to ₩118,000, ~₩3.87T of additional equity value is required. At least 2 of the following must materialize:

Value bucket	Conservative	Bull case	Required confirmation
Dividend normalization	Floor support	Telco-flow rotation back	2026 DPS ~₩3,540
AIDC / GPUaaS	₩1.0–1.6T	₩2.5T+	Revenue growth, margin, customer contracts
Anthropic stake	₩1.0–2.0T	₩3.0T+	Stake %, dilution, IPO / new round
Edge AI / AI-RAN / Network API	₩0.3–1.0T	₩2.5T+	Paid PoC, API billing, RAN savings
Owner / Holdco incentive	Catalyst	Stronger IR / capital return	SK Inc. / SKT capital policy

My call: ₩118,000+ is achievable. But the path is not via dividend alone. At least 2 of {AIDC, Anthropic, AI-RAN} must show up as numbers.

7. Where the Market May Misprice

Likely under-pricing

AIDC seen as plain DC. When connected with GPUaaS, Sovereign AI, Edge AI, and Network API, AIDC can clear higher multiples than vanilla IDC.
AI-RAN treated as distant 6G theme. First-order value comes earlier — RAN energy efficiency, network automation, edge-AI infrastructure-ization — not 6G revenue.
Owner / Holdco incentives under-weighted. SKT re-rating is favorable to SK Inc. NAV and the chairman’s collateral capacity, creating a structural push for AI-narrative emphasis and dividend recovery.
Anthropic seen as plain unlisted equity. It’s not a financial stake — it’s a strategic asset for telco-specialized AI cooperation.

Likely over-pricing

Mistaking AI-RAN for near-term revenue. Currently in standardization / PoC / tech-display phase. Limited 2026 P&L contribution.
Assuming GPUaaS is high-margin. Capex, power, depreciation, and price competition are large. No high multiple before margin confirmation.
Treating SK Group AIDC totals as SKT shareholder value. Ulsan AIDC and group AI infra projects require careful economic-attribution between SKT, SK Broadband, AWS, and group affiliates.

8. Red Team

Macro failure modes

Rate re-acceleration or AI-capex-bubble fears compress AIDC / GPUaaS multiples.
Higher-than-expected power / cooling costs lower AIDC ROIC.
AI infra capex re-concentrates on hyperscalers, capping telco edge-AI take rate.

Micro failure modes

2026 DPS read at ≤ ₩3,200 → dividend-recovery thesis weakens.
AIDC growth decelerates or GPUaaS margin disappoints → no AI-infra multiple.
Anthropic stake / dilution disappoints market expectation → SOTP mark-up shrinks.
AI-RAN / Network API fails to convert into paying contracts → stays “tech whitepaper.”
Post-hack subscriber recovery and brand-trust restoration drags → core normalization lags.

9. Factor Checklist

Read matrix

Cohort	Factor read
Existing exposure	Dividend normalization and AIDC growth are the key factors
New interest	Price and earnings confirmation still matter
Higher-conviction setup	Below ₩90,000, or after earnings / dividend confirmation
Above ₩118,000	Valuation burden needs to be rechecked
Above ₩130,000	Need ≥2 of {Anthropic, AIDC, AI-RAN} confirmed in numbers

Confirmation conditions

Price: ≤ ₩90,000. With DPS ₩3,540, yield ≈ 3.93% — floor improves.
Dividend: Quarterly DPS recovers to ₩830–900 range — signal of return to normal 2024-level payout.
Earnings: Guidance points 2026 OP back to 2024 normal level — 2025 lowered base alone is not enough; core competitiveness must restore.
AIDC: AIDC revenue growth maintained ≥ 30% or new long-term customer contracts disclosed. GPUaaS must be characterized as recurring infra revenue, not short-lease.
AI-RAN / Edge: ≥ 1 of {paid PoC, Network API billing, RAN energy-savings %} disclosed. Option value can then be priced more aggressively.

Catalysts

Catalyst	Expected impact
2026 quarterly dividend normalization	Telco-yield rotation back
1Q / 2Q earnings show core recovery	2025 one-off base clears
AIDC revenue + margin disclosure	AI-infra business value can be underwritten
Anthropic IPO / new round	SOTP mark-up
Haein / GPUaaS private-customer expansion	Reduces government-project dependency
Edge AI paid contract	AI-RAN / Network API option becomes real
SK Inc. / SKT shareholder-return upgrade	Holdco-owner incentive aligns with minority

Invalidation

2026 DPS sticks ≤ ₩3,200.
AIDC growth decelerates ≤ 20%.
GPUaaS margin so low that AIDC = ROIC-dilutive line.
Anthropic stake confirmed materially below market expectation.
AI-RAN / Edge AI / Network API stuck at tech-display level with no monetization evidence.
Subscriber recovery + ARPU defense fails post-hack.

10. Final Read

The SK Telecom re-rating thesis is valid. But at ₩100,000 the first leg is done. Dividend normalization alone offers limited additional upside. The remaining upside is decided by AIDC / GPUaaS earnings conversion, Anthropic stake re-pricing, and commercialization evidence for Edge AI / AI-RAN / Network API.

From an investor’s perspective, SK Telecom is no longer a simple telco — it is a telecom operator attempting an AI-infrastructure transition. But it is not yet structurally an AI-revenue explosion story like NVIDIA or the hyperscalers. What is currently underwriteable in numbers: AIDC and dividend normalization. AI-RAN and Edge AI remain long-dated options.

In short, the setup is constructive but no longer early. The cleaner confirmation points are ≤ ₩90,000 on price, or additional dividend / AIDC / Anthropic data. To see ₩118,000+, the market must reclassify SK Telecom from “telco dividend stock” to “AI infrastructure operator.” Only some of that evidence has arrived.

Appendix — Evidence Tier

[Fact]

2025 consolidated revenue ₩17.0992T / OP ₩1.0732T / NI ₩375.1B.
2025 AIDC revenue ₩519.9B, +34.9% YoY.
SKT positions AIDC / GPUaaS / Edge AI as the three pillars of “AI Infrastructure Superhighway.”
Haein = 1,000+ NVIDIA B200 GPUs in a single cluster.
SKT additionally invested $100M in Anthropic.
SK Inc. owns 30.57% / 65,668,397 shares of SKT.

[Inference]

At spot, dividend normalization is a floor-support factor more than an upside factor.
AIDC / GPUaaS is the most-underwriteable AI line for SKT today.
Edge AI / AI-RAN / Network API can change the multiple but remain option value.
SKT re-rating is favorable to SK Inc. NAV and owner collateral capacity, creating a structural incentive to highlight the re-rating.

[Speculation]

The current value of the Anthropic stake cannot be fixed without disclosed ownership and dilution.
AI-RAN may grow into a real new revenue line for SKT around 2030, but is in tech / standardization phase today.
Network API becomes a meaningful revenue line only with API billing, SLA, and a developer ecosystem.

[Blocked]

Exact SKT stake / post-dilution ownership / liquidity terms in Anthropic.
Exact economic attribution of Ulsan AIDC across SKT / SK Broadband / AWS / SK Group affiliates.
GPUaaS gross margin, AIDC EBITDA margin, power tariff, customer-level long-term contract terms.
Real RAN cost-savings %, energy-savings %, and commercial-network deployment timing for AI-RAN.

Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.

Pearl Abyss Re-Rating: Crimson Desert Is More Than a Unit-Sales Story

Sat, 25 Apr 2026 22:00:00 +0900

Pearl Abyss research hub: This note is part of the Pearl Abyss and Crimson Desert investment research hub. Related reads: Patch 1.04 re-rating · KRW 60,000 support · Shinhan target-price gap.

Series — Pearl Abyss Crimson Desert Thesis (Part 8) Prior posts traced the launch sales arc, the BDO back-catalog re-rating, the short-sale absorption, the developer update, and the 1.04.00 patch as a Witcher-3 moment. This post zooms out: the thesis is no longer about Crimson Desert as a single product. It is about Pearl Abyss as a platform.

TL;DR

The core of the Pearl Abyss thesis is no longer “how many copies of Crimson Desert sold.” The more important question is whether Crimson Desert is forcing the market to reclassify Pearl Abyss as one of the few Korean studios operating as a global AAA production platform with a proprietary engine.

As of April 25, 2026, the data supports that possibility. Crimson Desert has officially crossed 5 million global units sold. Post-1.04.00 new-review positivity has jumped sharply. Steam sales rank in the three markets that actually matter — Global #3, Korea #3, US #5 — has re-strengthened. Short balance is still elevated, but that reads less like a confirmed bearish signal and more like the market has not yet conceded the platform re-rating.

The conclusion is not “buy at any price.” What is confirmed: the patch reduced user friction and slowed the decay rate. What is not confirmed: that the CCU floor has structurally moved up. The next checkpoints for this thesis are the expected 1.05.00 patch in early May, the 1Q26 print, and whether 6M units gets formalized.

1. The Market’s Default Frame: “Big Launch → Peak-Out”

The market’s base-rate framing of Pearl Abyss is simple:

Crimson Desert sold well. But package games concentrate revenue in the launch quarter. CCU naturally decays. So 1Q or 2Q earnings can spike, but profit collapses afterward. Therefore: cap the multiple.

This logic is not wrong. CCU has come down from launch peak. Local tracking shows the April 19 average CCU at 118.8K, and on the April 23 patch day average CCU dropped to 63.5K. Decay is real.

What the market is missing: a package game’s success is not summarized by a CCU print. In an AAA title with strong single-player DNA, CCU does not move 1:1 with units sold. The variables that matter more for trajectory are: new-purchase flow, review quality, sales rank, patch responsiveness, UGC propagation, and — most importantly — production capability that transfers to the next title.

That is exactly where the Pearl Abyss thesis lives. If Crimson Desert is not just “a game that sold well” but a commercial validation of Pearl Abyss’s proprietary engine and global AAA pipeline, then the multiple the market should apply is different.

2. 5 Million Units Is Already an Official Fact

Pearl Abyss officially announced global cumulative units of 5,000,000 on April 15, 2026 — the fastest pace in Korean console-game history at 26 days from launch. The same release disclosed that between March 20 and April 14, Crimson Desert generated 5,700+ Twitch streams and 108,000+ YouTube videos.

Two things matter here.

First, Crimson Desert did not ship under the Korean industry’s traditional success formula — mobile gacha or live-service MMORPG. It is a PC/console AAA package game targeting global Western markets. Selling 5M units in 26 days in that market means dev capability, brand reach, distribution, and Western user reception all fired simultaneously.

Second, UGC propagation may matter more than the unit number itself. Per the company release, the US accounted for 23.3% of YouTube video creation and 46.3% of views. Crimson Desert is not an Asia-only hit. It is being actively consumed in the West.

The post-5M sales path is not officially confirmed. There is no 6M release yet. So 8.5M / 10M trajectories are inference and speculation, not fact. But the 5M anchor alone fundamentally changes Pearl Abyss’s 2026 P&L structure.

3. Patch 1.04.00 Is Not a Bug-Fix. It Is Evidence of a Live-Ops Capability.

On April 23, 2026, Crimson Desert shipped patch 1.04.00. Headline contents:

Easy / Normal / Hard difficulty options
New storage and housing-related storage system
New pet
Keyboard/mouse and controller preset options
Inventory category tabs
New skill additions
Distance background quality improvement
Boss rematch system flagged for future patch

This patch matters not because of content volume but because it targeted the exact friction points the community had been venting about. Difficulty, inventory, controls, storage, pet, view-distance quality — all classic post-launch complaint vectors. Pearl Abyss responded inside 30 days with a single coordinated patch.

A day later, on April 24, 2026, hotfix 1.04.02 shipped — font scaling, difficulty-related issues, UI cleanup, HDR input, pet-name bug, localization, character movement-speed tuning. From an investor lens this patch cadence is not an operating event. It is a test of whether the proprietary engine and the dev org can absorb global user feedback fast enough to matter.

The deeper message: Pearl Abyss is grafting its MMORPG live-service operating DNA onto an open-world AAA package game.

A typical single-player package release ships → patches bugs and balance for a window → goes silent. The unspoken grammar is “we shipped a finished product; we’ll do post-hoc cleanup.” MMORPG operations are different. The MMO operator has to ingest user complaints in real time, re-prioritize the patch queue, defend community trust, and push content + QoL on a metronome. Pearl Abyss has been building that operating muscle for years through Black Desert.

What is happening with Crimson Desert is precisely that capability bleeding outward. Pearl Abyss is operating an AAA package game closer to a service than to a product — friction-point patches on tight cadence are not just satisfaction levers. They build user belief: this game won’t be abandoned. The studio is listening. Two weeks from now it’ll be better.

That belief is long-tail durability. Long-tail in package games is not built by content volume alone. It’s built by users feeling comfortable recommending the game to a friend, by negative reviews softening, by lapsed users finding a reason to relaunch, by new buyers concluding “right now is fine to buy.” If Pearl Abyss’s MMO live-ops DNA is genuinely transferring to Crimson Desert, this is not a post-launch patch — it is a structural capability that lowers the decay rate itself.

So the investor read on 1.04.00 is not “they added difficulty options.” It is: Pearl Abyss demonstrated it can iterate a global AAA package title at MMORPG live-service speed. If that capability repeats, it does not just extend Crimson Desert’s tail — it lowers the discount rate on DokeV and the next project.

4. Post-Patch User Data: CCU Bouncing, Review Quality Step-Change

Local tracking sampled every 20 minutes from April 6 23:40 KST through April 25 15:20 KST. Splitting around 1.04.00 makes the picture clean.

Pre-Patch 3 Days (April 20–22, 2026)

Metric	Value
Avg CCU	80,870
Peak CCU	155,732
New positive reviews	1,982
New negative reviews	422
New-review positivity	82.4%
Steam Global rank avg	#3.3
Korea rank avg	#3.3
US rank avg	#6.7

Sales rank was holding, but review quality was wobbling. New-review positivity bottomed at 78.1% on April 22. This was the window the bear case loved: “launch was strong, but user friction is compounding.”

Post-Patch Full 2 Days (April 23–24, 2026)

Metric	Value
Avg CCU	67,009
Peak CCU	94,827
New positive reviews	2,422
New negative reviews	106
New-review positivity	95.8%
Steam Global rank avg	#5.2
Korea rank avg	#5.2
US rank avg	#7.7

The headline isn’t CCU. It is new-review positivity moving from 82.4% → 95.8%, with negative reviews collapsing 422 → 106. That is strong evidence the patch actually reduced user friction.

Yes, average CCU was lower than pre-patch — but patch day mixed maintenance windows + weekday effects, and CCU recovered from April 24. Calling 1.04.00 a failure on average-CCU alone is overreach.

Last 24h (April 24 15:20 → April 25 15:20 KST)

Metric	Value	Comparison
Avg CCU	78,382	+16.8% vs prior 24h
Peak CCU	96,997	—
New positive reviews	848	+19.8% vs same hour prior week
New negative reviews	81	-52.6% vs same hour prior week
New-review positivity	91.3%	—
Avg CCU vs same hour prior week	-13.8%	—

This snapshot describes the present accurately. CCU is below prior week. But CCU is up vs prior 24h, and review quality is materially better than a week ago. Crimson Desert is not “returning to launch peak.” More precisely, it is decaying naturally, but with the patch reducing friction and slowing the decay rate.

5. Sales Rank: Core Markets Alive, Periphery Soft

Latest Steam rank from local tracking, April 25 15:20 KST:

Market	Rank
Global	#3
Korea	#3
US	#5
China	#10
Germany	#9
France	#17
Japan	#28
Taiwan	#56
Brazil	#33

The core is GL / KR / US — all still Top 10. Global #3 / Korea #3 / US #5 is incompatible with “sales momentum has fully evaporated.”

The weakness is real too. Japan, Taiwan, Brazil, France are soft. Crimson Desert’s long-tail is not uniformly distributed globally — it’s compressing into the core markets. Bull thesis discipline: do not claim “perfect long-tail everywhere.” The accurate claim is “core-market sales momentum holds; peripheral retention is weak.”

For positioning this distinction matters. If GL / KR / US Top 10 holds, Crimson Desert continues to support upward earnings revisions. If 2 of 3 fall out of Top 10, sales-tail expectations need to come down.

6. Short-Sale Data: Not a Bear Confirmation, but Lingering Doubt

The KRX short-sale CSV covers 24 trading days from March 24 through April 24, 2026. Key reads:

Date	Net short balance (shares)
2026-03-24	2,698,216
2026-04-08 (low)	1,579,992
2026-04-16	1,912,458
2026-04-22 (latest valid)	1,765,579

Date	Short volume	Short volume share
2026-04-23	269,865	32.3%
2026-04-24	76,600	9.6%

Caveat: April 23 and April 24 net balance prints as “0” in the CSV — that is almost certainly a reporting-lag artifact, not a real liquidation. Use April 22 1,765,579 as the latest reliable balance.

The interpretation is two-sided.

First, balance is meaningfully off the highs. From 2,698,216 on March 24 to 1,765,579 on April 22 — roughly -34.6%. Some short cover has already happened post-launch.

Second, balance has rebuilt off the April 8 low (1,579,992). Shorts have not fully exited. A cohort of the market is still betting “Crimson Desert peaked; post-1Q decay will be quick.”

Third, short volume share of 32.3% on April 23 — patch day — is a real pressure signal. Some of the market read peak-out as more important than patch impact. By April 24, that share dropped to 9.6% and the stock rebounded to ₩57,000. Shorts have not captured the tape.

So short data is not a confirmed bearish signal. More precisely: the market has not yet conceded that Pearl Abyss is a global AAA platform company, and that doubt is what’s left in the balance. If 1.05.00 and the 1Q print land well, that balance becomes incremental demand. If CD metrics and earnings deteriorate together, that balance becomes the right bet.

7. Why “Engine / Platform Re-Rating,” Not Just “Game Hit”

Treat Pearl Abyss as a single-game studio and the math is easy. CD sold well → 1Q/2Q strong → package games decay → cap the FY26 multiple.

Treat Pearl Abyss as a global AAA production platform with a proprietary engine and the conclusion shifts. Crimson Desert demonstrated five things at once:

A Korean studio can drive mass volume in the global PC/console AAA market.
A proprietary-engine open-world action RPG can ship commercially.
The studio can absorb post-launch friction and ship a coordinated mega-patch inside 30 days.
The title can drive Western-led UGC propagation on YouTube/Twitch and global communities.
The studio can graft MMORPG live-ops DNA onto an AAA package game, restoring user trust and lowering long-tail decay rate.

These five capabilities matter more than a single SKU’s economics, because they transfer — to DokeV, to DLC, to expansions, to the next project.

Proprietary engines are not free. Maintenance cost, debugging, console optimization, tech debt — all real. But Crimson Desert’s units + patch cadence suggest BlackSpace Engine can carry a commercial AAA project. That alone changes the option value.

The valuation framework needs to bend with that.

Frame	Market Read	Fair-Value Feel
One-off package hit	CD = single P&L event, fast decay	₩55,000 – ₩68,000
Earnings reset	1Q/2Q drive FY26 EPS upgrade	₩72,000 – ₩80,000
Platform re-rating	Proprietary engine + DokeV + DLC/expansion option	₩86,000 – ₩93,000+

Spot near ₩57,000 is closer to frame #1. The data already supports frame #2. Frame #3 — platform re-rating — is not yet fully proven. But CD’s units, patch response, UGC, and the persistent short balance argue the option is not in the price.

8. Investment Conclusion: Hold, Add Conditional

The current call is clean.

Existing position: hold. Trimming in the high ₩50Ks contradicts the “Pearl Abyss = global AAA platform / proprietary engine” thesis. The current price does not look like it adequately reflects either CD’s 1Q earnings or the platform option.

Adding is a separate question. If position size is already large, neither the short balance nor the patch response alone is sufficient reason to size up further. Multiple variables remain unconfirmed: actual 1Q26 revenue recognition, gross/net accounting notes, platform mix, refund rate, marketing expense, transaction-fee leakage.

Add conditions (all must hold):

Stock retraces to ₩52,600 – ₩55,000
GL / KR / US Steam Top 10 maintained
New-review positivity near 90%
April 26–27 weekend peak approaches or re-breaks 100K
No close break of ₩53,000s under short pressure

Thesis-kill conditions (any of these):

1Q26 OP ≤ ₩170B or OPM < 45%
Post-1.05.00 D+1 CCU rebound under +5%
New-review positivity sticky below 85–90%
2 of 3 (GL / KR / US) fall out of Top 10
May CCU collapses to 30–40K range
Short balance re-expands above 2.0M shares alongside deteriorating game metrics

9. Final Read

The Pearl Abyss thesis is no longer “Crimson Desert sold well.” The market already knows that.

The actual thesis is: Crimson Desert is the first commercial proof that Pearl Abyss can be re-rated as a global AAA production platform / proprietary-engine company.

5M units proved demand. Patch 1.04.00 proved live-ops responsiveness. Post-patch new-review quality proved user friction actually fell. Core-market Steam rank shows sales momentum is alive. The persistent short balance shows the market has not yet conceded the re-rating.

Not everything is proven. The CCU floor lift is not confirmed. Console sales data is gated. 1Q26 accounting notes are pending. But what matters in investing is not waiting for everything to be confirmed — it is catching the change before the market’s classification flips.

Pearl Abyss is on that boundary now. The market is still pricing “post-launch peak-out.” The data is tilting toward “decay slowing + platform re-rating.” That gap is the alpha source.

Appendix — Evidence Tier

[Fact]

Pearl Abyss officially announced 5M cumulative units on April 15, 2026.
Company release cited 5,700+ Twitch streams and 108,000+ YouTube videos for March 20 – April 14, 2026.
Patch 1.04.00 shipped April 23, 2026 with difficulty modes, storage, pet, input presets, inventory tabs, skills, and visual quality improvements.
Hotfix 1.04.02 shipped April 24, 2026.
Local tracking April 25 15:20 KST: Steam rank Global #3, Korea #3, US #5.
Last-24h average CCU 78,382; new-review positivity 91.3%.
Steam API snapshot: cumulative reviews — positive 117,772, negative 22,426, total 140,198.
Latest valid net short balance: April 22, 2026 — 1,765,579 shares.

[Inference]

Patch 1.04.00 actually reduced user friction.
Core-market sales momentum is still intact.
The market may be over-discounting CCU decay as if it were unit-sales decay.
The persistent short balance is doubt that the re-rating has happened yet, not a confirmed bearish signal.

[Speculation]

BlackSpace Engine can improve dev efficiency and incremental margin on DokeV, DLC, expansions, and the next project.
A successful 1.05.00 patch can further slow the Crimson Desert decay rate.
A formal 6M units announcement would weaken the peak-out short thesis.

[Blocked]

Console-only unit numbers and console sales rank.
Region-by-region revenue mix.
Refund rate.
1Q26 gross/net revenue recognition footnotes.
Actual marketing expense and transaction-fee leakage.
Real net short balance for April 23–24, 2026.

Pearl Abyss: Why Crimson Desert Patch 1.04 Could Re-Rate the Stock

Thu, 23 Apr 2026 22:30:00 +0900

Pearl Abyss research hub: This note is part of the Pearl Abyss and Crimson Desert investment research hub. The follow-through sequence is platform re-rating · post-patch weekend data · Shinhan target-price gap.

Series — Pearl Abyss Crimson Desert Thesis (Part 7) Prior posts track the launch (3/19), the four-week retention curve, the BDO back-catalog re-rating, the short-sale response, and the April Developer Update. This post covers patch 1.04.00 — the content update that forces the thesis upgrade.

What Shipped on April 23

Two events, one day:

Time (KST)	Event
Apr 23, 10:48	Patch 1.04.00 (Revised) — full release
Apr 23, 23:46	Patch 1.04.01 hotfix (all platforms)

Total download ~29GB. That is not a balance patch. That is a mini-expansion shipping 35 days after launch — on a two-week cadence that Pearl Abyss has now hit every single release since 3/19.

The Patch — Eight Categories

1. Difficulty System (biggest single change)

Easy / Normal / Hard selectable mid-game. Easy reduces damage, widens parry/dodge windows, reduces boss counter frequency. Hard adds extra boss patterns, tighter i-frames. “Boss rematch” flagged as coming next — so the content pipeline is visible one patch ahead.

This is the most important change in the entire update. Difficulty modes expand TAM by letting story-first casuals complete content that was gating review scores, without alienating the hardcore combat crowd. It is the exact mechanism The Witcher 3 used post-launch.

2. Housing / Storage Overhaul

Sturdy Gatherables Chest — 1,000 slots
Collectibles Chest — 1,000 slots
Wardrobe — up to 1,000 slots
Kuku Cooler / Enhanced — 40 / 330 slots (food)
Select House — 4 layouts (Compact / Standard / Spacious / Pailunese)
Key QoL: stored items usable for crafting/cooking without being in the active inventory

This is the quality-of-life gate that “the review complaints were actually about.” It got fixed in one patch.

3. Pet System Expansion

5 new cats + new pet + Abyss Heuklang petification, pet accessory slots (Sigil of Bonding), rename capability.

4. New Content

Four Constabulary regions, three new weapons (Sword of Starlight, Tree Branch, Sturdy Tree Branch), Baltheon outfit, Pororin secret shop (pet gear), 13 new tattoos.

5. Graphics

Distant scenery improvements, some 2D asset replacements.

6. Controls

KB/M and controller presets (Classic), dodge/roll input swap, aim control improvements.

7. Combat Balance

Boss i-frames during heavy attacks removed. Status-effect elemental damage up. New skills: Force Palm Pulse (3-stage charge), Weapon Throw (dual-wield).

8. QoL

Item lock (sell/discard protection), instant wake after rest, Abyss Artifacts re-obtainable on skill observation.

Global Press Reception

Outlet	Headline	Read
Push Square	“Huge and Highly Praised”	Fully positive
Kotaku	“Biggest Update Yet Transforms It Once Again”	Witcher 3 post-launch comparison
VULKK	Detailed review	Favorable
GosuGamers	“Overhauls housing, expands pets”	Positive

Player comments quoted in Kotaku:

“like 2.0”
“this is like 4.0, every patch is a huge W”
“They just added so much random stuff, I love it”

The Witcher 3 comparison is the specific framing that matters. That comparison was not circulating at launch. It is circulating now. Press framing shifts are leading indicators for sell-side narrative shifts.

Korean press (Cheongnyeon Ilbo, Inven, Kyunghyang Games, Ruliweb) uniformly positive on the QoL and difficulty expansion. Residual bear point: some users still flag the quest/story system as unresolved — valid but not a patch 1.04 scope issue.

The Numbers Moved With the Patch

Metric	Pre-patch	Post-patch	Read
Daily review volume	871	1,353 (+55%)	Users feel it’s worth reviewing again
New-review positivity	83.7%	94.9% (+11.2pp)	Very satisfied
Cumulative positivity	83.76%	83.89% (reversal)	First turnaround since launch
Global sales rank	#4	#4	Holding

94.9% new-review positivity is immediately below the 95% “Overwhelmingly Positive” Steam threshold. If that holds 3+ days, Steam’s recommendation algorithm materially weights the page upward in recommendations and discovery. That is a structural, algorithmic tailwind to new-user acquisition, not just a vibes read.

Pearl Abyss’s Service Cadence — The Real Thesis

Patch	Date	Type
1.00.00	Mar 19	Launch
1.00.03/04	Mar 23-25	Hotfix
1.01.00-03	Mar 29-31	Content + hotfix
1.02.00	Apr 4	Content
Dev Update	Apr 10	Communication
1.03.00-01	Apr 11-12	Content
1.04.00-01	Apr 23	Major (29GB)

Every two weeks. No misses. This is BDO-proven live-service muscle being ported to single-player. The market has been pricing Crimson Desert as a single-shot title. The cadence argues it should be priced as BDO-style perpetual content.

Revised Thesis — Valuation Re-Rating

The right way to read this patch is: the patch closed the single biggest bear-thesis element (QoL / difficulty / review scores stuck at 83%) and opened visibility into the content pipeline (boss rematch confirmed for next patch).

Applying P/E 15× (appropriate for a title that has now demonstrated BDO-style live-ops cadence, not 12× appropriate for a one-shot premium title):

Item	Pre-patch view (4/22)	Post-patch view (4/23)
Week 5 decay	-15 to -20%	-15% confirmed
Week 6 decay	-15%	-10% (patch catalyst)
D53 (May 12)	6.00-6.10M	6.10-6.25M
Y1 2026 sales	10.0-11.0M	10.5-12.0M
CD 2026E revenue	₩500-530B	₩520-560B
Consolidated 2026E revenue	₩970B-1.0T	₩990B-1.03T
EPS	₩5,400-5,500	₩5,500-5,750
Fair value (P/E 15×)	—	~₩82,500-86,250
Vs. spot ₩53,100	—	+55 to +62%

The 12× → 15× multiple expansion is itself a thesis bet. Argument: Pearl Abyss is no longer a one-shot CD company. It is a multi-IP live-ops company with BDO legacy and CD as a recurring-content franchise. Multi-IP live-service Korean gaming comps trade 15-20× on stable earnings. 15× is the low end of that range.

On the more cautious P/E 12× base case, fair value is ~₩66,000-69,000 (+24-30%). That is the floor. 15× is what happens if the market agrees the live-ops muscle is real.

Investment Read — Three Horizons

Short (through May 12 earnings):

Week 6 decay tightening to -10% (patch catalyst)
4/25-26 weekend CCU likely ≥150K (if confirmed, patch bump = confirmed)
D53 cumulative projection: 6.10-6.25M units

Medium (Q2):

Easy difficulty → casual TAM expansion
Boss rematch (“coming soon”) = next patch pipeline visible = continuous catalyst
Two major content patches per month = strong evidence of Y1 CCU-decay mitigation

Long:

The Witcher 3 model — post-launch sustained improvement → store-rank maintenance + sale-event spikes → extended tail
Pearl Abyss IP portfolio re-rating (BDO back-catalog revival + CD long-support)

Watch-List — This Week

Apr 24-26 weekend CCU peak — ≥150K confirms patch bump; <120K weakens it
Review positivity crossing 95% — 3-day hold triggers Steam algorithm premium
Hard-mode boss rematch announcement — tests whether “coming soon” lands in 1.05 on cadence
First discount event — 1-month-post-launch sales typically trigger in early May; size of unit lift is the read on elasticity

Lessons Booked

Yesterday’s “4/22 low 29K breakdown” read was a maintenance artifact. Daily aggregation alone is insufficient — need hourly trace when patches are in flight.
Google Trends “pearl abyss crimson desert update” rising +750% correctly led the 1.04.00 announcement. Trends rising queries are a real leading indicator — not sentiment noise.
The BDO-revival thesis strengthens. This patch is the specific evidence that made the “Pearl Abyss IP portfolio re-rating” argument concrete rather than thematic.

Bottom Line

Crimson Desert’s 1.04.00 patch closed the biggest bear point (QoL + difficulty gating reviews) and demonstrated that Pearl Abyss’s live-service muscle — proven on BDO — carries to single-player. New-review positivity jumped to 94.9%, 11pp above pre-patch. At P/E 15× on revised EPS of ₩5,500-5,750, fair value lands at ~₩82,500-86,250 vs ₩53,100 spot — a +55-62% re-rating setup. The multiple itself is the thesis: this is no longer priced as a one-shot game.