Status: Active planning baseline. Created 2026-06-12 to close the audit gap
(V1_V7_PLAN_SET_AUDIT_2026-06-12.md §6.1(4)): the exit criteria gate
matchmaking at "5× expected launch concurrency" (V4/V4_features.md:1825,
:2744, V4/V4_ARCHITECTURE.md:1881) but no document stated what expected
launch concurrency is. This document states it, derives it, and attaches
the revenue and cost envelope the live service runs against.
All figures not citing a spec line are planning assumptions adopted 2026-06-12, owned by the Finance/BizOps Lead (revenue), Online Services Lead (capacity), and Live Ops Director (engagement), and are re-forecast at Open Beta +7 days (≈2026-08-10) and at launch +7/+30/+90 days.
1. Expected Launch Concurrency (ELC) — the anchor number#
ELC = 250,000 global peak CCU in the first launch weekend
(2026-10-03/04), all platforms combined.
The 5× staging gate target is therefore 1,250,000 synthetic CCU, with
matchmaking p99 ≤ 35 s held at that load (V4/V4_features.md:2744).
Derivation#
Premium-priced launch, 9 platforms (V4/V4_features.md:300-304):
- First-week sell-through (base): 4.0 M units. Comparables band (2024–2025 launches, publisher-reported): Battlefield 6 >7 M units in 3 days (Oct 2025); Black Myth: Wukong ~10 M in 3 days (Aug 2024, single genre, heavily China-weighted); annual CoD historically 15–25 M first month; Helldivers 2 ~8 M in ~6 weeks. V4 is a new IP without an annual franchise's pre-order machine — base case set at roughly half of BF6's opening: 4.0 M week-1.
- Launch-weekend DAU: 55% of owners = 2.2 M (week-1 owners play in week 1; 55% on the same calendar day is the high end of normal).
- Peak CCU ≈ 11.5% of DAU at the global prime-time overlap (EU evening / NA midday) — within the 10–15% peak-CCU/DAU band observable from public Steam CCU vs. estimated DAU for major shooter launches.
- 2.2 M × 11.5% ≈ 253 k → round to 250 k.
Cross-check against public peaks: BF6 Steam-only launch peak ≈ 750 k (Steam is roughly a third-to-half of a multiplatform shooter's concurrency), CoD HQ Steam all-time peak ≈ 490 k, Helldivers 2 ≈ 458 k Steam. A 250 k all-platform peak for a new IP sits deliberately below all of these.
Scenario band#
| Scenario | Week-1 units | Peak CCU | 5× staging target | Trigger to adopt |
|---|---|---|---|---|
| Conservative | 2.0 M | 120 k | 600 k | Open Beta peak <55% of pro-rated base |
| Base | 4.0 M | 250 k | 1.25 M | default |
| Optimistic | 8.0 M | 480 k | 2.4 M | Open Beta peak >150% of pro-rated base; capacity must be pre-provisioned to this level regardless (see §3) |
Open Beta (2026-08-03, open enrollment, V4/V4_features.md:2724) pro-rated
expectation: 35–45% of launch peak (planning assumption) → expected Open Beta
peak 90–110 k base case. The Soak test is already specced at 200 k real
players (V4/V4_features.md:2725-2726), i.e. 0.8× base ELC — a deliberate
near-full-scale rehearsal.
Regional split of peak CCU (base case)#
Mapped to the 10 dedicated-server regions
(V4/V4_ARCHITECTURE.md:983-984). Shares are planning assumptions adopted
2026-06-12 from genre norms (premium shooter, strong EU/NA skew; RTS and
Wukong cells add APAC weight).
| Region | Share | Peak CCU (base) | 5× test |
|---|---|---|---|
| NA-East | 18% | 45,000 | 225,000 |
| NA-West | 12% | 30,000 | 150,000 |
| EU-West | 20% | 50,000 | 250,000 |
| EU-East | 9% | 22,500 | 112,500 |
| APAC | 17% | 42,500 | 212,500 |
| India | 6% | 15,000 | 75,000 |
| LATAM | 8% | 20,000 | 100,000 |
| ME | 4% | 10,000 | 50,000 |
| Africa | 2% | 5,000 | 25,000 |
| AU/NZ | 4% | 10,000 | 50,000 |
| Total | 100% | 250,000 | 1,250,000 |
Regional prime-time peaks exceed each region's share of the global peak; per-region capacity is provisioned at 1.25× the table row (regional-prime factor, planning assumption), and the 5× matchmaking gate is run per region at that factored load, not only globally.
Mode mix at peak (drives server demand)#
Planning assumption adopted 2026-06-12; re-measured in Open Beta:
| Mode family | Share of CCU | Server model | Players/instance |
|---|---|---|---|
| Tactical PvP (5v5/6v6/SvM) | 40% | dedicated client-server (V4_ARCHITECTURE.md:976-1035) |
10/12/4 |
| Battle Royale | 12% | dedicated, ReplicationGraph bands | 100 |
| RTS (lockstep relay) | 10% | input-relay server, 25 Hz (:890-932) |
2–8 |
| PvE co-op (Spec Ops, RTST co-op, Wukong 2P, Horde) | 18% | dedicated or host-authoritative session | 2–6 |
| Single-player (campaigns, Hitman sandbox solo) | 20% | no game server (services only) | — |
→ Server-attached share ≈ 80% of CCU; full-simulation dedicated servers ≈ 70% of CCU.
2. Hosting cost envelope#
Unit cost assumption#
Blended $0.009 per server-attached CCU-hour (planning assumption adopted 2026-06-12), decomposed:
- Compute: tactical 5v5/6v6 instances at ~0.9 vCPU each (60 Hz server
tick), BR 100-player instances at ~10 vCPU, RTS relays at ~0.05 vCPU
(lockstep servers only collect/broadcast input bundles,
V4/V4_ARCHITECTURE.md:900-905). On bare-metal/reserved fleet at ~$0.022/vCPU-hour effective, weighted by the §1 mode mix → ≈ $0.0021/CCU-h compute. - Egress: from the specced bandwidth budgets
(
V4/V4_ARCHITECTURE.md:1030-1035): 32 KB/s (5v5) ≈ 115 MB/h, 38 KB/s (6v6) ≈ 137 MB/h, BR 25 KB/s avg ≈ 90 MB/h, SvM 28 KB/s ≈ 101 MB/h. Weighted ≈ 0.11 GB/CCU-h; at a $0.02/GB blended (bare-metal-included + cloud-burst mix) ≈ $0.0022/CCU-h. - Burst premium + headroom: launch quarter runs 30% cloud-burst capacity on top of the reserved floor (autoscale validation per risk R-07) ≈ $0.0024/CCU-h amortized.
- Fleet orchestration, DDoS scrubbing, relay/QoS probes: ≈ $0.0023/CCU-h.
Launch-quarter monthly cost (base case)#
- Average CCU ≈ 38% of peak over the 24 h global curve (planning assumption) → 95 k avg CCU; server-attached 80% → 76 k.
- 76 k × 730 h × $0.009 ≈ $499 k/month game-server hosting.
- Platform services (the Rust service fleet: matchmaking, login,
leaderboards, telemetry, workshop, store —
V4/V4_ARCHITECTURE.md§ services), databases, observability, CDN for the ≤6 GB day-1 patch (V4/V4_features.md:2727) and ongoing content: +45% overhead ≈ $225 k. - Total infra ≈ $725 k/month at base launch quarter (conservative ≈ $390 k; optimistic ≈ $1.35 M). Per-MAU infra cost ≈ $0.15–0.18/month — comfortably inside MTX ARPU (§5).
AI-inference cost caps#
The specced AI surfaces are the Calliope commentary stack — end-of-match
voiced highlight packages, per-match analysis, live spectator commentary, and
P3 extensions (V4/V4_features.md:2571-2601,
V4/cinematics/live-ai-commentary-expansion.json,
V4/cinematics/ai-generated-commentary-extended.json); the AI Director is
local heuristics with no inference cost (V4/V4_features.md:1527-1564). The
manifests spec consent, bias review, and opt-out but no spend caps; the
following caps are adopted as planning assumptions 2026-06-12 and should be
promoted into the runtime manifests:
- End-of-match packages are generated on demand (player opens the results reel), not for every match. At base: ~340 k matches/day; 18% open rate; $0.01/package (short LLM pass + TTS) ≈ $18 k/month. Hard cap: ≤3 generations per player per day.
- Live spectator commentary only on partnered/esports broadcasts, ≤200 concurrent commentary streams; per-stream $0.60/h → ≤ $86 k/month worst case, expected <$15 k.
- Podcast recaps batch once per season per cell (60 generations/season — negligible).
- Global ceiling: AI inference ≤1.5% of monthly net revenue, alert at 1.0%. Degradation ladder when breached: (1) lower open-rate cap, (2) text-only analysis (drop TTS), (3) live commentary to esports Majors only. Degradation never silently substitutes unreviewed output (risk R-14 fail-closed posture).
3. Capacity commitments this forecast creates#
- Reserved fleet sized to base peak × regional 1.25 factor; burst path validated to optimistic peak (480 k) within 30 minutes/region.
- The 5× gate (
V4/V4_features.md:2744) is executed as: 1.25 M synthetic clients distributed per the §1 regional table × 1.25, mode mix per §1, matchmaking p99 ≤35 s per region, run twice — once before Open Beta entry and once at Soak. - Login queue with ETA active from minute zero (graceful-degradation posture,
V4/V4_features.md:296-299); queue triggers at 92% regional capacity.
4. Monetization model#
Specced inputs: battle pass, 100 tiers / 90-day season, free + premium tracks,
tier-skip purchase (V4/V4_features.md:1946-1950); operator DLC seasonal
(:1954-1957); mission DLC $9.99–$19.99, content-not-power (:1958-1960);
cosmetic-only store (:1951-1953); three currencies
(V4/V4_ARCHITECTURE.md:1163-1165); PPP regional pricing
(V4/V4_features.md:2250); creator-code revenue share (:1994-1997).
Planning assumptions adopted 2026-06-12 where the spec is silent:
- Premium SKU price: $69.99 standard / $99.99 deluxe (12% deluxe mix). Blended net revenue per unit $38 after 30% platform fee, PPP regional pricing, launch-window discounts on PC, and returns.
- Battle pass price: $9.99 (1,000 Combat Points); attach rate 22% of 90-day actives base case (premium-shooter band 15–30%; F2P comparables run higher and are not used). Tier-skip + cosmetic store adds 40% on top of pass revenue from the same payer base.
- Mission DLC: two drops in Year 1 at $14.99 average, attach 8% of owners per drop.
- Funnel/retention targets: Open Beta → purchase conversion 25% of beta players; D1/D7/D30 retention 52% / 27% / 13%; payer conversion (any MTX) 14% of 90-day actives; MTX net ARPU per MAU $1.40/month base.
5. Year-1 revenue scenarios (net of platform fees)#
| Line | Conservative | Base | Optimistic |
|---|---|---|---|
| Year-1 units | 4.5 M | 9.0 M | 15.0 M |
| Premium net (×$38; conservative ×$36 due to deeper discounting) | $162 M | $342 M | $570 M |
| Avg MAU (Y1) | 1.8 M | 4.0 M | 7.5 M |
| Battle pass (4 seasons; attach 14% / 22% / 26% of quarterly actives, ~$7.0 net/pass) | $7 M | $25 M | $55 M |
| Store cosmetics + tier skips (0.4× pass line… scaling with ARPU) | $10 M | $42 M | $95 M |
| Mission DLC (2 drops; attach 5% / 8% / 10%; ~$10.5 net) | $5 M | $15 M | $32 M |
| Year-1 net revenue | ≈ $184 M | ≈ $424 M | ≈ $752 M |
| Infra + AI (12 mo, §2 scaled) | $5 M | $9.5 M | $17 M |
| Live team + support 12 languages + watchtower (planning assumption: 140 / 190 / 240 FTE-equivalent at $175 k loaded) | $24 M | $33 M | $42 M |
| Live-service margin before marketing/UA | $155 M | $381 M | $693 M |
LTV check (base): 12-month net LTV per payer ≈ premium $38 + MTX ($1.40 × 12 / 14% payer share ≈ $120 per payer... expressed per-owner: $38 + $16.8 MTX = ≈ $55 net per owner), against a marketing/UA envelope to be set ≤ $12 per acquired owner (planning assumption — gives ≥4.5× LTV/CAC).
Floor test (links risk R-11): the conservative scenario still covers the
live cost base ~6×; the true downside is conservative units and attach
<10%, which puts MTX (~$15 M) below the live team line (~$24 M) — the §2
degradation ladder plus season-length stretch (90→120 days) is the pre-agreed
response, never power monetization (V4/V4_features.md:1951-1953 is
non-negotiable).
6. Re-forecast checkpoints#
| Date | Input that lands | Action |
|---|---|---|
| 2026-08-10 (Open Beta +7 d) | real CCU curve, mode mix, funnel conversion | re-pick scenario; resize reserved fleet; re-run 5× gate if scenario changes |
| 2026-09-25 (post-Soak) | 200 k-player soak telemetry | final launch capacity sign-off |
| Launch +7/+30/+90 d | sell-through, attach, retention | re-forecast revenue; battle-pass attach trigger check (R-11); per-cell metrics feed market-analysis.md §6 |
Owners: Finance/BizOps Lead (model of record), Online Services Lead (capacity), Live Ops Director (engagement inputs), Program Director (scenario adoption decision).