VNET Group

Phase A — Understand the business

Lens 1 · Company Overview

What it is. A leading carrier-neutral and cloud-neutral internet-data-center (IDC) services provider in China, with "one of the largest carrier-neutral data center networks in China" — 889 MW wholesale capacity in service + 49,863 retail cabinets in service as of 2025-12-31 . Cayman holding co (VNET Group, Inc.), Nasdaq-listed since April 2011 (ticker VNET; 1 ADS = 6 Class A ordinary shares) . Operations conducted almost entirely in China through PRC subsidiaries + consolidated VIEs (Beijing Yiyun Network Technology Group and affiliates) — a classic China-ADR contractual-control structure ``.

How it makes money (three lines):

• Wholesale IDC — build-to-suit hyperscale sites for "internet giants and large-scale cloud computing service providers," constructed to customer spec since the 2019 "dual-core" pivot. The growth engine ``.
• Retail IDC — colocation (single cabinet up to MW-scale), interconnection (BGP/multi-line), and value-added (bare metal, hybrid IT, firewall, backup). The legacy cash cow ``.
• Non-IDC — cloud services (notably operating Microsoft Azure + Microsoft 365 in China since 2013 via partnership — VNET is one of the few licensed operators of foreign public cloud in China) and enterprise VPN/SD-WAN ``.

Contract structure. Recurring-revenue model: ~90% of net revenues are recurring, billed monthly over contract terms . Average monthly recurring revenue per retail cabinet RMB9,045 in 2025 (RMB8,769 in 2024) . Extremely low churn: 0.1% monthly (2025) vs 0.2% (2024) and 0.4% (2023) — very sticky, mission-critical infrastructure . Wholesale contracts are long-dated, high-commitment build-to-suit (95.3% commitment rate on in-service capacity) .

Customers. >1,500 managed-hosting customers, >7,000 enterprise customers total . Wholesale is concentrated: **two entity customers each generated 10–20% of total net revenue in 2025** (one such customer in 2023 and 2024) — hyperscale/internet-giant anchors (the FY2025 marquee win is a ByteDance 500 MW agreement, the largest order in company history ``).

Suppliers. Bandwidth + partnered-cabinet capacity from the three state carriers (China Telecom, China Unicom, China Mobile); top-5 suppliers = 46% of bandwidth/cabinet resources in 2025 (rising from 42% in 2023) ``.

Lens 2 · Supply Chain

Upstream → VNET → end customer, with named stakeholders:

• Land & power (upstream, the true chokepoint): greenfield sites / industrial buildings acquired or leased; permits and power allocation obtained via local government authorities and electric utilities. In China's tier-1 markets, "high-power capacity + optical-fiber connectivity" real estate is the binding constraint the 20-F flags as its #1 growth risk . VNET's wholesale footprint is concentrated in the **Greater Beijing Area (578 MW, 65%)** and **Yangtze River Delta (311 MW, 35%)** , plus AI hubs like Ulanqab (Inner Mongolia) for high-density builds ``.
• Connectivity & partnered cabinets (single-source-ish): China Telecom, China Unicom, China Mobile lease bandwidth and partnered cabinets to VNET (partnered = lower margin than self-built). Top-5 suppliers 46% of these resources — a real dependency on the three SOE carriers ``.
• Construction & equipment: third-party developers, contractors, cooling/power-management suppliers; RMB4.00B of purchase commitments for machinery/equipment/CIP/bandwidth/cabinet capacity due within 12 months of FY2025 ``.
• Cloud upstream: Microsoft (Azure/365 licensing for the China cloud business; VNET prepays Microsoft for cloud services — a named prepaid-expense item) ``.
• VNET (the node): designs, builds, powers, cools, secures and operates the DC; guarantees 99.9% power + connectivity uptime for self-built sites ``.
• Downstream (buyers): internet giants / hyperscale cloud (wholesale) — e.g. ByteDance ``; ~7,000 enterprises + 1,500 hosting customers (retail); Microsoft cloud end-users (non-IDC).
• Capital "supply chain" (distinctive): VNET recycles stabilized DC assets into Shanghai-listed asset-backed securities (ABS / pre-REIT) — a RMB860M holding-type real-estate ABS listed Dec 2025, plus two Feb-2026 ABS schemes (RMB1.702B + RMB4.654B) via Sinolink Securities, ~70% sold to institutions, VNET retaining ~30% ``. This is a genuine, named funding channel, not generic.

Chokepoints: (1) power/land permits in tier-1 markets (the growth governor); (2) the three SOE carriers for connectivity; (3) offshore USD funding vs onshore RMB cash (VIE/FX transfer restrictions — see Lens 10). Names present → lens passes.

Lens 3 · Competitive Advantages (moats)

• Scale + interconnection density (real): one of China's largest carrier-neutral networks, interconnected with all major carriers; proprietary "smart routing." First carrier-neutral operator in China (1999) and first to offer global cloud (Microsoft, 2013) . In retail carrier-neutral, #1 by revenue at 12.3% share (2025) .
• Switching costs / stickiness (real): 0.1% monthly churn, ~90% recurring revenue, mission-critical placement — customers "only place mission-critical servers" in premium carrier-neutral sites ``. This is the strongest, most durable moat.
• Location/power banks (real but capital-hungry): secured land + power in Greater Beijing / YRD tier-1 markets where supply is scarce; 2.48 GW of total wholesale resource capacity secured as of Q1 2026 ``. Being early on power allocation is the wholesale moat.
• Licensing moat (narrow but rare): operating Microsoft Azure/365 in China is a regulatory-scarce franchise ``.
• Bargaining power — mixed. Over customers: LOW-to-MODERATE in wholesale (two 10–20% hyperscale anchors have leverage; build-to-suit ties VNET's capital to their demand). Over suppliers: LOW vs the three SOE carriers (46% top-5 concentration). VNET is more a price-taker on inputs and a capacity-taker to hyperscalers than a franchise with pricing power — the moat is stickiness + location, not pricing.
• Emerging strategic moat: the pending CATL anchor could add an energy edge (power procurement, efficiency, storage) that rivals lack (Lens 8/9/12).

Lens 4 · Segments

Revenue by business line (RMB, consolidated) + segment split:

Wholesale mix went 13.8% (2023) → 23.6% (2024) → 34.8% (2025); wholesale grew +90.4% in 2024 and +77.4% in 2025, while retail/non-IDC is essentially flat ``. Read that plainly: all the growth is AI/hyperscale wholesale; retail is a mature ~RMB6.4B annuity. The company reports one operating segment (managed hosting) but discloses the wholesale/retail split from Q1 2024 onward.

Geography: wholesale 65% Greater Beijing / 35% Yangtze River Delta; retail self-built ~56% Greater Beijing, ~20% YRD, ~14% Greater Bay, ~9% other . Effectively an all-China, tier-1-concentrated footprint — no meaningful overseas revenue (an early LERETECH HK "overseas expansion" cooperation exists but is immaterial) .

Margins by mix: gross margin 22.0% (2025) vs 22.2% (2024) vs 17.4% (2023) . Self-built > partnered on margin; as wholesale/self-built scales and move-ins ramp, adjusted EBITDA margin has climbed to ~30% (2025) from ~28% (2024) .

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print)

Reference prints: FY2025 (reported 2026-03-16) and Q1 2026 (reported ~2026-05; latest available).

FY2025 +:

• Revenue RMB9,949.3M (+20.5%); wholesale +77.4% to RMB3.46B drove it.
• Gross profit RMB2,192.5M (+19.7%); GM 22.0%.
• Operating income ≈ RMB780.3M (GP 2,192.5 − opex 1,412.2); operating margin ~7.8% ``.
• Adjusted EBITDA RMB2.98B (+22.6%), 30.0% margin — beat 2025 guidance (RMB2.76–2.82B) ``.
• GAAP net loss attributable RMB256.8M (vs +RMB183.2M in 2024, −RMB2,643.8M in 2023) ``.
• The bridge from EBIT to a net loss: interest expense RMB598.6M (+49.3%, the key structural drag), a RMB314.3M convert fair-value loss (2027 notes mark), offset by a one-time RMB469.8M gain on ABS deconsolidation, then crushed by income tax RMB557.5M at a 131.5% effective rate (valuation-allowance + intra-entity-transfer + Cayman effects) ``. Strip the one-time ABS gain and the loss is materially worse — flag this.

Q1 2026 (latest) ``:

• Revenue +19.8% YoY; adjusted EBITDA RMB891.5M (US$129.2M), +30.6% — beat.
• Net loss attributable widened to RMB531.8M (US$77.1M) (vs RMB237.6M Q1 2025), driven by RMB486.2M capital-transaction-related income tax from the ABS/asset-restructuring — the tax line is again the villain.
• Wholesale utilized capacity +64 MW to 687 MW; utilization 75.7% (+5.6 pp QoQ). New orders YTD 519 MW (incl. a 400 MW + 110 MW internet-customer order in Greater Beijing) ``.
• Stock fell ~3.9% on the print — the widening GAAP loss overshadowed the EBITDA beat ``.

Balance-sheet flags ``: cash + ST investments RMB6,580.9M (US$941M); PP&E net RMB22,775.6M (+32% YoY — the buildout); AR net RMB2,222.1M (+34%, outrunning revenue +20.5% → DSO extending, doubtful-debt allowance up to RMB267.7M); OCF RMB1,918.6M (flat/slightly down YoY despite +20% revenue — a yellow flag, see Lens 10).

Read: operationally excellent (wholesale/AI demand, guidance beats, EBITDA +30%), but the print you actually receive at the net line is a loss, and it's getting wider, not narrower, because of tax + interest + convert marks. Management wants you to watch adjusted EBITDA; a skeptic watches the cash and the net line.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on the shelf (transcripts/ empty); this lens is , from the FY2025 (Mar 2026) and Q1 2026 (May 2026) calls .

• Consistent, intensifying theme: "wholesale / AI / move-ins." Each recent call leads with wholesale delivery speed, move-in ramp, commitment/pre-commitment rates and order wins (ByteDance 500 MW; 517 MW of new wholesale orders YTD 2026). Tone is confidently bullish and guidance-raising in spirit (repeated beats) ``.
• Rising vocabulary: "AI," "high-density," "power," "delivery/move-in cadence," "pre-commitment rate 85.8%," "2.48 GW total resource," and now "CATL / energy synergies / strategic partner" ``.
• Faded vocabulary: the 2022–2023 language of "cost optimization," "impairment," "consolidation," and defensive balance-sheet talk has receded as the AI cycle turned demand favorable.
• The tension management downplays: the GAAP loss and the tax/interest drag. Calls steer hard to adjusted EBITDA and operating KPIs; the widening net loss is framed as non-cash / one-off tax. Sentiment: operationally euphoric, financially defensive — a gap worth watching.

Lens 7 · Comps

Provenance-critical; unsourced multiples marked n/a.

Takeaway: the only clean apples-to-apples China comp is GDS at 15.3x EV/EBITDA. VNET at ~10x trailing / ~8.4x forward trades at a ~30–45% discount — justified in part (VNET's 30% EBITDA margin vs GDS's 47%, higher retail mix, more leverage, GAAP losses, governance overhang) but arguably over-discounted given VNET's faster wholesale growth (+77% vs GDS +11%) and the CATL catalyst. The valuation gap-close vs GDS is the core value angle. (I did not fabricate EQIX/DLR multiples — marked not-sourced.)

Lens 8 · Stock-Price Catalysts (>5% moves, ~5-yr pattern)

Mostly ``:

• Sep 2014 — Trinity Research short report: −27% over two days on 42× volume, worst since IPO; company rebutted ``. Establishes VNET's China-ADR short-attack history.
• Feb 2022 — Blackstone US$400M convertible (2027 notes + Series A preferred): balance-sheet event ``.
• Dec 28, 2023 — Shandong Hi-Speed US$299M investment (SOE becomes largest holder): stock jumped ~30% ``.
• 2024–2025 — the AI-datacenter re-rating: +208% in 2025 (best-performing datacenter stock YTD), +180% 12-month by Sep 2025, "surged ~700% off the lows on AI demand" . Driven by wholesale revenue +90% (2024) / +77% (2025) and the **ByteDance 500 MW** order .
• Sep 24, 2025 — +14% to $10.62 on AI-demand momentum ``.
• May 13, 2026 — CATL 38% stake deal: surged on the ~US$942M CATL-affiliated purchase from SDHG ``.
• May 2026 (Q1 print) — −3.9%: EBITDA beat but net loss widened on tax ``.

Pattern: the market trades VNET on (1) wholesale/AI order flow and capacity move-ins, and (2) strategic-investor / capital events — not on GAAP EPS (there is none to speak of). It ignores the net loss when EBITDA and orders are strong, and punishes tax/leverage surprises. This is a narrative + capacity stock, priced on the AI-infrastructure story and the balance-sheet backers.

---

Phase C — Judge people & books

Lens 9 · Management

• Founder / Executive Chairperson / interim CEO — Sheng Chen ("Josh" Sheng Chen). Founded the business May 1996, started China's first carrier-neutral DC July 1999; Tsinghua EE; co-runs Tsinghua's Energy Internet Research Institute . **Track record:** built China's carrier-neutral IDC category and the Microsoft-in-China cloud franchise — genuinely pioneering. **But he has been *interim* CEO since April 2024** — i.e. **~2 years with no permanent CEO**, and the 20-F names him as a key-person dependency . That leadership vacuum at the top of a capital-intensive scale-up is a real governance flag.
• Skin in the game — with a catch. Chen controls the vote (directors/officers group = 32.9% economic / 42.3% voting via 10-vote Class B + a proxy over SDHG's Success Flow shares to Dec 28, 2026; plus an authorized-but-unissued 500-vote Class D entrenchment mechanism) . **Red flag:** his holding vehicles (GenTao, Beacon, Fast Horse, Sunrise) have **pledged 68.4M Class A + 27.8M Class B shares as collateral** on personal financing (a prior US$50.25M Bold Ally margin loan settled July 2024, replaced by a Shining Rich promissory note up to US$24M) . Founder margin-pledged shares = forced-selling tail risk if the stock craters — a recurring China-ADR governance hazard.
• CFO — Qiyu Wang (since Jul 2023): ex-VP China Telecom Global, ex-CFO China Communications Services — deep telecom/SOE-financing rolodex, apt for onshore debt + carrier relationships . Principal accounting officer **Zhihua Zhang** leads domestic debt financing + the Microsoft partnership (ex-Anta/Lenovo/TDK) .
• Board quality — a genuine positive. Audit-committee chair Sean Shao is a well-known China-ADR governance figure who chaired the special committee that investigated the Luckin Coffee accounting fraud — a real fraud-buster on the audit chair, plus UTStarcom/Trina Solar CFO history . Directors include Acer co-founder **Kenneth Tai** (now chairman of Zettabyte, a Taiwan-listed AI-datacenter systems firm — relevant expertise) and a Hongshan (Sequoia China) operating partner **David Chen** .
• Capital allocation: aggressive reinvestment (capex RMB3.6B→5.0B→8.2B over 2023–25, guided RMB10–12B for 2026) funded by debt + converts + NCI + ABS recycling; a large 2023 goodwill/long-lived impairment (RMB1.87B) marks prior M&A missteps ``. ROE is negative/volatile (losses). Archetype: founder-operator/visionary re-taking the wheel, backed by SOE then strategic capital — high conviction, high leverage, thin permanent-management bench.

Lens 10 · Forensic Red Flags

Act-as-forensic-analyst; every figure labeled.

• Cash flow vs earnings. OCF RMB1,918.6M (2025) is below D&A of RMB2,126.0M and roughly flat YoY despite +20.5% revenue ``. Earnings quality is carried by the ABS one-time gain (RMB469.8M) and heavy D&A add-backs; underlying cash generation is not accelerating with the top line. Watch.
• Receivables outrunning revenue. AR net +34% vs revenue +20.5%; doubtful-debt allowance RMB189.5M→267.7M; annual doubtful-debt provisions RMB54.5M/106.3M/97.5M (2023–25) — collections lagging, some customer-credit stress ``.
• The tax line is the standout anomaly. 131.5% effective tax rate in 2025 (RMB557.5M on pre-tax income), driven by a valuation-allowance build (RMB241.4M) — i.e. writing off deferred-tax assets the company doesn't expect to use (a signal certain entities are structurally loss-making), plus intra-entity-transfer and Cayman effects. Q1 2026 added RMB486.2M of capital-transaction tax from the ABS restructurings. Tax is eating the P&L ``.
• Convert fair-value volatility. The 2027 notes are carried at fair value → RMB314.3M non-cash loss in 2025 as they marked up; a "critical audit matter" per the auditor ``. Adds non-operating EPS noise both ways.
• ABS deconsolidation gains flatter results. RMB469.8M gain on deconsolidating a DC subsidiary into the Dec-2025 ABS — legitimate capital recycling, but a non-recurring gain propping up the print; strip it and the loss widens ``. Watch for repeat use to smooth results.
• Leverage & funding dependence. Interest-bearing debt ~RMB20.8B (US$2.98B: ST/current RMB3.23B + LT RMB11.58B + redeemable-NCI borrowings RMB1.25B + US$680M converts); Q1 2026 LT debt alone RMB17.77B (US$2.58B). Net debt ~US$2.0–2.1B ≈ ~5x adjusted EBITDA ``. Capex (RMB10–12B guided) is >2× OCF — perpetual external-funding dependence. Onshore RMB debt is cheap (WA rates 3.18% ST / 3.90% LT) but rising.
• Structural China-ADR risks (label, don't hand-wave): VIE contractual control (not equity ownership) of the operating entities; RMB non-convertibility + SAFE controls trapping ~65% of cash onshore (only RMB0.2M of RMB6.58B sits at the parent; ~RMB1.36B offshore) ``; HFCAA/audit-oversight and potential delisting tail; up-to-10% dividend withholding.

Regulatory findings (required) — from regulatory/regulatory-findings.md + 20-F Item 8 + web:

• SEC EDGAR EFTS: No Litigation Releases and no AAERs naming VNET Group in 2021-07-10 → 2026-07-10 ``.
• 10-F/Item 8 (own disclosure): VNET is a defendant in a pending shareholder class action; it has settled shareholder class actions before, and one current action has reached a settlement with a motion for court approval pending ``. Recurring securities-litigation exposure — typical for a volatile China ADR with a short-attack history (Trinity 2014).
• Non-SEC / web: no material FTC/DOJ/other enforcement surfaced; the salient items are the 2014 Trinity short report (rebutted) and the ordinary-course class-action history ``.
• Net: no SEC enforcement or accounting-fraud finding on record; the live legal exposure is shareholder class-action litigation, and the forensic risk is earnings-quality (tax/leverage/one-off gains), not proven fraud. Verified via SEC EDGAR EFTS (LR + AAER), 20-F Item 8, and web as of 2026-07-10.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

VNET is loss-making on EPS and valued on EBITDA, so adjusted EBITDA + EBITDA-to-interest coverage + FCF-after-growth-capex are the right anchors, not EPS. Built bottom-up from FY2025 actuals + FY2026 guidance ``. No forecast.ts logged (watchlist rule).

FY2026 (base = guidance midpoint) ``:

• Revenue RMB11.65B (+17%); adjusted EBITDA RMB3.65B (+22.5%, ~31% margin); capex RMB10–12B.

FY2027 base ``: revenue ~RMB13.4B (+15%, wholesale keeps compounding as 450–500 MW/yr delivers + move-ins ramp toward 80%+); adjusted EBITDA ~RMB4.4B (~33% margin on self-built mix + operating leverage).

FY2028 base ``: revenue ~RMB15.3B (+14%); adjusted EBITDA ~RMB5.2B (~34%).

Bull ``: move-in utilization 70%→85%+, near-100% pre-commitment on deliveries, CATL energy synergies lift margins and lower power cost; FY2028 EBITDA ~RMB6.0B; the ABS/pre-REIT channel funds growth with minimal equity dilution; net-debt/EBITDA falls toward ~3.5x. Re-rates toward GDS's ~15x.

Bear ``: China AI-capex digestion or a hyperscale anchor pushes out move-ins; power/permit constraints slow deliveries; rate/refi pressure (2027 Blackstone converts, 2030 converts) + continued punitive tax; FY2026 EBITDA at the RMB3.55B low end and decelerating; net-debt/EBITDA stuck ~5x; equity raise dilutes.

GAAP net result: likely a small, narrowing loss through FY2026–27 (interest ~RMB700M+ and growing, convert-FV noise, ABS-related capital-transaction taxes), turning to breakeven/positive only once EBITDA growth outpaces interest and the tax drags roll off — plausibly FY2027–28 in the base case ``. I decline to publish a precise EPS figure — it is not a meaningful valuation anchor here and any point estimate would be false precision (n/a).

What to track quarterly: MW delivered & move-in/utilization rate; wholesale order intake (MW); adjusted-EBITDA margin; EBITDA-minus-cash-interest; net-debt/EBITDA; and the CATL deal close (expected Q4 2026).

Lens 12 · Bull vs Bear

Bull case. VNET is the cheapest scaled way to own China's AI-datacenter capacity build. Wholesale/AI revenue compounded +90% (2024) and +77% (2025) and the order book is accelerating (ByteDance 500 MW; 519 MW new orders YTD 2026; 2.48 GW total secured resource; 85.8% pre-commitment) . Adjusted EBITDA is compounding >20% at a widening ~30% margin, guidance is beaten repeatedly, and the retail annuity (0.1% churn) funds the base. It trades ~10x EV/EBITDA — a big discount to GDS's 15x — with a **Strong Buy** consensus and PTs of **US$18–20 (Citi $20)** vs an **US$8.05** price (~2× implied upside) . The kicker: a ~US$1B CATL strategic entry (38% from SDHG, closing Q4 2026) that re-anchors the register with a power/energy champion and validates the AI-infra thesis ``. Capital-light ABS/pre-REIT recycling (RMB7B+ raised) reduces equity-dilution risk.

Bear case — three permanent-impairment risks. (1) Balance sheet: ~5x net-debt/EBITDA, capex >2× OCF, chronic external-funding dependence, and convert maturities (2027 Blackstone / 2030) — a rate/liquidity shock or a stalled AI cycle forces dilution or distress. (2) China-ADR structural tail: VIE (contractual, not equity, control), trapped onshore RMB, HFCAA/delisting risk, RMB depreciation — any of which can permanently impair the equity independent of operations. (3) Earnings never convert to owner cash: a 131.5% effective tax rate, growing interest, and one-off-gain-dependent prints mean adjusted EBITDA may not translate into GAAP profit or free cash flow for years; the "cheap on EBITDA" multiple is a trap if EBITDA doesn't become cash. Pre-mortem (18 months out, thesis broke): the CATL deal slipped or fell through on SDHG-shareholder/regulatory grounds, a hyperscale anchor deferred move-ins, a US$ funding window shut, and the founder's pledged shares got margin-called into a China-ADR risk-off — the stock halved. Contrarian view the market is missing: the bull tape treats VNET as a pure AI-capacity call and ignores that its net income is going backwards — the tax + leverage structure means the AI boom is showing up in EBITDA and the buildout, but not yet in a single dollar of GAAP profit or free cash flow.

Lens 13 · Devil's Advocate (short-seller)

• What structurally breaks it: the model is build capacity → move in hyperscalers → recycle assets via ABS → repeat, all on borrowed money. If AI-capacity demand in China digests (a very real risk given hyperscaler self-build and policy swings), VNET is left with 452+ MW under construction, ~5x leverage, and thinning pre-commitment — the exact 2022–23 over-build setup that produced the RMB1.87B impairment ``.
• Concentration: two customers each 10–20% of revenue; the wholesale growth is a handful of internet giants. Lose or renegotiate one and the growth narrative and the build-to-suit capital both sour ``.
• Weaker moat than bulls think: VNET is a capacity- and price-taker to hyperscalers and a supply-taker from the three SOE carriers (46% top-5). Hyperscalers self-building, and GDS (better assets, 47% margins), are the underestimated competitive pressure.
• Capital-allocation / incentive flags: founder margin-pledged shares, a 2-year interim-CEO vacuum, an authorized 500-vote Class D entrenchment class, prior value-destroying M&A (goodwill impairment), and results leaning on non-recurring ABS gains ``.
• What must hold for US$8 (let alone US$18): uninterrupted AI-wholesale demand, on-time move-ins lifting utilization from ~76% toward ~90%, cheap continuous funding, the CATL deal closing, and no China-ADR risk-off. If growth disappoints 20–30%, EBITDA misses, ~5x leverage becomes ~6–7x, the ABS window narrows, and a ~10x multiple compresses toward the historical distressed range — 40–60% downside is easy to underwrite.
• Single permanent-impairment scenario (and plausibility): a China-ADR delisting/VIE-enforcement or forced-deleveraging event during an AI-capex air-pocket — low-probability but non-trivial for a levered VIE, and it's a zero-or-halve tail, not a dip.

Lens 14 · Management Questions (ordered by information value)

1. The 38% CATL-affiliated purchase is a secondary from SDHG — VNET receives no primary capital. What operational rights, power-procurement or energy-cost synergies, and governance changes does CATL bring, and what are the exact closing conditions and break scenarios (SDHG-shareholder vote, PRC/US regulatory)?
2. Your effective tax rate was 131.5% (2025) and Q1 2026 took RMB486M of capital-transaction tax. What is the normalized cash-tax rate once the ABS restructurings are done, and when does GAAP net income turn positive?
3. Walk us from adjusted EBITDA to free cash flow after growth capex: at RMB10–12B capex, when does VNET become self-funding, and what's the equity-issuance plan if it doesn't?
4. Net-debt/EBITDA is ~5x with 2027 (Blackstone) and 2030 converts outstanding. What is the target leverage, the refinancing plan for the 2027 notes, and the covenant headroom?
5. Two customers are each 10–20% of revenue. Name the concentration trajectory and the take-or-pay / termination terms protecting the build-to-suit capital if a hyperscaler defers.
6. Who becomes permanent CEO, and when? What has a 2-year interim arrangement cost in execution, and how is the succession bench being built?
7. Founder shares are margin-pledged (Shining Rich note). What is the current LTV, the trigger levels, and the plan to de-risk forced-selling exposure?
8. Utilization is 75.7% (wholesale) / 64% (retail). What's the realistic stabilized ceiling and the move-in cadence over the next 8 quarters?
9. How much power/land resource of the 2.48 GW is secured with grid allocation vs aspirational, and where is the binding constraint by market?
10. ABS/pre-REIT recycling raised RMB7B+ at ~30% retained. What is the steady-state cadence, the cost of that capital vs debt, and the accounting for future deconsolidation gains?
11. What is the churn and pricing outlook for the flat retail book — is it a durable annuity or a slowly eroding one as customers migrate to hyperscale/cloud?
12. Microsoft Azure/365-in-China: contract renewal terms, regulatory durability, and its revenue/margin contribution trajectory?
13. Item 8 flags a pending shareholder class action. Status, potential exposure, and D&O coverage?
14. RMB depreciation and trapped-onshore-cash: how do you fund US$-denominated converts and offshore obligations without repatriation, and what's the FX-hedging posture?
15. What single metric should the market hold you accountable to over the next 24 months — and what would you consider a broken thesis?

---