NEXTDC

Phase A — Understand the business

Lens 1 · Company Overview

What it is. NEXTDC is Australia's largest independent (carrier- and cloud-neutral) data-centre-as-a-service operator: it designs, builds, owns and operates hyperscale and enterprise colocation facilities, then leases power-and-space capacity to hyperscalers, cloud providers, enterprises and government under long-dated contracts. It is a REIT-like infrastructure/real-estate business dressed as a tech stock — it owns the land and buildings (A$3.6bn of owned land & buildings on balance sheet at 1H26 ) and monetises megawatts of critical IT load, not compute itself. The unit of account across the whole business is the MW of contracted utilisation.

How it makes money. Recurring colocation revenue (space + power + cooling), interconnection/cross-connect fees, and cloud on-ramp services. Revenue is a function of (a) built capacity, (b) how much of it is billing, and (c) the lag between contracting a customer and that customer ramping to billing. The gap between contracted and billing MW is the forward order book — pre-sold revenue that converts to cash as customers fit out and power up. NEXTDC guides that ~85% of the forward order book translates to billings/revenue/EBITDA over the following ~2 years.

Scale (the operative numbers).

• Contracted utilisation: 244.8MW at 30-Jun-2025 → 416.6MW at 31-Dec-2025 (+137% YoY) → 667MW on the 20-Apr-2026 update (+60% in a single quarter, +250MW).
• Billing utilisation: 110.9MW (FY25) → 119.8MW (1H26).
• Forward order book: 133.9MW (FY25) → 296.8MW (1H26) → 544MW (Apr-2026).
• Contracted future EBITDA from existing utilisation + forward order book: >A$1.0bn, ~4x+ FY26 EBITDA guidance. This one line is the entire bull case in a sentence.

Customers. Cloud/hyperscale is the dominant demand pool (AWS, Microsoft Azure, Google are the standard tenants of a neutral wholesale operator in AU, though NEXTDC does not routinely name-and-shame individual hyperscalers in releases — customers.csv is empty ). A concrete, named recent win: Sharon AI, a US$950m five-year cloud deployment across multiple NEXTDC Australian sites, with revenue commencing Q3/Q4 2026 — a "neocloud" GPU-as-a-service tenant, i.e. NEXTDC is now landing the AI-native buyer, not just the classic hyperscalers.

Contract structure & payment terms. Long-term (typically 10–15yr) take-or-pay-style capacity contracts with contracted ramp schedules; the forward order book is the contracted-but-not-yet-billing backlog. This is the crucial de-risking feature versus a speculative "build-it-and-they-will-come" developer: NEXTDC is increasingly pre-leasing shells before pouring concrete (the 667MW contracted vs ~120MW billing tells you demand is running years ahead of supply).

Verdict on Lens 1: A high-quality, contracted infrastructure annuity in the making — but one whose reported P&L (persistent statutory losses) badly understates the economic value being locked into the ground, because capex hits slowly via depreciation while the revenue is deferred by the ramp lag.

Lens 2 · Supply Chain

Map: inputs → NEXTDC → end customer. Named stakeholders along the chain (supply-chain.md missing, so this is web-assembled):

• Land & power (the true upstream chokepoint): grid connection is the binding input. NEXTDC's S4 Western Sydney (350MW) is the first data centre approved to directly interconnect to the NSW transmission grid — a structural advantage in a market where the NSW pipeline is 44 sites / 11.4GW and connection queues are the real constraint.
• Construction / EPC: Multiplex (S4 Sydney; four NSW facilities), Kapitol (M2 Melbourne portfolio). Construction partners matter because build-out speed = revenue timing.
• Critical M&E kit: generators, UPS, switchgear, and — critically for AI — liquid-cooling / high-density racks (NEXTDC is marketing 1MW-per-rack densities for its Melbourne facility ). Suppliers here are the usual global M&E vendors (Vertiv, Schneider, Cummins/Caterpillar for gensets) — ``, not company-disclosed.
• Silicon (indirect): NEXTDC does not buy GPUs, but its demand is a derivative of Nvidia's shipment cadence — an AI-training tenant only signs when it has an allocation. This is the transmission mechanism by which "the AI capex cycle" reaches NEXTDC's order book.
• Capital (a supply input for this business): the debt syndicate and equity/hybrid providers are a genuine part of the "supply chain" for a company that consumes ~A$2.7–3.0bn of capex a year. Named: a A$2.9bn syndicated debt platform (Dec-2024 refinance), a A$1.7bn hybrid programme anchored by La Caisse (CDPQ, Québec), and A$1.8bn of new senior debt (May-2026).
• End customers: hyperscalers (AWS/Azure/Google), neoclouds (Sharon AI), enterprise, government/"sovereign" workloads.

Chokepoints / single-source dependencies: (1) grid power & transmission — the number-one constraint on the whole sector, and the thing NEXTDC's incumbency + grid-connect approvals convert into a moat; (2) construction labour & long-lead M&E gear — sector-wide scarcity can slip delivery dates and push out revenue; (3) capital markets access — the business physically cannot fund its build-out from operating cash flow, so a closed equity/debt window is an existential input risk (see Lens 13).

Lens 3 · Competitive Advantages (moats)

The moat is real and it is physical, not technological. NEXTDC does not have proprietary IP; it has irreplaceable brownfield positions, grid connections, and a permitting/planning head start in supply-constrained metros.

1. Power & land oligopoly. Australia's wholesale market is an effective three-name oligopoly — AirTrunk (#1), NEXTDC (#2), CDC Data Centres (#3) — that collectively drives most large-scale capacity. In a market where connection queues, not capital, are the bottleneck, owning already-approved, grid-connected sites is a durable barrier: a new entrant can raise money but cannot conjure a transmission connection in Western Sydney on demand. S4's direct transmission interconnect is the sharpest example.
2. Switching costs / stickiness. Once a hyperscaler or enterprise installs racks, cross-connects to an ecosystem, and certifies a site, migrating is expensive and risky. Colocation churn is structurally low; the "neutral meeting-point" network effect (many networks + clouds in one building) compounds it.
3. Scale & cost of capital. A larger contracted book + investment-grade-adjacent funding lowers the cost of the next MW. The >A$1.0bn contracted future EBITDA gives lenders visibility that a sub-scale developer cannot match.
4. Sovereign/government positioning. NEXTDC leans hard on "sovereign capability" and Australian-owned critical infrastructure — relevant as Canberra tightens data-centre rules and prioritises 15 projects (A$51.9bn) for expedited approval.

Bargaining power — nuanced. Over customers: rising, because power/space is scarce and hyperscalers are desperate for AI-ready capacity — hence the 667MW pre-let. Over suppliers (esp. capital & the grid): weak — NEXTDC needs the debt markets and the grid operator more than they need NEXTDC. That asymmetry is the crack in the moat (Lens 13).

Versus AirTrunk: AirTrunk is larger and, since Blackstone/CPPIB's ~A$24bn 2024 acquisition, sits inside a deep-pocketed private balance sheet — it can out-spend NEXTDC without quarterly-loss scrutiny. NEXTDC's edge is being the listed, liquid, sovereign-branded pure-play with an enterprise interconnection ecosystem AirTrunk (pure hyperscale wholesale) lacks.

Lens 4 · Segments

segments.csv is empty, so no `` segment split is available. From disclosure, NEXTDC reports essentially one operating segment (data-centre services) and does not break out product-level EBITDA. The meaningful cuts are geographic and by facility maturity:

• Geography. Historically ~100% Australia across three metros — Sydney (S-series), Melbourne (M-series), Brisbane (B-series) plus Canberra/Perth/Adelaide. Now a deliberate APAC expansion: KL1 Kuala Lumpur (live; ~A$1bn long-term, 65MW planned) — the first international facility and the fastest-growing contracted site (KL1 hit ~10MW / 15% of planned early); TK1 Tokyo (28MW, with CBRE Investment Management, targeting late-2030); AK1 Auckland (10–15MW, ~A$140m). International is small today but is the multi-year TAM extension.
• Facility maturity (the real driver of margin). Mature metros (Melbourne pro-forma contracted 114MW = 161% of end-2024 built capacity ) throw off high-margin cash; new builds (S4, KL1, S5/S7) drag reported EBITDA while they lease up. The trend is accelerating contracted demand (137% YoY contracted growth into 1H26; +60% in one quarter to Apr-2026), decelerating billing conversion relative to contracting — i.e. the order book is inflating faster than it can be delivered, which is a good problem (demand) that creates a bad optic (billing/contracted ratio collapsing to ~18%).

Cause of the trend: the AI training/inference build-out arrived in APAC. The single-quarter +250MW step-change to 667MW is the fingerprint of one or more very large hyperscale/AI leases landing at once.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print: 1H26, to 31-Dec-2025)

Headline:

• Net revenue: A$189.2m, +13% (vs A$167.8m 1H25).
• Total revenue: A$231.8m, +13% (vs A$205.5m).
• Underlying EBITDA: A$115.3m, +9% (vs A$105.4m). Implied underlying EBITDA margin on net revenue ~61%; on total revenue ~50%.
• Statutory net loss after tax: A$(39.4)m, improved 8% from A$(42.7)m 1H25.
• Capex (half): A$1,285m (vs A$1,003m 1H25) — i.e. the company spent ~11x its EBITDA on capex in six months.
• Balance sheet: total assets A$7.0bn (A$3.6bn owned land & buildings); cash A$278m; undrawn facilities A$3,940m; total liquidity ~A$4.2bn.
• Operational: contracted 416.6MW (+137% YoY); billing 119.8MW; forward order book 296.8MW; 33MW of built capacity added in the half.

Beat/miss vs consensus: revenue and EBITDA were in line / modestly ahead and FY26 guidance (net rev A$390–400m, EBITDA A$230–240m) was reaffirmed; capex guidance was raised to A$2.4–2.7bn (from A$2.2–2.4bn) — later raised again to A$2.7–3.0bn at the Apr-2026 update. Precise consensus deltas: n/a cleanly.

What drove it: volume — new billing capacity ramping + interconnection growth, not price. Margin was roughly flat (new-facility drag offsetting mature-site operating leverage).

Balance-sheet flags: the tension is structural, not a one-quarter blip — A$1.28bn capex against A$115m EBITDA means deeply negative free cash flow, funded by fresh debt + equity. Net-debt/EBITDA sits around the mid-5x range. Cash of A$278m is thin relative to the burn, which is why the A$1.5bn equity raise + hybrid expansion followed in April.

Market reaction: despite the "record" print and record contracted book, the stock made 2026 lows (A$11.44) — the market rewarded none of it, because the print re-confirmed the FCF-negative, dilution-dependent profile rather than resolving it. Read-through: the market is no longer paying for demand; it wants to see the capital cycle turn (a JV monetisation or self-funding inflection).

Lens 6 · Earnings Calls (sentiment trend)

transcripts/ is empty; this is web-derived. Management's narrative arc across the last ~18 months:

• FY24 (Aug-2024): "record result", pivot language to AI-ready capacity, big capex ambition tied to AI demand. Tone: aggressive growth.
• 1H25 / FY25 (Aug-2025): the recurring drumbeat became "record forward order book" and "contracted utilisation" — management deliberately steering attention away from the statutory loss and toward the pre-sold backlog and the >A$1.0bn contracted future EBITDA. New phrase set: "sovereign capability", "AI factories".
• 1H26 (Feb-2026) → Apr-2026 update: tone shifted to urgency and scale — "unprecedented demand", "step-change", accelerating inventory (hence the two capex upgrades in a quarter). The subtext is a management team that has decided land-grab now, fund it, monetise later is the right trade in a supply-constrained window.

Phrases they lean on: "contracted utilisation", "forward order book", "AI-ready", "sovereign", "record". Phrases they've de-emphasised: anything anchoring on statutory EPS/profitability or near-term free cash flow — deliberately, because those numbers don't flatter the story. That selective framing is itself a (mild) yellow flag: the metrics management promotes (bookings) are leading indicators they control the narrative on; the metric they bury (FCF) is the one the bears weaponise.

Lens 7 · Comps

Reading it honestly: NEXTDC is not comparable to Equinix/DLR on earnings multiples because it is pre-profitability by design — it is a build-phase asset, not a stabilised-yield REIT. The correct comp frame is private-market MW value / cap rates: AirTrunk changed hands at ~A$24bn EV (2024), and the S4/S7 JV process is effectively a test of what NEXTDC's ground can fetch at private-infrastructure cap rates. If a JV values NEXTDC's Sydney assets near AirTrunk-style marks, the listed equity is arguably cheap; if it prints at a discount, the bear case wins. The comp that matters is the next monetisation cap rate, not a P/E. I am explicitly declining to fabricate an EV/EBITDA line I could not verify to primary source.

Lens 8 · Stock-Price Catalysts (moves >5% over ~5 years)

Pattern of what actually moves NXT:

• 2020–2021 (COVID digital surge): strong rally on cloud-demand tailwind; revenue +23% FY21.
• 2022 rate shock: de-rating — a long-duration, capital-intensive infra name got hit hard by rising rates (present-value compression). Confirms interest-rate sensitivity is a primary driver.
• May–Jun 2022: Global Switch APAC-bid headlines (NEXTDC hired Macquarie Capital) — an M&A optionality spike (note: this is a 2022 story, not a live 2026 catalyst; do not treat as current).
• Mid-2024: all-time high A$18.20 (9-Jul-2024) on the AI-infra narrative + the A$2.9bn debt facility funding aggressive AI-driven growth.
• Late-2024 → 2025: ~40% drawdown from the peak; DeepSeek-driven "is AI capex real?" risk-off hit picks-and-shovels names first; fell to ~A$11 area.
• Apr-2026: the 667MW / +60% contracted + A$2.2bn capital plan + La Caisse announcement — a demand blockbuster, but paired with a 1-for-5.4 dilutive raise at A$12.70 (8.6% discount to TERP), so the stock did not re-rate; instead it drifted to 2026 lows (~A$11.44).

What the tape reveals: the market reacts to (1) rates / risk-appetite more than fundamentals, (2) dilution events (raises reliably cap the upside from good demand news), and (3) AI-capex-durability sentiment (DeepSeek-type scares). It does not yet reward booking records. The catalyst that would break this pattern is a capital-cycle inflection — a JV monetisation crystallising asset value, or a rate-cut cycle lowering the discount rate on the long-dated EBITDA.

---

Phase C — Judge people & books

Lens 9 · Management

• Craig Scroggie — CEO & MD since June 2012 (~14 years); on the board since the 2010 IPO. 30+ years ICT (Symantec VP/MD Pacific, Veritas, CA, EMC, Fujitsu). Track record: took NEXTDC from a single Brisbane facility (B1) and ~A$1.2m revenue (FY12) to an ASX Top-100, ~A$400m-revenue, 9-facility national operator now expanding across APAC — a genuinely impressive multi-decade build. This is a founder-operator archetype in spirit (long-tenured builder), even if not the legal founder.
• Skin in the game — thin. Scroggie directly owns ~0.068% of shares (~A$6m). For a A$7bn company that is a small personal stake — alignment is via options/incentives and reputation, not a controlling holding. Not disqualifying, but not the founder-with-half-his-net-worth-in-it alignment that de-risks a capital-hungry story. Board/management average tenure ~6 / ~8.9 years.
• Capital-allocation history — the crux, and it cuts both ways. Bull read: he has consistently pre-funded ahead of demand and been proven right (the 667MW book vindicates years of "over-building"); the Dec-2024 refinance materially cut cost of funds; landing La Caisse (CDPQ) as a hybrid + equity anchor (A$1.7bn total) is a top-tier institutional endorsement of the asset base. Bear read: the company has serially diluted shareholders (A$678m placement 2024; ~A$1.3bn 1-for-6 entitlement; A$1.5bn 1-for-5.4 in 2026) and has never generated positive free cash flow — ROIC/ROE have been negative-to-negligible on his watch (statutory losses most years). The honest verdict: excellent operator and asset-assembler, unproven capital-returns generator — the question the whole stock turns on is whether his contracted book finally converts assembled MW into cash ROIC, or whether the treadmill just keeps spinning.
• Red flags: (1) the FY23 prior-period error restatement on capitalisation of borrowing costs (see Lens 10) — a governance/accounting-quality demerit; (2) narrative that steers relentlessly to bookings and away from FCF; (3) low insider ownership. No related-party or promotional-behaviour flags surfaced in web search.

Lens 10 · Forensic Red Flags

Acting as a forensic analyst. All figures `` (no filings on shelf).

• The signature flag — capitalised borrowing costs, and a restatement. NEXTDC capitalises interest on qualifying assets under construction. With a A$2.7–3.0bn/yr build programme, a large slice of interest never hits the P&L — it goes onto the balance sheet and is depreciated over decades. This is standard and GAAP-permitted, but (a) it flatters "underlying EBITDA" (which also excludes D&A) versus economic reality, and (b) NEXTDC restated FY23 for a prior-period error in exactly this area. A capitalisation-of-interest error is precisely the kind of thing a forensic analyst flags, because it sits at the seam between "aggressive but legal" and "misstatement". Watch this line every reporting period.
• EBITDA-to-cash gap is enormous and structural. Underlying EBITDA A$115m (1H26) vs capex A$1.28bn vs statutory net loss A$(39.4)m. "Profitable on paper (EBITDA), deeply cash-negative" is the defining feature — independent research cited FCF margins exceeding −300%. The reliance on external capital is not a risk around the model; it is the model until lease-up outruns build.
• Underlying vs statutory gap. NEXTDC leans on "underlying EBITDA" as its headline. The bridge from A$217m underlying EBITDA (FY25) to a statutory loss is D&A + net finance costs + tax — i.e. the real cost of the capital intensity. Always read the statutory line alongside the promoted one.
• Leverage. Net-debt/EBITDA ~mid-5x against a loss-making statutory base — elevated, and rising as capex front-runs EBITDA. Mitigant: A$4.2bn liquidity, long-dated maturities (FY30–FY33), and >A$1.0bn contracted future EBITDA that de-risks the ratio prospectively.
• Interest income tailwind. Large cash/deferred-raise balances have historically generated meaningful interest income that props the P&L — a quality-of-earnings caveat (non-operating).

Regulatory findings (required sub-section). Per regulatory/regulatory-findings.md: NEXTDC has no CIK and is not an SEC filer, so no EDGAR Litigation Releases or AAERs are searchable — total_sec_findings: 0 reflects inapplicability, not a clean SEC record. Non-SEC / web check for "NEXTDC" (FTC OR DOJ OR FDA OR CFPB OR consent decree OR settlement OR fine OR penalty): no material enforcement actions, consent decrees, fines or penalties surfaced in web search as of 2026-07-06. The only accounting-integrity item of note is the self-disclosed FY23 prior-period restatement on capitalised borrowing costs (above) — a financial-reporting correction, not a regulator-driven action. Australian-specific: the live regulatory theme is grid/connection rules (AEMC draft rule on large inverter-based loads; NSW/Commonwealth data-centre approval tightening) — a sector operating-condition, not an enforcement matter against NEXTDC. Conclusion: No material regulatory or legal enforcement findings — verified via the (inapplicable) SEC EDGAR path, web search (FTC/DOJ/etc.), and public accounts commentary as of 2026-07-06; one self-disclosed prior-period accounting restatement (FY23) stands as the sole reporting-quality blemish.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (FY26 → FY28)

Built bottom-up from guidance + the contracted book. NEXTDC is loss-making at the statutory line, so the honest projection is on revenue/EBITDA + the path to statutory break-even, not an EPS series — an EPS "estimate" here would be a fabricated precision. Everything below with arithmetic; anchors are.

Revenue & EBITDA (A$m):

• FY26 (guided): net revenue A$390–400m; underlying EBITDA A$230–240m. Base: A$395m / A$235m.
• FY27: as the forward order book (~544MW contracted vs ~120MW billing) converts (~85% flows through over ~2yrs), billing MW should step materially. If ~80–120MW of net new billing lands over FY26–FY27 at mature-site economics, net revenue plausibly reaches ~A$470–560m and underlying EBITDA ~A$290–360m. Range is wide because the ramp timing is the single biggest swing variable.
• FY28: with S4 (350MW) and international sites contributing, ~A$580–720m net revenue and ~A$360–460m underlying EBITDA. The >A$1.0bn contracted future EBITDA implies the multi-year trajectory is well above these interim points once fully ramped.

Base / Bull / Bear (FY27 underlying EBITDA, A$m):

• Bull ~A$360m: order book converts fast, mature-site leverage bites, KL1 ramps ahead of plan.
• Base ~A$310m: steady conversion, ~mid-single-digit MW adds per half, margins flat.
• Bear ~A$260m: ramp delays (construction/grid slippage), higher finance costs eat the EBITDA gain, no operating-leverage inflection yet.

Statutory break-even: the real inflection investors should track is when billing MW is large enough that operating cash flow covers interest + maintenance capex — plausibly not until billing utilisation roughly doubles from ~120MW, i.e. FY28–FY30 on current cadence. Until then, statutory losses and external funding persist.

Per SKILL, forecast.ts create is skipped in the --watchlist loop (no genuinely committed base case logged as a Brier forecast). The trackable binary I would register when promoting this: "NXT FY27 underlying EBITDA ≥ A$300m", p≈0.55.

Lens 12 · Bull vs Bear

Bull case. NEXTDC owns the scarcest thing in APAC AI-infrastructure — grid-connected, permitted, brownfield megawatts in supply-choked metros — inside a rational three-name oligopoly. Demand is no longer a question: 667MW contracted, 544MW forward order book, >A$1.0bn contracted future EBITDA (>4x FY26). Landing La Caisse (CDPQ) as a A$1.7bn anchor and running an S4/S7 JV process signals that private-infrastructure capital values these assets far above where the listed equity trades — a monetisation event could crystallise NAV and re-rate the stock, potentially at AirTrunk-style (~A$24bn EV) marks. The founder-operator has been building ahead of demand for a decade and is finally being proven right. Secular tailwind: AI training + inference + sovereign-AI mandates in Australia are a decade-long demand curve, and power scarcity means incumbents capture disproportionate value. Earnings-surprise vector: a hyperscaler mega-lease or a JV struck at a fat cap rate.

Bear case (2–3 things that could permanently impair).

1. The funding treadmill breaks. The model requires perpetual external capital (FCF < −300%). If risk appetite for AI-capex sours (DeepSeek-style), or rates stay high, or a raise fails/prices punitively, the equity is diluted repeatedly or growth stalls — "outside capital can change its mind" is the whole risk in one line.
2. ROIC never shows up. If capex keeps front-running EBITDA and mature-site returns disappoint (power costs, price competition from AirTrunk/CDC/Goodman), the company stays a permanently-loss-making capital sink that never earns its cost of capital — in which case the historical premium multiple is unjustifiable and the stock re-rates down to asset value, not up.
3. Grid/power delays or policy drag. Tightening NSW/Commonwealth connection rules and the AEMC inverter-load rule could slow the ramp, pushing revenue right and lengthening the cash-burn window.

Pre-mortem (18 months out, thesis broke — what happened?): Rates stayed higher-for-longer, a global AI-capex air-pocket hit sentiment, the S4/S7 JV either didn't close or printed at a disappointing cap rate (revealing the assets are worth less than bulls assumed), and NEXTDC had to do another dilutive raise into a weak tape — the stock is at A$8, and the "record contracted book" is still mostly un-billed.

Are multiples too high? On any earnings metric, NEXTDC screens "expensive/negative" — but that's the wrong lens (it's pre-stabilisation). The right question is whether book value / MW value is fair and whether the JV validates it. Today the market is pricing skepticism (2026 lows despite record bookings), so the multiple is arguably de-rated, not stretched — the debate is asset value, not P/E froth.

Contrarian view (what the market refuses to see): The market is treating NEXTDC as a speculative AI momentum name and punishing the dilution — but it may be mispricing a de-risked infrastructure annuity whose contracted backlog (>A$1.0bn future EBITDA, ~85% conversion) makes the forward cash flows far more certain than a loss-making tech stock's. If La Caisse — one of the world's most sophisticated infra investors — is anchoring A$1.7bn and a JV process is live, the smart money is validating the asset base while the listed market sells the quarterly loss. The re-rate trigger is a monetisation print, not a better P&L.

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

Dismantling the bull case.

• What structurally breaks the model: it doesn't self-fund. FCF is negative by hundreds of percent and only turns if lease-up outruns capex — but management keeps raising capex guidance (A$1.8–2.0bn → A$2.2–2.4bn → A$2.7–3.0bn in months), pushing self-funding further out. A business that must sell equity below TERP (A$12.70, 8.6% discount) to fund growth is a price-taker in the capital markets — the most dangerous position for an equity holder, because your ownership is continually diluted at whatever price the market dictates in a downturn.
• Concentration risk: the 667MW contracted book almost certainly leans on a handful of hyperscale/AI mega-leases (the +250MW single-quarter jump screams one or two giant deals). If one anchor tenant renegotiates, delays fit-out, or (in the neocloud case — Sharon AI) proves financially fragile, a large slab of "contracted" backlog is softer than it looks. Contracted ≠ billing ≠ cash; the ~18% billing/contracted ratio is the tell.
• The moat may be weaker than bulls think: power scarcity is a tailwind for all three incumbents — AirTrunk (Blackstone money), CDC (Infratil), and now Goodman Group (A$11.3bn dev pipeline, ~75% data centres, scaling 300→500MW) are all chasing the same hyperscale MW. Blackstone-backed AirTrunk can out-fund NEXTDC without caring about a quarterly loss. NEXTDC's "oligopoly" could become a capital arms race NEXTDC is the weakest-balance-sheet participant in.
• Accounting soft spots: capitalised interest (with a prior restatement in that exact line), "underlying EBITDA" headlining over a statutory loss, and interest income flattering the P&L. None are fraud; all reduce earnings quality and give the bull case a rosier gloss than cash reality.
• What must hold for today's price: that (a) the order book converts to billing on schedule, (b) rates fall / risk appetite for AI-infra holds, and (c) a JV monetises assets at a premium. If growth disappoints by 20–30% (ramp slips, one anchor wobbles), FY27 EBITDA lands nearer A$260m, the mid-5x leverage creeps toward danger, and the equity needs another raise — a reflexive down-spiral.
• The single scenario that permanently impairs: a sustained AI-capex retrenchment (hyperscalers cut data-centre spend) coinciding with higher-for-longer rates — NEXTDC is left holding half-built, half-let capacity it cannot fund to completion, forced to sell assets into a buyer's market. Plausibility: low-to-moderate, but non-trivial, and it is the tail the mid-5x leverage makes lethal rather than merely painful.

Lens 14 · Management Questions (ordered by information value)

1. What is the precise timeline for the S4/S7 JV monetisation, and at what implied cap rate / A$-per-MW would you strike it — and would proceeds fund growth or de-lever? (This single answer most changes the thesis — it tests private-market asset value vs the listed multiple.)
2. Of the 667MW contracted, what share sits with your top three customers, and what are the take-or-pay / termination protections if an anchor delays or cancels fit-out?
3. Walk us through the path to statutory break-even and positive free cash flow — at what billing-MW level does operating cash flow cover interest + maintenance capex, and in which fiscal year?
4. Capex guidance has risen from ~A$1.8bn to A$2.7–3.0bn in one year. What is the maximum annual capex you'd fund, and what leverage ceiling (net-debt/EBITDA) will you not breach?
5. What is your expected stabilised ROIC per new MW, and how does it compare to your marginal cost of capital today?
6. How much of "underlying EBITDA" is interest income on raise/deferred-cash balances, and how do you think about earnings quality ex that item?
7. On the FY23 restatement of capitalised borrowing costs — what changed in controls, and how do you assure investors the current capitalisation policy is conservative?
8. What is the realistic conversion timeline for the 544MW forward order book into billing, quarter by quarter, and what has historically caused slippage?
9. How exposed is the ramp to NSW/Commonwealth grid-connection queues and the pending AEMC inverter-load rule — which specific projects are gated on transmission?
10. AirTrunk (Blackstone), CDC (Infratil) and Goodman all have deep capital and are chasing the same hyperscale MW. Why do you win, and does your listed-equity funding model disadvantage you versus their private balance sheets?
11. What are the economics and ramp risk of the neocloud cohort (e.g. Sharon AI's US$950m deal) versus investment-grade hyperscalers — how do you underwrite counterparty risk there?
12. What returns and strategic rationale justify international expansion (KL1, TK1, AK1) versus concentrating capital on higher-certainty Australian metros?
13. How should we think about power procurement and energy cost as a share of the cost base as AI density rises to ~1MW/rack — and your exposure to Australian electricity prices?
14. What is the dividend/return-of-capital philosophy once the business self-funds, and when might that begin?
15. If the equity window closed for 12 months, could you complete your contracted commitments from existing liquidity + debt alone, and what would you cut first?

---