Galaxy Digital · GLXY

The thesis in one paragraph.

Galaxy Digital is a four-business conglomerate trading at ~40% discount to a defensible sum-of-parts intrinsic value of ~$48/share, driven by a market that struggles to reconcile a cyclical crypto trading parent with a long-duration, A3-tenanted, 90%-EBITDA-margin data center asset. The Helios CoreWeave lease (526 MW of contracted critical IT load, 15-year triple-net, $1B+ avg annual rent) is the dominant value driver — and the additional 830 MW of ERCOT-approved capacity is an under-priced call option on AI infrastructure scarcity. We rate BUY with conviction 4/5, 12-month base target $38 (+28%), bull $68 (+130%), bear $18 (–39%).

Three things matter for the 24-month thesis. First, the Phase I data hall went live in April 2026 with revenue starting to recognize in Q2'26 — this transforms the income statement from "crypto-trading volatility plus capex burn" to "crypto-trading volatility plus $240M of contracted lease revenue scaling to $720M+ by 2028." Adjusted EBITDA inflects from –$188M in Q1'26 to a +$90M run-rate (per management's pre-earnings April disclosure) and onward to consensus $649M by 2027 — a $1.3B annual EBITDA swing over 24 months. Second, the 1 January 2026 ERCOT approval of an additional 830 MW (taking total approved Helios capacity to 1.63 GW vs the 800 MW committed to CoreWeave) created roughly 1,000 MW of uncommitted optionality on a site that already has substation, fiber, and water entitlements. At Applied Digital's Ellendale comp ($1.85M/MW/year NNN to CoreWeave), monetizing even half of that headroom adds ~$5–8B of NPV. The market is not pricing this option. Third, the structural disconnect between GLXY's $14.6B EV and Nebius's $52.6B EV (against 3,500 MW contracted, 6–7× forward ARR) reveals not that GLXY is overvalued at the company level, but that Helios specifically is being valued like a colocation REIT, not like an AI-infrastructure call option. The asymmetry resolves through milestones, not multiples.

"These are 15-year contracted cash flows that will meaningfully diversify our revenue and earnings profile … the convergence of BTC and AI is the single most important macro trend of 2026." — Mike Novogratz, Q1 2026 earnings call, 28 April 2026.

Four businesses under one founder-controlled holding.

Galaxy reports four functional segments: Global Markets (institutional spot/derivatives/OTC, prime, lending, investment banking), Asset Management (GAM) (passive ETF via the Invesco-Galaxy BTCO partnership, plus active hedge fund and venture/PE mandates), Digital Infrastructure Solutions (Helios HPC/AI data center plus a residual mining tail), and Treasury & Corporate (principal balance-sheet crypto and venture investments). The Q1'26 segment data — taken at the cyclical trough of crypto — exposes the gear ratio between trading volatility and the latent landlord franchise that is about to flip on.

The compositional reality: the Digital Assets segment (Global Markets + GAM + Mining + Treasury) delivered $505M of adjusted gross profit in FY'25 — up 67% from $303M in FY'24 — driven by record Q3 2025 trading volumes during the post-election BTC rally. The trading desk's 1,691 institutional counterparty count by Q1'26 is a clean proxy for franchise momentum; the $1.43B average lending book is the credit-intensive layer earning spread income on tri-party repo-equivalent crypto collateral. Global Markets quarterly adjusted gross profit has ranged from $30M (Q4'25, cyclical trough) to $295M (Q3'25, cyclical peak) — a near 10× spread that explains why the market refuses to capitalize trough earnings at peer-bank multiples. GAM is more linear: $6.4B managed AUM at YE'25, $5.0B at Q1'26 (decline driven by BTC MtM, not redemptions — Q1 saw +$69M net inflows), with a 65bps blended fee rate generating ~$40-50M of run-rate management revenue. Compensation-to-adjusted-gross-profit was 59% in FY'25 — a Wall Street-style ratio that signals the cost structure scales with revenue but compresses to single-digit comp-to-EBITDA ratios once Helios's nearly headcount-free $650M+ of stabilized lease EBITDA layers on.

Helios — the segment that actually drives the valuation — was de-minimis on Q1'26 financials ($3.1M adjusted gross profit, –$0.9M EBITDA) because the first data hall did not deliver until April 2026. This is the cleanest possible setup: a quarter that fully captured the bear case (crypto down 20%, treasury bleeding $140M) while the upside engine had not yet turned over. Phase I full delivery (133 MW critical IT, $240M Year-1 rent) is targeted for end of Q2 2026; Phase II (260 MW) for H1 2027; Phase III (133 MW) for 2028. The mining footprint that historically operated at the Helios site has been almost entirely repurposed for HPC — Galaxy now holds approximately 2 EH/s of residual mining (<5% of company EV attribution), which we value at $80–120M in the SOTP as a strategic-orphan run-off line item, a 40-50% discount to the pure-mining peer EV/EH range of $80-110M observed at CLSK, MARA, and BITF.

Each business deserves its own multiple.

A consolidated EV/EBITDA on GLXY is meaningless — the Treasury MtM line alone moves $200M+ per quarter. The correct framing is segment-by-segment, with multiples sourced from the cleanest available comp universe. Below: each business, the comp set, the multiple range, and the equity contribution to bear/base/bull. Toggle the scenarios on the chart to see how the value stack rebuilds.

The SOTP delivers a base-case equity value of $9.2B on current segment economics — implying $23.53/share on a fully-diluted 390M share count (which includes the 198.4M Class B LP units that exchange 1-for-1 with Class A, and which a careless reading of public market-cap quotes will miss). The base 12-month price target of $38 layers in the milestone-driven re-rating: as Phase I cash recognizes through Q2-Q4'26, the market discounts a stabilized Helios EBITDA run-rate of $270M (Phase I alone, 90% margin on $300M avg rent) at progressively tighter cap rates — from the current ~12% (skeptical) toward the 7-8% peer cap rate for contracted hyperscaler data center landlords. That cap rate compression alone moves the Helios segment EV from $3.5B (bear) to $7B (base) to $12.5B (bull), independent of any tenant addition on the 830 MW headroom. The bull case adds explicit credit for a single new 400 MW tenant deal at Ellendale-comparable terms, worth ~$3B of incremental NPV.

Each milestone is worth a precise dollar of equity.

The Helios story is not a single deal — it is a sequence of nine de-risking events that each unlock a discrete piece of equity value. Below we map every milestone to its incremental EV contribution. The bridge starts at the pre-Phase I delivery EV (March 2026) and walks forward through Phase III completion and the monetization of the 830 MW headroom. Toggle milestones on/off to see the assembled valuation.

The valuation math behind these milestone steps follows directly from disclosed lease economics. Phase I generates $240M of Year-1 rent on 133 MW critical IT, escalating at 3% annually — total 15-year revenue of $4.5B. At 90% lease-level EBITDA margins (the explicit guidance from the Q1'26 call), Phase I stabilized EBITDA reaches $270M. At a 7% cap rate (in-line with Equinix/Digital Realty trading multiples for contracted, A3-tenanted hyperscale assets), Phase I's stabilized asset value is $3.86B, less ~$1.4B of project debt drawn → ~$2.5B Phase I equity at stabilization. Phase II (260 MW IT) at the same per-MW economics yields ~$650M of revenue and ~$585M of EBITDA, worth ~$8.3B EV stabilized, less ~$3.4B of incremental debt → ~$5B incremental equity. Phase III (133 MW IT) mirrors Phase I economics → another ~$2.5B equity. The cumulative Helios equity at full 526 MW stabilization is therefore ~$10B on a base case, before adjusting for the 830 MW of uncommitted ERCOT capacity. The headroom is harder to value — it depends on tenant demand timing and Galaxy's willingness to finance the build — but at $5-6M/MW of "developed land + power + interconnection" optionality (the Applied Digital land-bank comp), 830 MW is worth $4-5B of pre-lease optionality even before any tenant signs.

Numerical DCF with explicit assumptions and a live sensitivity heatmap.

We build a five-year explicit forecast (2026–2030) plus a terminal value, with WACC ranging from 8.5% (predominantly contracted lease cash flows) to 11% (volatile crypto trading). The terminal year reflects steady-state Helios at $1.05B revenue × 90% EBITDA, plus Digital Assets normalized at $400-500M EBITDA. Probabilities: Bear 20% / Base 55% / Bull 25%. Use the heatmap to test alternative WACC × terminal-growth assumptions.

The DCF construction is mechanical but the key sensitivity is not WACC — it is the cap rate at which the market chooses to capitalize Helios's contracted EBITDA. At a 10% cap rate, Helios 526 MW stabilized EBITDA of $950M values the asset at $9.5B EV; at 7%, it values at $13.6B; at 6%, $15.8B. The current implied cap rate baked into GLXY's $14.6B EV is approximately 11–12% for the contracted Helios cash flows — that is, the market is treating Helios as a development-stage data center asset, not a stabilized, A3-tenanted lease. As Phase I cash recognition begins to flow through the income statement in Q2-Q3 2026 (the consensus 2027 adjusted EBITDA jump from $169M to $649M is essentially the Helios stabilization), the cap rate compresses mechanically. Each 100bps of cap rate compression on Helios adds ~$1.0B of equity value, or ~$2.56 per share.

The same MW does not pay the same dollar — and that is by design.

Nebius (NBIS) at $52.6B EV on 3,500 MW contracted trades at $15M/MW. IREN at $25.4B EV on 5 GW secured (480 MW deliverable) trades at $5.1M/MW of secured power but $53M/MW of in-year contracted MW. GLXY at $14.6B EV on 526 MW contracted IT trades at $28M/MW. The headline ratios are not comparable. The economic ratios are.

* GLXY full-company EV; the Helios segment alone is implicit ~$7-8B per base SOTP.

Why the gap exists — six structural reasons.

1. Revenue density. NBIS captures $10M of ARR per MW; Helios captures $1.9M of rent per MW. The 5.3× gap is not arbitrage — it is the difference between selling GPU-hours (the operator) and selling square feet of conditioned data hall (the landlord). The operator monetizes the chip; the landlord monetizes the building. Even after applying Helios's 90% EBITDA margin vs NBIS's 45%, NBIS still earns $4.5M of EBITDA per MW vs Helios's $1.7M — a 2.6× advantage that is permanent and structural.

2. Compounding upgrade cycle. NBIS owns the GPUs. Every 18-24 months, the H100 fleet gives way to H200, then GB300, then Vera Rubin — and NBIS swaps the silicon and re-bills the customer at the higher rate per GPU-hour. The landlord captures none of this. Helios's 3% annual rent escalator is a fixed-rate annuity (a good one — but an annuity). CoreWeave gets the compounding chip improvements; Helios gets CPI plus 1%.

3. Sentiment overhang from the parent. GLXY is a crypto-trading firm with a data center, not a data center with a crypto-trading subsidiary. The Treasury & Corporate segment marks $1.36B of digital assets to market every quarter and posted a $140M loss in Q1'26 from BTC depreciation. The Global Markets desk's adjusted gross profit swung from $295M (Q3'25) to $30M (Q4'25) — a 90% decline in 90 days. Pure-play AI infrastructure investors discount this volatility heavily; the equity moves on crypto headlines, not on data center milestones. That is a 20-30% multiple discount we estimate vs a hypothetical Helios spin-out.

4. The LUNA scar tissue. Galaxy's $200M settlement with the NY AG (March 2025) over Terra/LUNA-related misrepresentations is structurally provisioned ($60M payments due 2027 and 2028), but the franchise damage is harder to quantify. Institutional buy-side memos still cite this as a "regulatory overhang" in the BUY/HOLD margin, particularly at Goldman ($24 PT, Neutral). NBIS has no such overhang. IREN has no such overhang. This is unique to GLXY.

5. Founder concentration. Mike Novogratz controls 51.9% of the economic interest through Galaxy Group Investments LLC. He sold 3.5M shares in October 2025 at ~$36 — above the current $29.62 — which raised questions about founder conviction on the near-term cycle. Founder-controlled cap structures with insider selling at higher prices tend to attract small but persistent governance discounts. NBIS's float is wider; IREN's is more dilute (52% share count growth YoY but no concentrated founder block).

6. Capital structure unfinished. Phase II/III aggregate capex of ~$5.1B is unfunded. Management's plan is to refi Phase I once stabilized, unlocking equity. But this depends on (a) Phase I delivery going smoothly, (b) macro credit conditions for HPC project debt remaining accommodating, and (c) Helios cap rates compressing to the 7-8% range that justifies an aggressive refi quantum. The $445M of exchangeable notes due December 2026 adds near-term refinancing tail risk. NBIS has $9.3B of cash (vs $9.6B of debt — net debt $217M); IREN raised $2.6B in fresh converts in May 2026 and holds $2.6B cash. GLXY's cash + stables of $2.6B is comparable in absolute terms but smaller relative to its remaining capex commitment.

What price is the market currently asking us to believe?

At $29.62 and a fully-diluted 390M share count, GLXY's market cap is $11.6B and EV is $14.6B. Strip out the segments we have high conviction on (Global Markets $1.2B + GAM $0.8B + Treasury $1.1B + Mining $0.1B = $3.2B), and the implied Helios EV is approximately $11.4B — which sounds aggressive until you do the math.

The market is currently pricing Helios at an implied cap rate of ~11% on the contracted 526 MW stabilized EBITDA. At Equinix's 5-6% cap rate or even Digital Realty's 6-7%, the contracted Helios is mathematically worth $14-17B alone, not $11B. The gap reflects three things the market is discounting: (i) execution risk on Phase II/III delivery and financing, (ii) tenant concentration risk on CoreWeave (now A3 IG but still a single counterparty), and (iii) the conglomerate discount from owning Helios inside a crypto-trading parent. Of these, only (iii) is unfixable on a 24-month horizon. Items (i) and (ii) compress mechanically as milestones land — and the catalyst calendar is dense between Q3'26 and end-2027.

If we take the reverse direction — solving for the implied EBITDA CAGR that the market is currently pricing — we get the following: at a stabilized peer multiple of ~12× EV/EBITDA (the blend of EQIX 22×, DLR 25×, APLD 8×, CRWV 14×) and current EV of $14.6B, the market is implying ~$1.2B of 2028E EBITDA. That is in line with our base case ($1.0-1.3B at full Phase III stabilization in 2028) but well below the bull case ($1.5-1.8B if the 830 MW headroom is partially monetized). The asymmetry is real and quantifiable.

Twelve risks, ranked.

BUY · conviction 4/5 · 12-month base target $38.

The 24-month event horizon.

The verdict.