Mara Holdings vs. Soluna Holdings: Which AI Pivot Is More Compelling?

The article argues that Soluna Holdings has a larger AI data center pipeline at 4.3 GW, while Mara Holdings is closer to monetization with 2.2 GW of combined operational and development capacity and a near-term goal of more than 1 GW of IT capacity. Soluna generated $9.4 million in Q1 revenue, up 58% year over year, while Mara posted $174.6 million in revenue, down 18% year over year, with both still reporting heavy losses. The piece is constructive on both names' AI exposure, but favors Mara for lower execution and financing risk in the nearer term.

Analysis

The market is increasingly pricing AI infrastructure as an option on future power optionality, not current earnings. That favors names with the shortest path to contracted utilization and external capital, which is why the relative winner is likely not the company with the biggest pipeline but the one that can convert megawatts into bankable cash flow before financing costs and permitting friction compound. In that framework, near-term execution matters more than headline gigawatts: each quarter of delay pushes equity value further into the terminal-discount bucket and increases dilution risk.

The second-order winner is the rest of the AI infrastructure stack: equipment vendors, switchgear, cooling, and grid-interconnection beneficiaries should see faster order conversion than the operators themselves. If these developers keep winning joint ventures, lenders and infrastructure partners are effectively being paid to absorb duration risk, which can cap upside for common equity even if the projects succeed. That creates a subtle rotation: the more crowded the “AI power” trade becomes, the more value migrates away from speculative development equities and toward toll-booth suppliers and financiers with contractual revenue.

Consensus is probably underestimating how violently financing sensitivity changes with rates over a 12-24 month build cycle. A modest move down in funding costs could re-rate the slower builder disproportionately, while a stickier rate environment would likely compress long-duration project equity multiples first. The real catalyst to watch is not general AI demand commentary but the first credible hyperscaler offtake or project-finance package that turns capacity from narrative into amortizing assets; without that, the stock remains a macro rate bet wearing an AI costume.

The contrarian angle is that larger pipelines can be a liability if they require continuous capital recycling and repeated milestone risk. Investors may be overpaying for optionality that has not yet cleared permitting, interconnection, and financing gates. In contrast, the name closer to monetization may have less upside asymmetry but better path-dependent probability of value capture over the next 6-18 months.

Key Decisions for Investors

• Prefer a relative-value long IREN / short SLNHP basket for 3-6 months: IREN should benefit more from the market rewarding closer-to-cash AI capacity, while SLNHP remains exposed to dilution and project-timing slippage.
• For event-driven exposure, buy NBIS or NVDA on any pullback tied to skepticism around AI demand; these are cleaner second-order beneficiaries of accelerated data-center buildout with less single-project execution risk than operator equities.
• Use call spreads on IREN into the next 1-2 quarters rather than outright equity: upside should be driven by de-risking milestones, but financing and delivery timing can still create sharp drawdowns.
• Avoid chasing SLNHP common until a signed hyperscaler contract or project finance close appears; absent that, the equity is effectively a long-duration funding instrument with poor downside protection.
• If wanting a hedge, pair long GS against short SLNHP over 6-12 months: financing and advisory activity should accrue to capital providers faster than to speculative developers, especially if rates remain elevated.