What LPs Are Actually Pricing Into 2026 Data Center Infrastructure Funds

What changed in LP diligence

Through 2023 and into 2024, the dominant pitch structure for data center infrastructure funds leaned heavily on demand narrative. AI compute was growing. Hyperscalers were committing to multi-year capacity roadmaps. Power-constrained markets were driving up pricing. The story practically wrote itself, and LPs who knew little about data center engineering were allocating based on the macro thesis.

That era ended in late 2024. A small number of high-profile projects came in over budget or behind schedule. One CMBS structure settled at spreads that surprised credit analysts who had assumed the sector's infrastructure designation would compress risk premiums further. Two development-stage funds raised against demand projections that didn't materialize into signed off-take within the commitment window.

The LP community - pension funds, insurance allocators, sovereign wealth, and family offices writing $1M-$50M checks - started asking different questions. Not "is the macro trend correct" but "how does this specific project clear the execution risk?" The shift wasn't uniform and it didn't happen overnight. But by Q1 2026, the average diligence timeline on new data center infrastructure fund commitments had extended meaningfully, and the checklist driving those timelines had professionalized.

The projects clearing that checklist fastest are those that answer three specific questions before the LP asks them. The projects stalling in diligence are those still leading with the macro thesis.

The three questions that clear the room

Experienced infrastructure allocators at pension funds and insurance companies have converged on a short list of project-level questions that separate fundable opportunities from narrative-dependent ones. These questions aren't academic - they map directly to the risk categories that caused the 2024 project failures.

The demand-first model and why it's winning capital

The projects that are closing LP capital fastest in 2026 share a structural characteristic: demand commitment precedes construction, not the other way around. JLL's data center investment outlook shows 87% of capacity pre-leased before delivery in primary markets. That's not just a statistic about the market tightness - it's a description of the underwriting logic that institutional capital now requires.

When off-take is LOI-backed before construction financing closes, the project's risk profile changes in concrete ways. Revenue is semi-contracted rather than speculative. The time-to-revenue window compresses because there is no leasing period between delivery and first dollar - the tenant is waiting for the building. The construction loan underwrites against committed demand rather than projected absorption.

For LP allocators underwriting against a return target, the difference between a contracted infrastructure asset and a speculative development is not a matter of degree - it's a different risk category. Insurance pension allocators, who have historically stuck to core and core-plus infrastructure (toll roads, airports, regulated utilities), are extending into data center infrastructure specifically because the demand-first model gives these assets the contracted cash flow profile that resembles what they already hold.

The implication for fund managers pitching institutional LPs is direct: a project without signed off-take before the fund raise is a development bet, not an infrastructure allocation. Many LPs are now explicit about this distinction in their mandate language.

Where family offices fit differently than institutional funds

The LP universe for data center infrastructure isn't monolithic. Pension funds and sovereign wealth funds operate on return targets, mandate constraints, and governance timelines that shape how they can access this asset class. Family offices - particularly single-family offices with direct investment capability - are approaching the same opportunity from a structurally different position.

The check sizes are different: family office commitments to project-level data center opportunities typically run $500K to $5M, below the threshold that most large-scale fund vehicles are built to accommodate. The liquidity requirements are different: family offices can hold illiquid direct investments through multi-year construction timelines in ways that public pension funds, constrained by beneficiary liquidity obligations, cannot always justify. The return threshold is often different: a 20-25% IRR target that a PE-style infrastructure fund requires to justify its fee structure may be less relevant to a family office that is underwriting purely on absolute return and portfolio correlation.

This is creating a distinct market segment: project-level co-invest structures and smaller direct SPV vehicles designed for family office and HNWI capital that want direct exposure to data center infrastructure economics without going through a fund with 2-and-20 fee drag. The SuperReturn Amsterdam and World Family Office Forum Montreux gatherings in June 2026 are reflecting this - data center infrastructure is one of the more discussed real asset categories, and the discussions are specifically about direct and co-invest access, not fund fund allocation.

The diligence questions at the family office level are the same three - power, off-take, execution - but the decision process is faster and the tolerance for direct developer relationships is higher. Family offices are typically not running LP advisory committee approval processes. If the GP can get in front of the principal with a credible project, a clean power story, and signed LOI documentation, the path to a term sheet is shorter than with institutional capital.

What the $2 trillion AI spending forecast actually means for underwriting

Gartner's September 2025 forecast of $2.02 trillion in global AI spending in 2026 has become the most-cited number in data center investment decks. It's real, it's significant, and in most fund pitches it's doing a job it can't fully support.

The macro figure tells LPs that the demand environment is favorable. It doesn't tell them whether this specific project has access to that demand, or at what point in the supply chain. A $2T AI spending forecast benefits the hyperscalers who are procuring GPU clusters and the GPU cloud operators who are building out inference capacity. Whether any of that spending flows through a specific 20 MW colocation campus in a secondary US market depends on that campus's tenant relationships, its connectivity profile, and its position in the operator procurement hierarchy - none of which are visible from a Gartner forecast.

Sophisticated LP allocators are now separating the macro from the project-specific. "AI demand is growing" has become table stakes - it's the minimum to get a meeting. The meeting itself is about how this specific project captures demand from that macro. The fund managers advancing fastest in LP processes are the ones who come with the macro context already abbreviated and spend their time on the project-specific evidence.

The construction cost advantage that fund decks underuse

One comparison that appears in more fund decks in 2026 than in prior years is the build cost differential between modular and traditional construction approaches. The gap is significant: modular data center construction costs run $4-8M per MW at utility scale, versus $11.3M per MW for traditional large-campus construction. That's a 30-65% cost reduction depending on specifications and scale.

What's changed in LP diligence is that this cost figure is no longer taken at face value. The first follow-up question is about operational equivalence - does the modular approach deliver comparable PUE, comparable connectivity, comparable power quality? The second question is about off-take acceptance: will the tenant sign a lease on a factory-built facility with the same terms they'd sign on a traditional campus? The third is about warranty and lifecycle: what happens to a modular facility in year 12 if the original manufacturer has changed product lines?

Funds that have done the work on those follow-up questions are in a stronger position than those presenting modular cost figures without the supporting technical documentation. The LPs who have looked at this most carefully are the ones treating modular as a legitimate capital efficiency argument rather than a speculative cost projection.