As has been widely reported, digital infrastructure has become one of the fastest growing investment structures in recent years, most recently driven by the explosion in demand from firms in the artificial intelligence (AI) industry. This in turn has led to unprecedented needs for capex spending for the construction, expansion and upgrading of data centers, cell towers and networks, fiber optics and other data transmission facilities and power production and transmission.
AI infrastructure strategies seek to build and profit from operating long-lived physical assets, primarily data centers. These assets include the buildings themselves, the advanced computer chips and equipment within those buildings, the internal structural support, specialized cooling and other technical infrastructure necessary to operate that equipment and all of the necessary power and data transmission connections, all of which involves rapidly evolving technical, regulatory, local permitting and other requirements, not to mention coordination among a host of suppliers, operators and other counterparties. For a private fund structure capitalizing such complex projects, that mix tends to concentrate attention on (i) how the strategy is defined, (ii) how offering and marketing materials describe assumptions and underwriting inputs, (iii) how conflicts are managed, (iv) how asset-level contracts allocate remedies and operational downside risks and (v) how the fund and the assets could be impacted by trade regulation and foreign policy considerations.
From a US federal securities law perspective, the core discipline is familiar: investor communications should be accurate, complete in context and not misleading. From a private civil litigation perspective, obligations and claims can be framed through state-law contract principles, including under LPAs and side letters, and related theories tied to statements made in PPMs or otherwise during fundraising or ongoing reporting. The specifics of AI infrastructure investments add complexity to those baseline frameworks.
Defining Strategy and Scope
“AI infrastructure” is not a standardized asset class, and the label can mean different things across sponsors, investors and intermediaries. Depending on the mandate, it may include fully provisioned data centers, “powered shell” projects (buildings with power capacity but no servers), connectivity adjacent to compute sites, power procurement and grid interconnection arrangements and investments or financings tied to critical compute hardware and enabling equipment. It may also include, or exclude, adjacent software and investments in venture-stage businesses.
Because the term is elastic, alignment often starts with defining the scope in fund documents and marketing materials. Where AI infrastructure will be a material focus, fund sponsors and investors can benefit from well-aligned expectations, derived from clear descriptions of what is in-scope and out-of-scope, the expected mix of development-stage versus operating-stage exposure and any stated constraints or flexibility regarding leverage, structured products, co-investments, warehousing and related matters. If the strategy is intentionally flexible, then setting out guidelines for prioritized assets can reduce later confusion as to mandate scope.
Clear scope drafting can also help manage expectations for both fund sponsors and investors regarding ways in which the portfolio could shift over time, including movement across asset stages (development to stabilized) or across the capital stack (equity, preferred equity, debt).
Marketing Discipline
In any private fund offering, statements made to investors (in PPMs, pitchbooks, case studies, diligence responses and other communications) should be accurate, complete in context, and not misleading. For AI infrastructure strategies, investor-facing materials often include underwriting narratives where key drivers are assumptions about future operations. Examples include utilization and ramp timelines, pricing, power availability, equipment delivery schedules, permitting and energization dates, useful lives of technical components and counterparties’ performance.
Some of those assumptions may be informed by contracted commitments (for example, customer service agreements or power arrangements), but even “committed” arrangements can include exceptions, conditions, termination rights, force majeure exceptions and credit support limitations that are material to the risk profile. Managers and investors alike benefit when marketing and other communications distinguish assumptions from commitments where that distinction is material, and when “commitments” are described in a way that does not unfairly downplay material qualifiers.
Third-party inputs (market analyses, engineering and power studies, vendor statements, customer discussions and consultant analyses) can change quickly, and the same is often true of operational facts in development-stage assets. This, in turn, can have follow-on effects in underwritten return targets. Where underwriting inputs are referenced in marketing or diligence materials, clarity improves when those inputs are dated and summarized carefully and are revisited when there are known material developments. For SEC-registered investment advisers, the Advisers Act marketing rule and related recordkeeping requirements can make substantiation and version control particularly important, especially for material statements. If a manager is not SEC-registered, the same practical considerations still apply: keeping a record of the basis for material statements, and clearly disclosing any material underwriting criteria and assumptions, can reduce later disputes about what was said, when and why.
Governance and Conflicts
AI infrastructure strategies can involve multiple vehicles, affiliates and overlapping relationships across development, procurement and operations. Those features can increase conflicts sensitivity in fund governance. Common pressure points include allocation of investment opportunities and expenses among funds and accounts, cross-fund investments at different points in the capital stack (including different seniorities), decisions around follow-on financings and recapitalizations and affiliate-provided services (development management, construction management, operations and maintenance, procurement or related services) as well as related compensation. It is not always feasible to eliminate all conflicts, but the objective should be to at least make the process predictable and defensible. Clear disclosure of expected conflicts, coupled with defined approval mechanics (including any applicable LPAC mechanisms), can shape how issues are escalated and resolved.
Contract Risk Allocation at the Asset Level
Both fund sponsors and investors can benefit from ensuring that there is a mutual understanding of the risks that can arise in AI infrastructure investments. Such risks can include: concentration dependencies (on key customers, suppliers, electricity providers or other counterparties); volatility of short-term demand (relative to project forecasts); uncertainty of future needs (today’s facility designs may not match tomorrow’s needs); changes to customer base (loss or turnover of major tenants can disrupt cash flow); erosion of asset value and uncertainty of assets’ useful lives (assets might lose value if they become under-utilized or technologically outmoded, difficulty in estimating assets’ optimal lifespan); limited access to compute/equipment (supply-chain bottlenecks for GPUs, power gear, etc., can delay or restrict operations); constraints on permitting/utility connections (extensions in project timelines due to delays in permitting processes and power grid/data connection hookups); financing/interest rate risks (rising interest rates or tighter credit can impact project viability and returns); cybersecurity and data breach risks (concentration of valuable data in AI data centers, with liabilities/losses from any resulting breaches/downtime); physical disaster risks (natural disasters/extreme weather and related grid reliability issues, potentially affecting insurability); limited ability to use licensed intellectual property (in certain cases, third-party license terms may restrict how the infrastructure’s technology can be deployed, especially for AI uses); and changes in laws, regulation and policy (including AI governance, environmental and antitrust policy).
It is also important to understand how these risks can be transmitted – and to whom – across a broader asset structure, where a setback in one area (for example, a construction delay or a supplier default) can cascade to other stakeholders. AI infrastructure projects typically allocate risk across counterparties through a set of interlocking agreements, rather than through a single “project” contract. These agreements can include land and permitting arrangements (securing the site and entitlements), interconnection and utility service (for power and network access), construction and equipment contracts (with builders and vendors, often with performance guarantees), purchase orders for compute capacity (ensuring demand via customer commitments), operations and maintenance agreements (governing uptime and management of the facility), power procurement arrangements (to supply the enormous energy needed, sometimes via dedicated utility or renewable deals), customer service or colocation agreements (outlining lease terms, service levels and security with end-users), financing arrangements (debt and equity deals that fund the project with their own covenants) and insurance policies (covering construction, liability, property damage, cyber incidents, etc.). Each layer of agreements may include its own remedies, caps, exclusions, cure periods, step-in rights and termination mechanics. In addition, obligations at one layer may be contingent on performance at another, meaning a failure in one contract can trigger consequences in others, creating a “risk transmission” effect throughout the project’s contractual structure. This complexity underscores the importance of mapping and disclosing material risks to the fund resulting from interdependencies and potential failure points across the structure.
Valuation, Performance Reporting and Updates
Complex assets can make for complex valuation, and complex mixes of complex assets even more so. Portfolios that mix development assets, stabilized assets and structured exposures can introduce consistency issues in valuation and performance reporting if not approached carefully. A thoughtful and consistently applied valuation framework (including guidelines on methodology and key inputs for different assets, any third-party involvement and governance procedures for overrides and methodology changes) can benefit all parties. Consistency in performance calculation methodologies can raise similar concerns, and similar benefits can be derived from having a thoughtful and consistently applied performance calculation framework that reflects key differences across these assets.
Because operational and market conditions can shift quickly, managers may also benefit from clear divisions of responsibility for monitoring intervening events to determine whether any investor facing descriptions, case studies or risk discussions should be refreshed. Depending on materiality, potential triggers could include permitting delays, interconnection changes, supply disruptions, customer changes or shifts in expected capex or in-service dates. Where confidentiality obligations limit detail, managers could consider whether higher-level disclosures would still convey the material nature of an issue without implying facts that cannot be disclosed.
Sanctions, National Security and Outbound Investment Controls
Depending on counterparties, jurisdictions, end uses and technology characteristics, AI infrastructure investments can implicate US sanctions screening, export controls, national security review, and, in some cases, outbound investment control considerations. These regimes are fact-specific and evolve, so current requirements should be verified for the relevant transaction and timeframe.
Practically, fund sponsors can seek to manage these risks by implementing robust processes for investor and investment onboarding and diligence including investor and counterparty screening, by performing end-use and location diligence where relevant and by seeking appropriate contractual protections.
* * * * *
AI infrastructure projects can evolve faster than a private fund’s offering and governing documents. But these challenges are addressable. Clearly delineating investment scope, providing thoughtful disclosures, establishing clear conflicts management processes, mapping project-level risks and interdependencies, being mindful of foreign policy shifts and implementing consistently applied valuation and reporting guidelines can help to manage these risks.