Tech giants are racing to build the physical backbone of artificial intelligence, pouring hundreds of billions of dollars into data centers, specialized chips, and power-hungry cooling systems. Yet as the concrete is poured and servers are racked, the true cost of this build-out is often buried inside opaque accounting categories that leave investors guessing. I see a widening gap between the scale of AI construction and the clarity of the financial reporting that is supposed to explain it.
The AI build-out is massive, but the numbers are murky
Across the largest platforms, AI has shifted from a software story to a concrete and steel story, with capital spending on data centers and related infrastructure climbing at a pace that would have seemed implausible only a few years ago. Companies are not just buying more servers, they are designing custom accelerators, locking in long term power contracts, and building new campuses that can support dense clusters of AI hardware. Yet in many quarterly reports, those costs are bundled into broad capital expenditure lines that obscure how much is going into AI versus more traditional computing or office projects, which is why even seasoned analysts now describe AI construction as an accounting black box.
That opacity is not just a matter of style, it has real consequences for how markets price risk and reward. When a company tells investors that capital expenditures are rising sharply but does not break out how much of that surge is tied to AI, it becomes difficult to judge whether management is making disciplined bets or chasing hype. Reporting by Mark Maurer, a Reporter at The Wall Street Journal, has highlighted how this massive AI build-out comes with a transparency problem, with his Mark Maurer’s Post describing how companies can capitalize AI related construction in ways that make it hard to see the full picture in less than three years.
Data center spending is a “financial black box” for investors
From an investor’s perspective, the most immediate problem is that data center construction costs are often aggregated into a single capital expenditure figure that covers everything from AI clusters to routine facility upgrades. I find that this aggregation turns what should be a clear investment thesis into a guessing game, especially when management teams talk up AI ambitions on earnings calls but do not provide matching detail in their financial statements. Without a breakdown of how much is going into land, buildings, power infrastructure, networking, and AI specific hardware, shareholders are left to reverse engineer the numbers using rough industry benchmarks and vendor disclosures.
Recent analysis has underscored that tech giants are investing hundreds of billions of dollars in AI infrastructure while their financial disclosures lack specific detail on individual cost components, a gap that has led some observers to describe data center construction costs as a financial black box. When capital spending is rising at that scale, the absence of granular disclosure is not a minor footnote, it is a central risk factor. It affects how investors model future depreciation, how they think about the sustainability of free cash flow, and how they compare one company’s AI strategy with another’s.
Accounting rules lag the AI infrastructure reality
The accounting framework that governs how companies treat AI construction costs was not designed for a world where a single hyperscale campus can cost tens of billions of dollars and where the line between research, software, and physical infrastructure is increasingly blurred. In practice, management teams must decide what portion of AI related spending should be capitalized as construction in progress, what should be expensed as research and development, and how to allocate shared costs like power and networking between AI and non AI workloads. I see those judgments as fertile ground for inconsistency, even when everyone is acting in good faith.
Regulators are aware of the tension, but any shift in formal standards will take time. The Financial Accounting Standards Board, which sets U.S. accounting rules, has been flagged as a key player in potential reforms, yet reporting makes clear that any such changes in U.S. accounting rules would be a while away, leaving companies to navigate the current framework with limited guidance on AI specific issues. That delay matters because it means the market will be living with today’s patchwork of practices for years, even as AI capital spending accelerates. As one detailed overview of these issues noted, Any such changes in U.S. standards will arrive only after extensive deliberation, long after the current wave of AI data centers is already on the books.
CAPEX, Construction in Progress, and the scrutiny to come
Within the existing rules, the most contested territory sits inside capital expenditure, construction in progress, and the treatment of interest and overhead during long build cycles. When a company embarks on a multi year AI campus, it can accumulate large balances in construction in progress that do not immediately hit the income statement, which can make profitability look stronger in the short term. I view that as a legitimate feature of accrual accounting, but in the AI context it also creates room for aggressive assumptions about useful lives, residual values, and the point at which assets are considered ready for use.
Accounting experts are already signaling that this area will attract more attention from regulators, auditors, and investors. Marc Siegel, who has closely followed these developments, has described how scrutiny is likely to intensify around CAPEX, Construction in Progress and the way companies capitalize interest on AI projects, noting that this topic drew strong interest at a recent conference. His Marc Siegel’s Post framed this as a Good signal that market participants are starting to ask tougher questions about how AI construction is reflected in the numbers, rather than simply applauding headline spending.
Why transparency on AI build costs matters now
The stakes in this debate go beyond technical accounting. When AI infrastructure spending is opaque, it distorts how capital is allocated across the entire market, from pension funds that hold the largest tech names to utilities that must plan for the power demands of new data centers. I see a direct link between disclosure quality and the ability of outside stakeholders to assess whether AI investments are likely to generate returns that justify their environmental and financial footprint. If investors cannot distinguish between disciplined long term infrastructure bets and short term arms races, they are more likely to misprice risk, which can fuel bubbles on the way up and deepen corrections on the way down.
Some finance professionals are already trying to pull this conversation into the open. Walden Siew has highlighted how the massive AI build-out comes with a transparency problem, amplifying reporting by Mark Maurer Tech on the way companies describe their AI infrastructure programs to the market, and his Walden Siew Post has helped push the issue into broader financial circles. Others, such as Kris Bennatti, have drawn attention to how AI Construction Costs Can Be an Accounting Black Box for auditors and analysts who must interpret these figures, with her commentary referencing how a professor at Little Rock framed the challenge for students learning modern financial reporting. In that context, her Construction Costs Can Be discussion underscores that the next generation of accountants will inherit these questions, not escape them.
More From TheDailyOverview
Grant Mercer covers market dynamics, business trends, and the economic forces driving growth across industries. His analysis connects macro movements with real-world implications for investors, entrepreneurs, and professionals. Through his work at The Daily Overview, Grant helps readers understand how markets function and where opportunities may emerge.
