Part 2: AI data center construction, CIP, all the cash flows and concentration

After Part 1 last week we're turning to CIP reporting, cash flow disclosures and an analysis of potential concentration exposure if AI demand falters.

This is collaborative effort, co-authored with Deep Quarry the newsletter from Olga Usvyatsky. If you value our work, please subscribe!

Data centers have always been a major component of tech CAPEX, but the scale of AI-specific facilities is transforming them into one of the most widely discussed capital-allocation stories of the past several decades.

Projects tied to OpenAI, Anthropic, Meta, Microsoft, Oracle, and Amazon do not necessarily raise new accounting questions. Rather, the projects increase the materiality of familiar issues — such as consolidation and lease accounting, disclosure of construction-in-progress balances and disaggregation by PPE category, and presentation of CAPEX-related accounts payable. The potential materiality of AI-related construction projects to the lead organizations’ balance sheets magnifies insufficient and inconsistent disclosure and, in some cases, suggests more specific accounting guidance is needed.

In Part 1, we examined how the AI-related projects are being structured and financed, and why the accounting for and disclosure of these projects should be evaluated in conjunction with the chosen structure of the project.

Key themes in Part 1:

• Deal structure matters. Some companies appear willing to keep more of the financial impact of the construction on their own balance sheets, while others rely more heavily on joint ventures, special-purpose vehicles, and sale-leaseback arrangements that obscure the full impact of the projects.

• Accounts payable are rising across the AI-related data center construction industry but not all increases are created equal. Across several major tech companies, total accounts payable as well as payables tied to property, plant, and equipment have risen sharply, suggesting that some companies may be relying on supplier financing. In our view, it is important to differentiate between ordinary cash management activity and financing-like arrangements embedded in the financing structure of some of the construction projects.

• Disclosure remains inconsistent. Companies do not all present CAPEX-related activity the same way, making comparisons more difficult for investors.

• Concentration is a material risk. If AI demand weakens or construction timelines slip, the economic pressure may not fall evenly across sponsors, developers, lenders, vendors, tenants, and investors.

In our view, the broad increase in days payables outstanding, or DPO, across the group of firms focused on AI-related construction projects suggests that large-scale AI and data center construction — including EPC contracts, long-lead GPU procurement, and phased buildouts — may be significantly influencing payable timing.

Differences in smoothness versus choppiness likely reflect variations in construction contract structures (fixed schedule vs. milestone-based), the cadence of PPE purchases, and whether companies are expanding capacity through steady multi-year programs or more episodic project launches.

As discussed above, however, DPO trends alone do not fully capture the mechanics of the cash flow presentation and construction-in-progress (CIP) balances, including the timing of “placed in service” decisions which drive depreciation expense.

Part 2 will turn to construction-in-progress, cash flow presentation, and who may be left exposed if the buildout proves too optimistic.

Source: https://thealtitudeagency.com/avoiding-the-development-money-pit/

We’ll examine CIP reporting and cash flow disclosures in greater depth, before concluding with an analysis of the potential concentration exposure that might occur, and where, if AI demand slows.