Big Tech’s AI buildout is larger than headline capex suggests, SEC filings show
Daniel Simon ·
The scale of Big Tech’s AI buildout is not fully visible in standard capital expenditure lines. Recent SEC filings by Microsoft, Amazon and Alphabet show rising finance lease liabilities and multiyear purchase commitments tied to chips, data centers and network equipment that sit outside headline capex.
Microsoft’s 2024 Form 10-K discloses material noncancelable purchase commitments for servers, networking gear and related cloud infrastructure alongside finance lease obligations for data center assets. In the commitments and contingencies note, the company details multiyear obligations for third party hardware and construction services in addition to power and hosting arrangements that do not run through capex on the cash flow statement.
Amazon continues to rely on a mix of cash capex and equipment under finance leases for AWS infrastructure. Its 2023 Form 10-K and subsequent 2024 quarterly reports provide separate line items for property and equipment acquired under finance leases and show lease liabilities for data center equipment and facilities, a structure that can depress reported capital spending while keeping long dated obligations off the capex line.
Alphabet’s 2024 Form 10-K reports significant purchase commitments for compute capacity, data center construction and networking services over multiple years, as well as lease liabilities for real estate and technical infrastructure. The company also added disclosure about supplier financing arrangements in response to SEC guidance, clarifying how certain payables tied to large scale equipment procurement are settled.
Chip supply adds a second layer of off line obligations. Nvidia’s 2024 Form 10-K describes long term inventory purchase and supply agreements with foundry and memory partners that expanded materially year over year as demand for AI accelerators surged. Hyperscalers that pre book capacity through these agreements or through direct purchase commitments record the obligations in footnotes rather than in period capex totals.
The SEC adopted new disclosure requirements in 2022 for supplier finance programs that took effect in 2023. The rules require registrants to present the size and balance sheet presentation of payables financed through such programs. The filings reviewed show that Big Tech has begun to add these details, but the presentation still fragments the true cost of the AI buildout across capex, leases, purchase commitments and supplier financing.
Investors who track free cash flow and capital intensity risk understating the scale and duration of AI infrastructure spending. Utilities and local governments face a parallel blind spot because power and water demand from data centers can ramp ahead of what cash capex implies, complicating grid planning and environmental review timelines. Rivals and smaller AI firms face higher entry barriers when incumbents lock in multiyear chip supply and colocation capacity.
The next test arrives with September quarter reports that begin in late October. Analysts will parse updates to lease liabilities and purchase obligations alongside capex guidance to gauge whether AI infrastructure costs are accelerating into 2027. The SEC could also press for more standardized breakouts of AI related commitments in upcoming comment letters if current disclosures continue to obscure comparability across firms.