Reporting from Bloomberg, Barron's, The Hindu BusinessLine and Tickeron shows a sharp pullback in share repurchases among large US tech companies as they redirect capital toward AI infrastructure. Bloomberg data cited by multiple outlets found that of the four largest AI spenders, Alphabet, Microsoft, Meta and Amazon, only Microsoft repurchased stock in Q1, at $3.4 billion, its smallest quarterly total in nearly a decade. The Hindu BusinessLine reports that Alphabet is planning an equity sale of about $85 billion to fund data-center capex, and that Meta is weighing a large offering. Tickeron reports combined buybacks by several majors fell to $12.6 billion in Q4 2025, a 74% decline from a $48 billion peak in 2021. Barron's frames the scale of AI capex as historically large compared with past national projects.
What happened
Reporting by Bloomberg and republications in Yahoo Finance and The Hindu BusinessLine show that share repurchases among the largest AI investors have declined materially as capital expenditures for AI infrastructure rise. Per Bloomberg data cited by those outlets, of the four biggest AI spenders, Alphabet, Microsoft, Meta and Amazon, only Microsoft executed buybacks in the first quarter, totaling $3.4 billion, the lowest among that group in nearly a decade. The Hindu BusinessLine reports that Alphabet is planning an equity sale of about $85 billion to help fund data-center spending, and that Meta is reportedly considering a stock offering that could raise tens of billions of dollars. Tickeron reports combined buybacks by Amazon, Alphabet, Microsoft, Meta, and Oracle fell to $12.6 billion in Q4 2025, a 74% decline from a roughly $48 billion peak in 2021. Barron's places the aggregate AI capex scale in historical terms, comparing it to major national infrastructure projects.
Technical details
Editorial analysis - technical context: Training and deploying large generative models requires sustained, high-dollar investments in data centers, accelerators, and specialised networking and power infrastructure. Industry reporting highlights three cost drivers common across vendors: higher GPU and accelerator procurement, denser data-center builds, and greater operating power and cooling requirements. These capital demands convert software-led businesses into more capital-intensive platforms, increasing near-term cash needs for physical infrastructure and custom silicon.
Context and significance
The shift away from buybacks toward capex alters common valuation and cash-return dynamics that supported Big Tech share performance after 2019. For investors and practitioners, larger and more persistent infrastructure spending can expand cloud and private-infrastructure capacity, lower per-inference costs over time, and change vendor economics for AI services. Reporting also notes some firms are supplementing balance sheets with equity raises rather than debt, which has different implications for ownership dilution and corporate flexibility.
What to watch
Editorial analysis: Observers should track quarterly capex and buyback line items in corporate filings, disclosures about data-center footprints and custom-accelerator projects, and any announced equity offerings tied to capex. Watch for changes in gross margin trends on cloud/AI services, utilisation rates of third-party versus in-house accelerators, and commentary from credit analysts such as the Bloomberg Intelligence team, which has flagged capex as a primary driver behind the decline in repurchases. Market responses to dilution events or sustained lower buyback activity will also matter for capital markets and compensation-linked metrics.
Scoring Rationale #
This story signals a notable capital-allocation shift among major cloud and AI players that affects market dynamics, infrastructure capacity, and cost structures for model deployment. It is important for practitioners and investors but does not introduce a new technology or regulatory change.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.