Microsoft Corp.: Winning Formula – Azure Estimates Up, Capex Reaffirmed

研报英文原文证据摘录

Microsoft Corp.: Winning Formula – Azure Estimates Up, Capex Reaffirmed

2HCY26 Azure growth to be higher than 1H would still hold.

Token Optimization: Ahead of the print, we flagged the risk that enterprise

token spend optimization in 2H might weigh on Azure growth but this topic didn’t

even come up, not on the public call and not in our call-back. That is a relief, and

while it’s very clear that enterprises are token-optimizing, perhaps any efficiency

improvements were filled/offset by strong new demand/prompts.

1P Allocation: Moreover, the outlook signals that the 1P internal allocation of

compute did not ramp materially and choke off 3P customer access to Azure AI

capacity.

Rev Rec’ing AI Lab Sales: This is an extraordinary ramp in sequential dollar adds.

While core CPU-based demand for Azure services may also have been strong

(more on this below), it seems very likely that Azure (as well as Google Cloud and

AWS) is benefiting materially from its role as a front-end procurement engine for

OpenAI and Anthropic model access and 1P products, as Azure and its hyperscaler

peers (as well as Databricks and others) recognize this as revs.

Capacity Adds: Microsoft stood up another 1 GW of capacity in 4Q/Jun, the third

straight quarter with 1 GW of adds. That’s 3 GW of new capacity (equivalent to

$30b+ of incremental revs) in the last 9 months alone, a material capacity unlock.

In our view, the willingness of the CFO to guide to an Azure growth acceleration in

2HCY26 signals that Microsoft has very good visibility into the addition of “warm

shells” or raw capacity going live in the coming quarter, although to be clear

Microsoft did not offer any capacity guidance. Microsoft argues that it has made

strides in driving efficiency gains in terms of data center go-lives, improving