Goldman Sachs reiterated its Buy rating on Microsoft (NASDAQ:MSFT) with a $515 price target in a note this week, citing strong returns on the company's investments in Generative AI (Gen-AI).
While some investors are concerned about the return on investment (ROI) for Gen-AI, Goldman Sachs sees parallels to the company's successful Azure cloud computing build-out.
"We draw several parallels to the Azure infrastructure build-out in which Azure CapEx intensity (CapEx / Azure Revenue) stabilized around the ~50% level over a decade after the launch of Azure," states Goldman Sachs.
Analysts show that Microsoft's Gen-AI CapEx efficiency is already comparable to years four or five of the cloud computing cycle, even though Gen-AI is only about a year into its rollout.
Microsoft has managed to balance rapid CapEx growth with tangible AI revenue generation. While the CapEx increase is similar to the Azure build-out, analysts don't expect a similar decline in gross margins (GM) as seen during the Azure build-out (which saw a 19pp drop).
Currently, AI revenue generates positive gross profit, but depreciation outpaces revenue, leading to gross margin dilution, notes the bank.
Analysts estimate that without these investments, Microsoft's gross margins could be 150-300 basis points higher. However, they also highlight that AI-specific gross profit dollars in FY24 are already comparable to what Azure achieved eight years after its launch (FY18).
Drawing another parallel to Azure, analysts point out that Microsoft's operating margin (OM) has increased by 10pp since Azure began contributing meaningfully to revenue (FY12/13), despite a 7pp contraction in gross margins.
They see a similar trajectory for Gen-AI, with AI revenue ramping up faster than Azure did. After just one year, AI-driven revenue is already equivalent to what Azure achieved in year seven.
Analysts believe that eventually, margins will benefit from increased scale, leading to slower CapEx growth and improved margins. They also expect Microsoft to reach a critical mass for Gen-AI much faster than with Azure, which took seven years.