Goldman Sachs just struck a remarkably confident tone on Microsoft (MSFT)stock just ahead of a key earnings report on January 28.
The firm slapped the software giant with a buy rating and an eye-popping $655 price target, which implies nearly 37% upside from current prices.
I can’t say how other analysts see it, but for me, Microsoft has been a unique stock.
Despite its continued fundamental outperformance and strong stock-market showing, it’s never quite been the “it” stock.
However, that unusual reputation hasn’t deterred the stock, which has posted a jaw-dropping 91% gain over the past three years, comfortably beating the broader market.
Last year, however, performance was mostly muted, with the stock up just 7%, as AI fatigue appeared to catch up with it, even considering its first-mover advantage.
Following the dip, though, the stock’s now ripe for the picking, trading at over 28-times non-GAAP forward earnings (12% lower than its five-year average).
It’s interesting to note that Goldman’s price target is the most bullish among the major analysts.
-
Morgan Stanley: $650 price target (overweight).
-
Barclays: $610 price target.
-
J.P. Morgan: $575 price target (buy).
-
Wedbush: $625 price target (outperform).
On top of that, Goldman’s rationale is perhaps what sticks the most, where it argues the market is underappreciating the long-term value its AI endeavors can unlock for the business.
Goldman Sachs says the broader market isn’t pricing in nearly as much value as Microsoft’s Copilot tools and AI agent-based workflows add to the company’s long-term case.
At the core of its approach is the belief that Microsoft’s approach is moving quickly beyond just experimentation and into practical, repeatable use cases.
Since its release back in 2023, Microsoft’s “family of Copilot apps” has been a juggernaut for the business. In fact, during the FQ4 2025 earnings call, CEO Satya Nadella said that it boasted 100 million monthly active users across commercial and consumer segments, according to the The Times of India.
More Tech Stocks:
Moreover, in the latest earnings call, Nadella remarked that “more than 90% of the Fortune 500” are using Copilot.
A big part of its success is its unique positioning, sitting directly inside products that customers are already using.
Which is why veteran analyst firm Wedbush argued that the power-packed combo of Copilot and Azure could potentially add $25 billion in sales by fiscal 2026.
That form’s a big part of why, in Goldman’s upside scenario, Microsoft may deliver more than $35 in earnings per share by fiscal 2030 (north of 20% EPS growth), trumping the mid-teens expansion expected from most mega-cap players.
On top of that, Goldman points to Microsoft’s growing focus on AI agents, effectively revolutionizing how work gets done inside organizations.
That, in turn, will drive stickier usage, along with more predictable long-term monetization with Microsoft’s vertical integration (AI compute, platforms, and applications), giving it a serious edge.
For these reasons, its power-packed AI strategy will likely translate into robust and measurable demand across its business.
-
Infrastructure (Azure and AI data centers): More AI agents in play will require stronger training and push inference demand higher. For perspective, Microsoft’s Azure business has already blown past $75 billion in sales in fiscal 2025, and that level of growth is unlikely to stop anytime soon.
-
Platform (Foundry and agent services): In July 2025, Satya Nadella said Foundry APIs processed a whopping 500 trillion tokens, up sevenfold.
-
Applications (Copilot, GitHub, Dynamics): Microsoft can efficiently embed agents into everyday workflows. For a little color, customers reportedly created 3 million agents through SharePoint and Copilot Studio in a year, while GitHub Copilot skyrocketed to 26 million users by late 2025.
It’s also important to note that the market for agent-based AI is massive.
For instance, Gartnersaid that 40% of enterprise apps will include task-specific AI agents by the close of 2026.
Wall Street might be excited about agentic AI, but the performance gap is impossible to ignore.
For instance, more than 40% of agentic AI projects might be canceled by the end of 2027, with costs outpacing results,Gartnerwarned recently. Another worrying trend it also flagged was widespread “agent washing,” with just 130 vendors being viewed as truly agentic rather than repackaged chatbots.
Real-world performance data also back up that caution.
Related: Cathie Wood drops $50 million on AI stock, dumps favorite
In TheAgentCompany benchmark, which effectively tested 175 realistic workplace tasks (stuff like web apps, documents, and code), even the premier AI model completed just 30% of tasks.
Moreover, the most common setups actually performed far worse, often getting stuck in loops or failing in complicated interfaces.
The Reddit community broadly agrees, and a pertinent post from user hkreporter2 in the AI_Agents forum (116,000 agents) said it best.
Related: Nvidia CEO sends biblical-scale reality check on AI
This story was originally published by TheStreet on Jan 17, 2026, where it first appeared in the Investing section. Add TheStreet as a Preferred Source by clicking here.
