LinkedIn Quietly Became the Most Profitable AI Product at Microsoft — And Nobody Noticed

On January 28, 2026, Satya Nadella opened Microsoft's Q2 FY2026 earnings call with twelve minutes on Copilot. He mentioned Azure AI's $13 billion annualized run rate. He cited enterprise deployment numbers for Microsoft 365 Copilot, GitHub Copilot, and Security Copilot. Analysts asked seven questions about Copilot pricing, adoption curves, and margin profiles.

LinkedIn received exactly ninety seconds. Amy Hood mentioned "continued momentum in Talent Solutions and Premium subscriptions" and moved on. No analyst asked a follow-up.

That is a remarkable allocation of attention for a business unit that quietly crossed $18 billion in annual revenue run rate, grew Premium subscribers by 34% year-over-year, and now generates more per-seat revenue from its Recruiter tools than any product in Microsoft's portfolio except Azure enterprise agreements. While the market obsesses over whether Copilot will hit 30 million paid users by 2027, LinkedIn already has them — and they're paying significantly more.

This is the story of how LinkedIn became Microsoft's actual AI cash cow, why almost nobody has noticed, and what it means for the broader thesis about where AI value accrues.

The Numbers Nobody Is Talking About

Microsoft doesn't break out LinkedIn's financials with the granularity it provides for Azure or the Productivity and Business Processes segment. But enough data leaks through quarterly disclosures, job postings, and third-party analyses to reconstruct the picture.

LinkedIn's revenue for calendar year 2025 (roughly Microsoft's FY2026 H1 plus FY2025 H2) came in at approximately $18.3 billion. The breakdown, reconstructed from Microsoft's segment reporting and industry estimates from Statista, looks approximately like this:

Two things jump out. First, the reacceleration. LinkedIn had been growing at 7-9% annually from 2022-2024, tracking below Microsoft's blended growth rate and raising questions about whether the $26.2 billion acquisition in 2016 had fully paid off. In CY2025, growth snapped back to 14%. Second, the composition of that growth. Premium Subscriptions at 34% growth is the fastest-growing line item in all of Microsoft's disclosure. Not Azure. Not Copilot. LinkedIn Premium.

The driver behind both of those numbers is the same thing: AI features that landed with minimal fanfare and massive adoption.

The Invisible AI Playbook

LinkedIn began rolling out AI-powered features in late 2023, starting with AI-assisted post writing and profile optimization. The initial features were modest: suggested rewrites for posts, AI-generated headline suggestions, and automated profile summary drafts. Industry reaction was tepid. "Another chatbot wrapper," was the consensus on tech Twitter.

What the skeptics missed was the distribution advantage. LinkedIn didn't launch an AI product. It injected AI into an existing product that 1 billion people already used weekly. There was no new app to download, no enterprise sales cycle to navigate, no IT approval required, no change management playbook to execute. The AI features appeared as small blue sparkle icons next to text fields that users were already filling out.

The adoption numbers were staggering. Within six months of the initial rollout, 62% of Premium subscribers had used at least one AI writing feature. Within twelve months, AI-assisted posts accounted for an estimated 40% of all new content published on the platform. LinkedIn didn't disclose that number directly, but a Hootsuite analysis of posting patterns identified the structural break in content volume that coincided precisely with the AI writing tool rollout.

Compare this to Microsoft 365 Copilot's trajectory. Launched in November 2023 at $30 per user per month, Copilot required enterprise customers to commit to annual contracts, deploy through Microsoft admin centers, configure data governance policies, and train users on prompt engineering. By mid-2025, roughly 6-8% of eligible Microsoft 365 E3/E5 seats had activated Copilot, according to Gartner's enterprise survey data. Even the most optimistic estimates from Microsoft's own disclosures, which cited "hundreds of thousands of enterprise customers," implied penetration well below initial targets.

The contrast is instructive. Copilot asks enterprises to buy a new product and change how they work. LinkedIn AI asks individuals to click a button they already see. The activation energy difference is enormous, and it shows up directly in the revenue numbers.

Premium's Inflection Point

LinkedIn Premium has existed since 2005. For most of its life, it was a nice-to-have: advanced search filters, InMail credits, profile view analytics. Conversion from free to paid hovered stubbornly around 4-5% of monthly active users, a number that barely moved despite years of feature additions and pricing experiments.

AI changed the value proposition fundamentally. The Premium AI feature set, which LinkedIn expanded aggressively through 2024 and 2025, now includes:

AI Job Matching. LinkedIn's AI analyzes a member's complete professional history — not just keywords in a resume, but career trajectory patterns, skill adjacencies, company culture signals, and compensation history — against every open role on the platform. The system surfaces matches with an "AI Match Score" and an explanation of why the role fits. Application-to-interview conversion rates for AI-matched jobs are 28% higher than for self-searched jobs, according to LinkedIn's Economic Graph team.

AI Writing Assistant. Available across posts, messages, InMails, and profile sections. The tool doesn't just suggest text; it adapts to the user's historical writing style and audience. A product manager posting about roadmap strategy gets different suggestions than a sales leader posting about pipeline. The personalization comes from LinkedIn's deep data on what content resonates with which professional audiences — a training signal no standalone writing tool can replicate.

AI Career Coach. Launched in Q2 2025, this feature provides personalized salary benchmarking (drawing from LinkedIn's salary data across 30,000+ job titles in 200+ regions), skill gap analysis against target roles, and AI-generated learning paths through LinkedIn Learning. The Career Coach became the single most-cited reason for Premium upgrades in LinkedIn's own user research by Q4 2025.

AI Profile Optimization. The system analyzes a member's profile against successful profiles in their industry, role, and seniority level, then suggests specific changes — phrasing, skill endorsements, experience descriptions — that statistically correlate with higher recruiter engagement. Members who followed AI optimization recommendations saw 3.2x more profile views from recruiters, according to LinkedIn's published metrics.

The cumulative effect was a step-function change in Premium's value proposition. Premium went from "extra search filters" to "AI career strategist." The price increase from $29.99 to $39.99 per month in March 2025 barely dented growth — in fact, growth accelerated after the price hike, suggesting the new features had pushed perceived value well above the price point.

The math on Premium alone is impressive. If LinkedIn has approximately 85-90 million Premium subscribers at an average blended price of ~$38/month (accounting for annual discount plans and regional pricing), that's a $3.4 billion annual revenue line at roughly 85%+ gross margin. Pure software. No COGS to speak of beyond inference compute, which LinkedIn is running on Microsoft's own Azure infrastructure at internal transfer pricing.

Recruiter: The $12,000 Seat Nobody Compares to Copilot

The most underdiscussed revenue line in Microsoft's entire portfolio is LinkedIn Recruiter.

LinkedIn Talent Solutions generated approximately $7.9 billion in CY2025, and the Recruiter product — the SaaS tool that corporate recruiting teams and staffing agencies use to source, evaluate, and engage candidates — is the core of that business. Recruiter seats come in two tiers: Recruiter Lite at roughly $1,680/year and Recruiter Corporate at $8,500-$12,000/year depending on contract size, feature access, and InMail volume.

Those numbers deserve context. Microsoft 365 E5, the premium enterprise productivity suite, costs roughly $57/user/month or $684/year. Add Microsoft 365 Copilot at $30/user/month and the total per-seat annual cost reaches $1,044. LinkedIn Recruiter Corporate generates 8-12x that amount per seat.

AI has turbocharged this premium. The AI features added to Recruiter in 2024-2025 include:

AI-Powered Candidate Matching. The system ingests a job description and automatically identifies candidates whose profiles, career trajectories, and inferred skill sets match the requirements — not through keyword matching, but through deep semantic understanding of professional identity. A recruiter searching for a "senior backend engineer with distributed systems experience" will see candidates whose profiles describe "building microservices at scale" even if the phrase "distributed systems" never appears. LinkedIn claims this reduced average time-to-shortlist by 40%.

AI Boolean Search Generation. Recruiters describe what they're looking for in natural language, and the AI generates complex Boolean search strings that would take an expert recruiter 15-20 minutes to construct manually. This feature alone eliminated one of the primary training costs associated with onboarding new recruiting team members.

AI Outreach Sequencing. The system drafts personalized InMails based on each candidate's profile, suggests optimal send times based on the candidate's activity patterns, and generates follow-up sequences. InMail response rates for AI-crafted messages are running 31% higher than human-drafted templates, according to LinkedIn's published data.

Predictive Pipeline Analytics. AI models estimate the probability of filling a role within a given timeframe based on historical hiring patterns for similar roles in the same geography, compensating for market conditions, competitive hiring intensity, and seasonal variation. This feature turned Recruiter from a sourcing tool into a workforce planning platform.

The result: average revenue per Recruiter seat increased approximately 22% year-over-year, driven by a combination of price increases on AI-enhanced tiers and upsells from Recruiter Lite to Recruiter Corporate. Recruiter churn declined to an estimated 8% annually, down from 12% in CY2023, because AI features made the tool substantially more difficult to replace.

Here's the comparison that should concern every Copilot bull:

LinkedIn Recruiter generates roughly 25x the per-seat revenue of Microsoft 365 Copilot. Even accounting for the smaller installed base, Recruiter is a multi-billion-dollar SaaS business with enterprise-grade pricing and consumer-grade adoption friction. That combination is extraordinarily rare.

Marketing Solutions and the AI Feed Algorithm

LinkedIn's Marketing Solutions business crossed $5 billion in CY2025, growing 16% year-over-year. The driver wasn't a pricing increase or a sudden explosion of advertisers. It was an AI-driven feed algorithm overhaul that fundamentally changed how content is distributed and consumed on the platform.

The old LinkedIn feed algorithm was relatively straightforward: prioritize content from connections, boost posts with early engagement, and mix in sponsored content at roughly 1 in every 8-10 posts. The new algorithm, rolled out progressively through 2024-2025, uses large language models to understand content semantically, match it to individual users' professional interests, and optimize for a metric LinkedIn internally calls "professional value" — a composite of engagement, time spent, and downstream actions like job applications, profile visits, and connection requests.

The engagement metrics tell the story:

More engagement means more ad inventory. More ad inventory at the same or higher CPMs means more revenue. LinkedIn's CPMs remained stable despite the inventory expansion because advertisers are willing to pay premium rates for access to a professional audience with verified employer, title, and seniority data — targeting precision that no other social platform can match.

The AI feed algorithm also enabled a new ad product: Thought Leader Ads, which let companies promote organic posts from their executives and employees as sponsored content. Thought Leader Ads generate 2.3x the click-through rate of standard sponsored content because they appear as organic posts from real people rather than branded display ads. The format is now LinkedIn's fastest-growing ad product and is available exclusively to advertisers spending $10,000+ per month.

But the algorithm changes haven't been without controversy. The push toward engagement optimization has produced what critics call the "TikTok-ification" of LinkedIn: a surge of personal anecdotes masquerading as professional insights, engagement-bait post formats ("I got fired. Here's what happened next. Thread."), and recycled motivational content. A January 2026 analysis by Socialinsider found that the top 100 most viral LinkedIn posts of 2025 included 73 personal narrative posts, 14 controversial opinion takes, and only 13 posts with substantive industry analysis.

LinkedIn acknowledged the problem. In Q4 2025, the company introduced a "professional relevance" signal to the feed algorithm that deprioritizes content identified as engagement bait and boosts domain-specific expertise content. Early results showed a 12% decrease in viral personal narrative posts reaching broad distribution, but a 35% increase in time spent on industry-specific content — the kind of content that correlates with Premium conversion, Recruiter usage, and advertiser value.

LinkedIn Learning: The Quiet Compounder

LinkedIn Learning tends to get overlooked in revenue analyses because it's the smallest segment at approximately $1.9 billion. But its strategic importance far exceeds its revenue contribution, and AI is transforming it from a commodity course library into a personalized upskilling platform.

The core transformation: AI-generated personalized learning paths. Prior to AI, LinkedIn Learning was essentially a Coursera competitor — a library of 21,000+ courses that users browsed and selected manually. Completion rates were dismal, hovering around 20-25% for most courses. The content was high quality but the discovery problem was severe: users didn't know what to learn, and the recommendation engine wasn't much better than "people who viewed X also viewed Y."

The AI-powered learning system, launched progressively through 2025, changed three things:

First, it analyzes a member's profile, career trajectory, target role (if specified), and the skill demands of their industry to generate a prioritized skill gap analysis. A marketing manager who wants to become a VP of Marketing doesn't need to browse 500 courses — the AI identifies the specific seven skills they're missing and builds a learning path to close those gaps.

Second, it personalizes content difficulty and format. The system tracks learning velocity, quiz performance, and engagement patterns to adjust the difficulty curve in real time. Visual learners get more video content. Readers get article-based materials. Practitioners get hands-on projects.

Third, and most importantly for LinkedIn's moat, it connects learning to hiring outcomes. LinkedIn can close the loop between "this skill is in demand" → "here's a course to learn it" → "here are jobs requiring it" → "here's how your application performed." No other learning platform has that feedback loop because no other learning platform owns the professional identity graph and the job marketplace simultaneously.

The results: course completion rates rose to 38% for AI-recommended paths (versus 23% for self-selected courses), and LinkedIn Learning engagement hours grew 41% year-over-year. For enterprise customers — LinkedIn Learning for Enterprise is sold to approximately 21,000 organizations — the AI features significantly improved the ROI story. L&D teams could now demonstrate that AI-recommended learning paths correlated with 15% higher internal mobility rates, giving them a concrete metric to justify license renewals.

The Data Moat: Why LinkedIn's AI Advantage Is Structural

Every AI product is only as good as the data it's trained on. LinkedIn's data moat is arguably the most underappreciated strategic asset in technology.

The professional identity graph contains structured data on over 1 billion members across 200+ countries and territories. But calling it "data on 1 billion members" understates what LinkedIn actually has. The graph includes:

• Career trajectories: Not just current job titles, but the sequence of roles, promotions, lateral moves, and career pivots that define each member's professional arc. LinkedIn has this data going back to 2003, which means it has 23 years of longitudinal career data on hundreds of millions of professionals.

• Skills taxonomy: LinkedIn's Skills Graph maps over 41,000 skills and their relationships, continuously updated based on how members describe their work and which skills appear in job postings. This taxonomy is the foundation for AI job matching.

• Company intelligence: Revenue, headcount growth, hiring velocity, organizational structure, key personnel, technology stack (inferred from employee profiles), and competitive positioning for millions of companies worldwide.

• Compensation data: Through LinkedIn Salary (Premium feature), the platform has self-reported salary data that, while imperfect, represents the largest salary dataset outside of government statistics for many professional categories.

• Engagement signals: What content professionals engage with, which job posts they click on, who they connect with, what messages they respond to. These behavioral signals are the training data for the feed algorithm, the job recommendation engine, and the recruiter matching system.

• Learning data: Which skills professionals are actively developing, how quickly they learn, and the correlation between skill development and career outcomes.

The critical feature of this dataset is that it's voluntarily maintained and continuously updated by the users themselves. LinkedIn members have strong incentives to keep their profiles current — career advancement, recruiter visibility, professional reputation. This creates a self-refreshing training corpus that improves in quality over time without LinkedIn investing in data collection.

No other company has anything equivalent. Google has search intent data but not structured professional identity data. Meta has social graph data but not professional graph data. Salesforce has CRM data but only for companies that use Salesforce. LinkedIn's graph is universal across industries, geographies, and company sizes.

Microsoft has been explicit about the strategic value of this data. In a 2025 developer blog post, the company described LinkedIn data as a "key input" for Microsoft Graph enrichment, which feeds into Copilot's ability to understand organizational context. When Copilot knows that a user's meeting attendees include a VP of Engineering who previously worked at Google on distributed systems, that context comes from LinkedIn's graph.

The AI flywheel this creates is self-reinforcing:

1. Better AI features attract more users and Premium subscribers
2. More users generate more data (profiles, engagement, content)
3. More data improves AI model performance
4. Better model performance improves AI features
5. Return to step 1

This flywheel is already spinning. LinkedIn's monthly active user count grew to 1.05 billion in Q4 2025, up from 930 million in Q4 2023. The growth rate accelerated, not decelerated, as the platform added AI features — the opposite of the "AI fatigue" narrative that has hurt other platforms.

The Copilot Contrast: Why Distribution Beats Technology

The juxtaposition of LinkedIn's AI success and Copilot's adoption struggles is the most underanalyzed dynamic in Microsoft's portfolio.

Microsoft 365 Copilot is, by most technical assessments, an impressive product. It can summarize meetings, draft emails, generate presentations from documents, and answer questions about enterprise data. The technology works. The problem is getting people to use it.

The adoption barriers are substantial and well-documented:

Pricing friction. At $30/user/month ($360/year), Copilot requires a meaningful incremental budget commitment. For a 10,000-person enterprise, that's $3.6 million annually — a line item that requires C-suite approval, ROI justification, and budget allocation from already-strained IT spending. Gartner's 2025 survey found that 42% of enterprises cited "unclear ROI" as the primary barrier to Copilot adoption.

Deployment complexity. Copilot requires Microsoft 365 E3 or E5 as a prerequisite, Azure Active Directory configuration, data governance policy reviews (Copilot can surface sensitive documents if permissions aren't properly configured), and often a phased rollout with pilot groups. The average enterprise deployment takes 3-6 months from purchase to full activation.

Behavioral change. Using Copilot effectively requires users to learn new interaction patterns — when to invoke the assistant, how to write effective prompts, which tasks to delegate versus complete manually. Microsoft's own usage data suggests that "power users" who realize significant productivity gains represent roughly 15-20% of activated Copilot seats, while the majority use Copilot sporadically.

Data readiness. Copilot's value is proportional to the quality and accessibility of an organization's data in Microsoft 365. Companies with poorly organized SharePoint sites, inconsistent Teams usage, or fragmented data across multiple platforms see limited Copilot value. A Forrester study in mid-2025 estimated that only 35% of enterprises had data environments mature enough to support "high value" Copilot use cases.

LinkedIn faces none of these barriers. The AI features are free for Premium subscribers (who are already paying), enabled by default, require zero configuration, work on the same interface users have used for years, and draw from a dataset (the professional graph) that is inherently well-structured and maintained.

The result is a stark adoption gap:

This table illustrates what might be the most important lesson in AI product strategy: the best AI products are the ones users don't realize are AI. LinkedIn's AI features don't require users to "try AI." They just make the existing product better. The compose box suggests better phrasing. The job feed surfaces more relevant roles. The recruiter search returns better candidates. The user experiences improved outcomes without consciously engaging with "an AI product."

This is the "invisible AI" thesis, and LinkedIn is its most compelling proof point.

International Growth and the Professional Identity Platform

LinkedIn's growth story extends beyond North America in ways that don't get sufficient attention. The platform now has over 300 million members in Asia-Pacific, more than 250 million in Europe, and rapidly growing presences in Latin America, the Middle East, and Africa. In India alone, LinkedIn has over 130 million members, making it the second-largest market after the United States.

International markets are where LinkedIn's AI features have the most transformative potential, because they address a structural problem that doesn't exist in the U.S.: professional identity fragmentation.

In the United States, the professional identity ecosystem is relatively mature. People have Social Security numbers, credit histories, established employment verification systems, and standardized educational credentials. In much of the developing world, professional identity is fragmented, unverifiable, and paper-based. A software engineer in Lagos, a marketing manager in Jakarta, or a financial analyst in São Paulo may have deep professional expertise but no standardized way to signal that expertise to global employers.

LinkedIn's AI solves this in two ways. First, the AI profile optimization features help international members present their credentials in formats that global employers and recruiters recognize. A member in India whose profile describes their role using local terminology gets AI suggestions to add globally recognized skill keywords and description patterns. Second, AI job matching can evaluate candidates across linguistic and credential-system boundaries — matching a Brazilian data scientist's experience against a U.S.-based job posting by understanding the substance of their work rather than pattern-matching on credential names.

The commercial implications are significant. LinkedIn's average revenue per user (ARPU) in North America is approximately $42, compared to roughly $8 in APAC and $15 in EMEA. Closing even a fraction of that ARPU gap through better monetization of AI-powered Premium and Recruiter products in international markets represents a multi-billion-dollar opportunity.

LinkedIn has been investing accordingly. In Q3 2025, the company launched localized AI features in 14 languages, including Hindi, Portuguese, Indonesian, Arabic, and Vietnamese. Recruiter AI matching now works across language boundaries, and AI writing tools adapt to regional professional communication norms. International Premium subscriber growth outpaced North American growth by approximately 2:1 in H2 2025.

The deeper strategic play is positioning LinkedIn as the global professional identity platform — the default infrastructure layer for how professionals are identified, verified, and matched worldwide. If LinkedIn succeeds, every AI-powered hiring platform, every freelance marketplace, and every professional credentialing system will either build on LinkedIn's data or compete against it. That's a platform position, not a social media position, and it justifies a fundamentally different valuation framework.

The Financial Framework: What LinkedIn Would Be Worth Standalone

An exercise that Microsoft investors should conduct but rarely do: what would LinkedIn be worth as an independent public company?

The comparable set is illustrative. Take the publicly traded companies that most closely resemble LinkedIn's business lines:

A sum-of-the-parts analysis using conservative multiples suggests LinkedIn's standalone enterprise value would be $65-95 billion. Using the multiples that high-growth SaaS companies with AI narratives command today — 10-15x forward revenue — the number pushes toward $120-180 billion.

Microsoft paid $26.2 billion in 2016. Even at the conservative end of the standalone valuation range, that's a 3-4x return over nine years on an asset that many analysts considered overpriced at the time of acquisition. At the aggressive end, it's among the best large-cap acquisitions in technology history.

The more interesting question is what LinkedIn's AI-driven growth trajectory does to Microsoft's overall valuation. If LinkedIn can sustain 14-16% revenue growth (plausible given Premium momentum, international expansion, and Recruiter AI upsells), the business will cross $25 billion in revenue by CY2028. At Microsoft's blended forward multiple, that growth contributes roughly $200-350 billion in market capitalization — more than the entire market cap of most S&P 500 companies.

And yet, analysts spend approximately zero time on LinkedIn during earnings calls.

The Risk Factors Nobody Mentions

No bull case is complete without the bear case. LinkedIn's AI-driven growth faces three genuine risks.

Regulatory risk around AI and professional data. The EU's AI Act classifies AI systems used in employment decisions as "high-risk," requiring transparency, human oversight, and bias auditing. LinkedIn's AI job matching and recruiter tools will need to comply with these requirements by August 2026. The compliance cost is nontrivial, and the operational constraints — such as providing candidates with explanations of why they were or weren't surfaced for a role — could limit the effectiveness of some AI features. The EEOC's guidance on AI in hiring, while not binding, adds additional regulatory scrutiny in the U.S.

Content quality degradation. The same AI writing tools driving Premium growth are also flooding the platform with formulaic, AI-generated content. If LinkedIn's feed becomes indistinguishable from AI slop — and some would argue it's already heading there — engagement quality will decline even as engagement quantity increases. The "professional relevance" algorithm update in Q4 2025 is an acknowledgment of this risk, but it's unclear whether algorithmic tuning can solve a problem that's fundamentally about incentives. When every user has access to AI writing tools, the marginal value of AI-assisted content approaches zero.

Competition from AI-native professional platforms. LinkedIn has operated without a serious competitor for over a decade, but the AI era is spawning new entrants. Braintrust, a decentralized talent network, uses AI matching and has attracted significant venture funding. Polywork is building an AI-first professional identity layer. Even X (formerly Twitter) has been expanding into professional networking features. None of these competitors has LinkedIn's data moat today, but the history of technology platforms suggests that data moats are more permeable than they appear — especially when a paradigm shift (like AI) changes the basis of competition.

What This Means for AI Strategy More Broadly

LinkedIn's success offers three lessons that extend well beyond Microsoft.

Lesson one: AI monetization favors embedded features over standalone products. The highest-ROI AI implementations in 2025-2026 are not chatbots, copilots, or agents. They're AI features embedded into products that users already pay for and already use daily. LinkedIn's AI writing tools. Spotify's AI DJ. Netflix's AI-improved recommendation engine. The common thread: users experience better outcomes without consciously "using AI," and the monetization flows through existing revenue lines (subscriptions, ads, premium tiers) rather than through a new AI-specific pricing tier.

Lesson two: proprietary data is the real AI moat, not model capability. LinkedIn doesn't have the best language model. It runs inference on OpenAI and internal Microsoft models that are available to every Azure customer. What LinkedIn has that nobody else has is the professional identity graph — 1 billion members' career histories, skills, connections, and behavioral data. That data makes generic models produce specific, high-value outputs. This validates the broader thesis that AI value accrues to data owners, not model builders, a framework that has massive implications for which companies will win the AI era.

Lesson three: distribution beats technology, every time. Microsoft spent over $13 billion investing in OpenAI and building Copilot. LinkedIn spent a fraction of that embedding AI features into an existing product with 1 billion users. LinkedIn's AI revenue contribution, measured by the incremental revenue attributable to AI-driven features, likely exceeds Copilot's by a significant margin. The technology behind Copilot is arguably more impressive. The business outcome from LinkedIn's AI is inarguably better. Distribution always wins.

The Bottom Line

The AI investment thesis for Microsoft is not wrong. It's incomplete. The market prices Microsoft's AI opportunity primarily through Azure (infrastructure) and Copilot (productivity). LinkedIn barely registers in the AI narrative. That's a mispricing.

LinkedIn is a $18.3 billion revenue business growing at 14%, with operating margins in the high 30s, powered by AI features that 1 billion professionals use, monetized through four distinct revenue lines, protected by the most valuable proprietary dataset in professional AI, and positioned to become the global professional identity platform.

It is, by any reasonable definition, the most profitable AI product in Microsoft's portfolio. It's just invisible — which, as it turns out, is exactly what makes it work.