Google Stock Breakdown: Crumbling Monopoly or Misunderstood Opportunity?

Get smarter on investing, business, and personal finance in 5 minutes.

Stock Breakdown

This newsletter is adapted from a recent video Drew Cohen did on Google.

Investors commonly treat Google as just another large technology company, yet Google differs from that group in one respect that matters most…

Its core business sits under direct threat from AI.

Google Search still generates about 75% of Google's earnings, and for the first time in decades, that business faces real competitive pressure.

Can Google transition its search business into an AI business that holds a comparable profit margin?

And what return will come from the hundreds of billions of dollars Google is committing to AI infrastructure?

In this week’s Five Minute Money, we break down Google's business segments and history, the risk facing Google Search, the AI infrastructure build inside Google Cloud, and a valuation built two separate ways.

Business Overview.

Revenue splits into four categories: 1) Google Advertising, 2) Google Subscriptions, Platforms, and Devices, 3) Google Cloud, and 4) Other Bets.

Google Advertising and Google Subscriptions, Platforms, and Devices together make up Google Services.

Let’s go into each revenue segment one by one.

Google Advertising.

Google Advertising separates into three distinct businesses: Google Search and Other, YouTube ads, and Google Network.

Google Search and Other generated $243bn in revenue over the past 12 months, growing 17 y/y, and represents 55% of Google's total revenue and is likely its highest-margin segment.

By one estimate, Google Search accounts for somewhere between 70% and 80% of the company's overall profit.

It remains the main driver of the business today.

The segment includes advertising shown on Google Maps, inside Gmail, and on the Google Play Store, in addition to the core search results page.

YouTube ads make up the second business inside Google Advertising, generating $43bn in revenue and growing 12% y/y.

This figure covers only advertising revenue on YouTube.

YouTube Premium and YouTube TV sit in a separate segment discussed below.

Google Network is the third business, and it functions differently from the other two.

Instead of placing ads on properties Google owns, Google Network places ads across the open web through products such as AdSense for publishers, AdMob for mobile app advertising, and Google Ad Manager.

When a website visitor loads a page and an ad appears, Google is running an auction in real time to determine which ad gets shown.

Google Network generated $30bn in revenue but contracted 2% over the past 12 months, a notable exception to the growth seen across the rest of Google's advertising business.

Part of the decline traces to rising use of large language models for informational searches, and part of it traces to AI overviews appearing directly inside Google Search results.

By one estimate, roughly 65% of searches now end without a click to any website, a pattern known as a zero-click search.

Zero-click searches hurt Google Network specifically because that business only earns money once a user clicks into a website and has the opportunity to view an ad there.

This represents a structural issue across the open internet generally.

The effect on Google itself is smaller, because Google Network carries much lower margins than the rest of Google's advertising business.

Google does not own the ad inventory in this business; it purchases that inventory, a cost known as the traffic acquisition cost, and passes most of that money to the publisher.

What does matter for Google Search specifically is that Google Network's structural decline offers an early signal of where search traffic has been heading, a topic we will covered in detail later in this breakdown.

Subscriptions, Platforms, and Devices.

The second major category, Subscriptions, Platforms, and Devices, generated $52bn in revenue and grew 18% y/y.

YouTube Premium represents one of the largest subscription products in this category, and it includes YouTube Music as part of the bundle.

YouTube TV sits separately, priced at roughly $85 a month for access to a full set of linear television channels delivered through a connected TV instead of satellite or cable.

The product has become particularly popular for live sports, and Google is now believed to rank as the third-largest pay television provider in the United States.

Hardware makes up another piece of this category through the Pixel device line, covering smartphones and home devices.

Hardware revenue has never been especially significant for Google, though Pixel increasingly serves as a showcase for the company's AI capabilities, and recent versions have incorporated AI features faster than Apple has matched on the iPhone.

In-app purchases add a further revenue stream to this category: whenever someone buys a virtual item inside a mobile app, Google collects a 30% share of that transaction.

Gemini subscriptions, tied to the AI Pro feature, complete the category.

Combined with Google Advertising, this category forms Google Services, a $370bn revenue business that grew 15% y/y and still represents the core of Google's business today.

Google Cloud.

Google Cloud generated $78bn in revenue over the past 12 months, though growth has been accelerating sharply: revenue grew more than 80% in the most recent quarter alone, and annual recurring revenue now stands at $100bn.

Operating margins sit in the mid-30s percent range today, a dramatic shift from several years ago, when the segment was still losing money.

The improvement reflects substantial operating leverage as revenue has scaled.

Cloud computing traditionally organizes into layers.

The infrastructure layer provides raw compute capacity.

The platform layer sits above it, offering developer tools, databases, and software kits that a company integrates into its own products.

Google's database product BigQuery is a good example: once a company builds its software around BigQuery, switching providers becomes considerably harder, since the integration reaches deep into the company's own code and becomes difficult to unwind.

This platform layer is what makes a cloud provider sticky and supports pricing power over time.

A company that simply rents raw computing capacity can switch providers far more easily.

More recently, a separate and newer architecture has grown alongside this traditional, enterprise-focused, CPU-first cloud model: a GPU-first architecture built specifically to run AI workloads.

Many of the newer cloud providers known as neo clouds, including CoreWeave, originated as cryptocurrency mining operations, which meant they already operated GPU-heavy data centers when AI workloads emerged as a far larger opportunity.

Those companies redirected existing GPU capacity toward AI compute, while the traditional hyperscalers, including Google, had to add this GPU-first infrastructure onto an existing CPU-first base.

Google also designs its own AI chip, called a TPU, as an alternative to NVIDIA's GPUs.

A lower-cost, more specialized chip gives Google a potential structural cost advantage in its own data centers, even without matching NVIDIA's top-end performance, and Google sells TPU chips directly to outside customers as well, including Meta and OpenAI.

Google increasingly bundles its cloud infrastructure together with its own AI models under the Gemini brand, selling what the company refers to internally as AI intelligence, a bundled product distinct from raw compute sold on its own.

Other Bets.

Waymo stands as the most significant business inside Other Bets.

The robotaxi service currently owns its entire fleet of vehicles and operates its own ride-hail product directly, though Google has discussed a future model built around licensing the underlying technology to vehicle manufacturers.

A driver buying a Ford or a BMW in the future could potentially get access to driverless technology built on Waymo's platform, offering a way for traditional automakers to compete against Tesla's current full self-driving offering.

Waymo's last funding round, held separately from the rest of Google, valued the business at $125bn, and Google retains roughly 80% ownership, putting its stake at close to $100bn.

A handful of smaller businesses also sit inside Other Bets.

Verily operates as a life sciences venture.

Calico focuses on longevity-related research.

Google Fiber provides internet service in a limited set of markets.

Google Ventures, the company's venture capital arm, has built a strong track record of investments over time, including a stake believed to include Anthropic, and the entire portfolio is valued at several hundred billion dollars.

Google also holds a stake in SpaceX, believed to be worth around $120bn.

Taken together, Other Bets generated roughly $1.5bn in revenue, down -7% y/y, alongside an operating loss of $9bn.

Altogether, Google generated $445bn in revenue over the past 12 months, growing 20% y/y.

On the operating income side, Google discloses far less detail than it discloses on revenue, reporting only three segments.

Google Services generated $154bn in operating income.

Google Cloud generated $24bn in operating income over the trailing 12 months, though annualizing the most recent quarter alone implies closer to $36bn, a 36% margin, with operating income up 200% y/y, a sign of a business still in the early stages of its profit ramp.

Other Bets lost $9bn.

A separate corporate overhead line, covering shared expenses that cannot reasonably be allocated to any single segment, subtracted a further $21bn.

The Google Search Risk.

Google Search faces its first serious competitive threat in roughly a decade and a half.

Understanding why requires looking first at how Google built its dominant position in the first place, since the same forces now shape how Google is responding to AI.

Google entered the search engine market relatively late, behind several competitors that were larger at the time.

Google won initially because it built a better product, but the bigger factor over time was how aggressively Google pursued distribution, a pattern that matters directly to how Google is positioning itself in AI today.

Google pursued three separate distribution strategies.

The first was the Chrome browser, released for free, with Google Search set as the default search engine inside it.

Anyone who downloaded Chrome became, without necessarily realizing it, a committed Google Search user.

The second strategy involved Android.

As computing shifted from desktop to mobile, Google worried that device manufacturers might build or favor their own search products instead of Google's.

Google responded by making Android free to use, on the condition that manufacturers preload the full suite of Google apps, including Chrome, with Google Search set as the default.

The third strategy involved Apple, the one major device manufacturer that did not run Android.

Google reached a deal with Apple built around a payment known as the traffic acquisition cost.

Antitrust proceedings later revealed that Google pays Apple 36% of the advertising revenue generated by Google searches performed on Apple devices, a payment that has become one of the largest contributors to Apple's services revenue.

These three strategies, taken together, explain why Google Search became so deeply embedded across the internet.

The lesson carries forward directly into how Google approaches AI.

Distribution has mattered to Google's position for a long time, even during the years when Google Search faced essentially no real competition.

Vertical Competition and the Rise of Zero-Click Search.

A decade ago, a different competitive concern dominated discussion around Google: vertical-specific search.

Sites such as Trivago aggregated listings across travel booking platforms, including Booking, Expedia, and hotels.com, giving users a reason to begin a travel search outside Google entirely.

Similar concerns existed around e-commerce search and around Yelp for local search.

Google's response was to build its own version of each vertical and surface those results directly inside Google Search itself.

Google Flights, Google Hotels, Google Finance, Google Local, and Google Shopping all emerged from this strategy.

Google effectively became its own meta-search engine, and no outside company needed to fill that role.

None of the standalone vertical-search competitors ever grew into a serious threat.

A separate trend developed alongside this one: users clicking on search results less often, even before AI entered the picture.

Google increasingly displayed information directly inside search results, often eliminating the need for a click to a separate website; a search for the weather, for example, could be answered directly on the results page.

This pattern, known as a zero-click search, has existed for years.

AI overviews and AI mode have intensified it considerably, since Google can now summarize an entire search response directly within the AI-generated answer, further reducing any need to click into an outside website.

Despite the rise of zero-click searches, Google's disclosed paid link volume has continued growing.

Paid links grew 6% in 2025, then accelerated to 13% growth in the most recent quarter, with the average price paid per click up 3% as well.

This pattern makes sense once monetizable searches and non-monetizable searches get treated as two separate categories.

Search activity currently moving away from Google, toward tools such as ChatGPT, Claude, Grok, or Gemini, tends to be informational in nature.

Little of it is commercial.

Someone looking to buy running shoes still tends to go to Google Search, click through listings, and complete a purchase there.

Google has preserved this commercial intent so far, and search revenue overall has grown faster than expected as a result.

One detail complicates the picture of advertiser value inside Google Search: branded keyword bidding.

A company such as Nike still has to bid on the keyword "Nike" inside Google Search, even though Nike would likely rank first organically regardless, because a competitor such as Adidas can bid on that same keyword and win a meaningful share of those clicks if Nike does not also bid.

Advertisers run lift studies confirming a positive return on this spending, and those studies are accurate as far as they go, but none of them compare results against a scenario with zero sponsored ads present at all.

Estimates vary, but research cited in this breakdown places the share of Google ad revenue coming from branded or competitor-name bidding somewhere between 20% and 35%, a potentially material figure if accurate.

Google's own strategy around AI reflects an understanding of these dynamics.

Over 1bn people used AI mode in the past month, 2.5bn people have used AI overviews, and 950 million people now use Gemini.

Google is deliberately pushing AI across every product it controls, from Search to Chrome to its full app suite, aiming to own the AI interface before a competitor does and introduce advertising into that interface later.

Advertising inside an AI conversation appears likely to arrive eventually; OpenAI's free tier already carries advertising.

What remains uncertain is whether that advertising will match the margin profile of a traditional Google Search ad, a product that costs little more than displaying a link and remains among the highest-margin products in all of technology.

An AI conversation, by contrast, consumes meaningfully more compute per interaction, which likely caps the margin such ads can ultimately achieve.

A more serious risk to Google Search sits beyond the advertising-margin question: the possibility that monetizable searches disappear from Google Search entirely through agentic AI.

Conversational AI has reached mass adoption, but agentic AI, meaning an AI system that takes real actions such as shopping or booking travel on a user's behalf, remains in an early stage.

Meta's Muse product, with roughly 10 million users, represents one of the first examples of agentic AI reaching a broad consumer audience.

Meta has disclosed that 25% of Muse users have shopped through the product, spending an average of $200.

Some users have also booked travel directly through Muse.

A traditional Google search for a hotel typically surfaces sponsored listings from Booking, Expedia, and other booking platforms, generating monetizable clicks for Google along the way.

None of that activity occurs, and none of that revenue reaches Google, if a user completes the same travel booking entirely inside an AI agent instead.

Google has its own agentic AI product, called Google Spark, though usage numbers remain undisclosed and awareness of the product appears limited.

Whether Google can scale Google Spark quickly enough to compete with agentic products from Meta, OpenAI, and others will significantly shape how much of this risk actually materializes.

A further long-term risk involves the next shift in computing hardware itself.

Computing has already shifted once from desktop to mobile, and a further shift toward augmented reality, virtual reality, or voice-first devices could change search behavior again, potentially moving toward a more faceless form of interaction where a user simply tells an AI agent what they need, without a traditional search interface involved at any point.

Google Cloud and the AI Infrastructure Bet.

Google Cloud has become a genuinely large business: $100bn in annual recurring revenue, operating margins between 35% and 36%, and revenue growth exceeding 80% in the most recent quarter alone, with growth still accelerating.

The contractual backlog exceeds $500bn, and Google expects to realize about half of that backlog over the next 24 months.

Customer quality varies considerably within this business.

Google does not disclose contract-level detail publicly, but known customers include Anthropic and OpenAI, both of which were built around a multi-cloud strategy from the start.

A customer built to run across multiple cloud providers can, in principle, shift workloads toward whichever provider offers the lowest price, a dynamic currently masked by the fact that compute capacity across the entire industry remains constrained.

An enterprise customer whose software is deeply integrated into Google's platform layer, through a tool such as BigQuery, represents a stickier and more durable relationship than a customer buying raw compute capacity alone.

The competitive field inside cloud computing has widened substantially.

What began as roughly a duopoly, followed by three serious players once Google entered the market, now includes Microsoft, Google, Amazon, Meta, and Oracle, alongside SpaceX and a growing list of neo clouds.

Nearly every company on this list is betting that AI compute demand will keep rising indefinitely, and several technology leaders appear to view underbuilding, and missing the AI opportunity entirely, as a far worse outcome than overbuilding and absorbing a few years of weaker margins.

Every capital cycle in history has followed a similar pattern, without exception so far: innovation attracts capital, early returns on that capital look strong, and more capital continues entering the space in response.

Supply eventually overshoots demand, and when that happens, pricing pressure follows, since excess capacity has to be absorbed by lowering prices to attract enough usage.

None of this implies that AI lacks real value or will fail to transform daily life and the broader economy over time; the concern here is purely a matter of timing.

If total AI infrastructure spending runs even a few years ahead of actual demand, a scenario sometimes described as an air pocket, the result is a period of real pricing pressure and contraction across the entire supply chain as companies reduce capital expenditure plans temporarily, prices on compute decline, and the industry moves toward a new equilibrium between supply and demand.

A scenario in which global AI infrastructure spending, now committed by essentially every major technology company at once, ends up perfectly matched to real-world demand with no overshoot at any point would be unusual relative to how previous capital cycles have unfolded historically.

That does not make such a scenario impossible.

If an air pocket does occur, Google appears better positioned than many of its cloud competitors, for reasons discussed below, though a better relative position would not necessarily spare Google from a difficult multi-year stretch if a correction of this kind actually occurs.

Why Google Could Be Better Insulated.

Three factors could leave Google better positioned than other cloud providers if an air pocket does occur.

The first is Google's TPU chip, which offers a potential structural cost advantage and the ability to deliver cheaper inference relative to competitors dependent entirely on NVIDIA.

The second is Google's strategy of selling AI intelligence, the bundled combination of cloud infrastructure and Google's own AI models, instead of raw compute alone; Google has stated that most new contracts now involve this bundled offering, which creates a stickier, less easily substituted customer relationship than a contract built around raw compute capacity.

The third is flexibility: if external demand for Google Cloud capacity falls short of expectations, Google can redirect that same capacity toward its own products, improving ad targeting across YouTube or expanding capacity for AI mode.

Over the trailing 12 months, even accounting for the scale of Google's AI spending, the estimated return on incremental capital still comes out to approximately 19%.

This remains a risk worth monitoring closely, since the underlying cycle could shift quickly: contractual commitments could be reduced, and a portion of this investment picture could unwind faster than expected.

Given that nearly every major technology company and cloud provider is currently pursuing the same opportunity simultaneously, often outspending one another to do so, a point at which supply exceeds demand across the industry looks like a reasonably likely outcome at some point.

None of this implies a negative view of AI as a technology; it simply reflects a question of timing and risk management around how quickly that demand will actually materialize.

Valuation.

Valuing Google is a genuinely complicated exercise given how many distinct businesses sit inside one company.

We are going to value Google two separate ways: a sum-of-the-parts analysis first, followed by a simpler, more traditional growth and multiple-based method.

Sum of the Parts Valuation.

Sum-of-the-parts analysis carries a well-known limitation: in a genuinely strong business, the individual pieces are usually worth more combined than separated, and in a weak business, a sum-of-the-parts calculation can overstate value, since a breakup rarely gets realized in practice outside of activist investor intervention.

Google's businesses are distinct enough from each other, though, that the method still produces something useful here.

Google trades at approximately $320 a share, implying a fully diluted market cap of $3.9 trillion.

Google holds roughly $260bn in net cash and long-term investments, including about $55bn in cash and the rest in marketable securities and investment stakes such as SpaceX, a stake that may now be understated by $30-40bn relative to its last formal valuation mark.

Subtracting that $260bn from the market capitalization produces an enterprise value of $3.65 trillion.

Consolidated operating income, taxed at an assumed 20% rate, produces $117bn in net operating profit after tax, or NOPAT, implying a trailing multiple of 31x.

From that starting point, individual pieces of the business get removed one at a time.

Excluding the $9bn operating loss generated by Other Bets, on the logic that a loss-generating business should not carry a positive multiple, and excluding Google's roughly $100bn Waymo stake, reduces the multiple to 28x.

Removing YouTube next, estimated at $65bn in revenue with an assumed 30% operating margin given the high revenue share YouTube pays to creators, produces $16bn in NOPAT; applying a 30x multiple values YouTube at approximately $500bn, which also gets excluded from the remaining enterprise value.

What remains after excluding Other Bets, Waymo, and YouTube is $3.05 trillion in enterprise value, covering Google Cloud and Google Services combined.

Google Cloud, a $100bn annual recurring revenue business, gets modeled forward to 2030 under two separate growth assumptions: a 25% compound annual growth rate and a 35% compound annual growth rate, producing projected 2030 revenue of $210bn and $300bn.

Applying margin assumptions to each scenario produces projected NOPAT of $60bn and $95bn.

A 22x multiple produces a Google Cloud valuation range of $1.3 trillion to $2.1 trillion once discounted back 10% to reflect present value.

Excluding Google Cloud from the remaining $3.05 trillion leaves Google Services, Google's advertising, subscriptions, and devices business with YouTube already removed, carrying an implied multiple of between 19-25x, depending on which end of the Google Cloud valuation range gets applied.

An investor who views Google Search as a durable secular grower, with AI engagement continuing to strengthen and more advertising likely to appear inside AI products over time, might reasonably argue that 19x looks too low and that a fairer multiple sits closer to 25x or 30x, a case for meaningful undervaluation at current prices.

An investor more concerned about Google Search facing a slow, ongoing decline, sometimes described as a melting ice cube, might instead view 19x as roughly fair, or even still somewhat generous.

Scenario-Based Growth Method.

This method grows Google Cloud and the rest of Google, meaning Google Services broadly, at different assumed rates out to 2030, holding Google Services margins flat and varying the margin assumption applied to Google Cloud across three scenarios: bear, base, and bull.

The bear case assumes 25% growth for Google Cloud, a relatively mild bear scenario, reaching approximately $200bn in revenue at a 25% margin, with the rest of Google growing 5%.

This produces projected 2030 NOPAT of $175bn.

The base case assumes 35% growth for Google Cloud at a 35% margin, with the rest of Google growing 10%, producing projected 2030 NOPAT of $230bn.

The bull case assumes the same 35% growth for Google Cloud but at a 40% margin, with the rest of Google growing 15%, producing projected 2030 NOPAT of $295bn.

Translating these NOPAT figures against today's enterprise value produces implied multiples of 20x 2030 NOPAT in the bear case, 16x in the base case, and 12x in the bull case.

Because Google carries no meaningful debt complicating this picture, the investor's forward return becomes primarily a function of which exit multiple gets applied at that point in 2030.

In the bull case, applying a 25x exit multiple, reasonable given a business still growing at a low-to-mid double-digit rate even in 2030, to a 12x current implied multiple roughly doubles the investment, close to 100% upside over the period.

In the bear case, applying a 22-23x exit multiple to a 20x implied multiple produces only modest upside over four years.

In the base case, applying a multiple in the low 20s to a 16x implied multiple corresponds to a return in roughly the 8% to 9% compound annual growth rate range.

Several distinct outcomes remain plausible from here.

Google Search could face disruption faster and more severely than current numbers suggest, as more monetizable search volume shifts toward agentic AI.

Alternatively, the broader AI infrastructure buildout could experience a moderate air pocket, where demand fails to meet committed capacity on schedule and pricing pressure follows across the industry.

A third outcome involves none of these risks fully materializing: Google Search transitions smoothly into AI-driven monetization, and computing demand continues growing in line with the capital every major company in this space has already committed.

Ultimately, investors have to make their own judgment and whether the potential return is worth it.

For more on Google, check out this video below.

YouTube FAQ

Nothing in this newsletter is investment advice nor should be construed as such. Contributors to the newsletter may own securities discussed. Furthermore, accounts contributors advise on may also have positions in companies discussed. Please see our full disclaimers here.