Alphabet's 70% Return

Programming note: our next issue lands 2 March, where we will be diving into the market pressures facing software platforms.

The $4 Trillion Vibe Shift

If you rewind the tape just a year or two, the consensus narrative on Wall Street was that Alphabet was a vulnerable incumbent trapped by its own success. Google's search monopoly was about to be disrupted by a clean, white text box from OpenAI and Perplexity. The "10 blue links" were obsolete. The ad business was doomed.

Today, Alphabet is a $4 trillion company. Its stock has surged by roughly 71% over the past 12 months alone, cementing its status as one of the biggest winners of the AI boom.

But if you actually look under the hood of that return, you find a fascinating mechanical truth. Google didn't invent 70% more profits. They just convinced the market to change its mind.

The Math of a Comeback

When a stock goes up, it generally happens for one of two reasons: the company is expected to make more money (earnings growth), or investors decide they are willing to pay a higher price for every dollar the company makes (multiple expansion).

A look at the breakdown of Google's 1-year run reveals exactly what kind of rally this is. We ran the decomposition of the stock's 71% return into multiple expansion versus earnings growth, and the results are lopsided:

• Forward Earnings Estimates: Only 16.3% of the return is attributable to Wall Street upgrading its earnings expectations for Google.
• Multiple Expansion: A massive 83.7% of the return is pure multiple re-rating.

In additive terms, roughly 59 percentage points of that 71% rally came from multiple expansion, and only 11 points came from EPS upgrades. Alphabet's forward price-to-earnings multiple went from a depressed 14x in mid-2024 to around 28x today.

In financial terms, Google's most successful product launch of the past year wasn't a new language model; it was a vibe shift.

Defense, Offense, and the Full-Stack Premium

Why did the multiple expand so violently? First, the defense held.

The core bear thesis was that AI chatbots would cannibalize search. But as it turns out, shoving an AI summary at the top of the search page actually increased engagement. Google's "AI Overviews" caused search impressions to jump 49% year-over-year. Furthermore, they secured their distribution moat the old-fashioned way: by writing a giant check to Apple to power iOS AI features, reminding the market that raw algorithmic intelligence doesn't matter if you don't control the real estate.

But the re-rating isn't just a "relief premium" for search not dying. Wall Street has increasingly accepted that Google is actually an AI leader.

As Bloomberg recently noted, the market is waking up to the fact that Google's incumbent advantages are virtually impossible to replicate:

The rapidly improving sentiment reflects of Alphabet is seen as having dominant positions across key areas of AI. Notably, rave reviews for the company's latest Gemini AI model helped ease concerns about competition from companies like OpenAI, while its tensor processing unit chips are seen as a potentially significant driver of future revenue growth.

“The company's competitive advantages are not only intact but growing, and no other player has the models, compute, applications, talent, and data to succeed in AI,” said Divyaunsh Divatia, research analyst at Janus Henderson Investors. “It is certainly being seen as a big AI winner, and that strength helps investors feel more comfortable giving it a more-premium valuation than it previously got.”

Google is proving it has the full stack. On the consumer side, the Gemini app rocketed from 9 million monthly users a year ago to 750 million today. On the enterprise side, they are one of the only companies with the financial might to actively challenge Nvidia's chip monopoly.

Google is currently expanding its Tensor Processing Unit (TPU) offerings to third parties and actively investing hundreds of millions into "neocloud" startups like Fluidstack. They are essentially subsidizing an entire alternative cloud ecosystem just to ensure the AI industry isn't entirely reliant on Nvidia GPUs.

Pricing the Infrastructure Moat

There is an irony to Google's current valuation, which becomes obvious the moment you compare it to its biggest rival.

Usually, when a high-margin software company starts spending tens of billions of dollars pouring concrete and buying chips, investors compress the multiple. Just look at Microsoft. Over the last year or two, as its AI capital expenditures have ballooned, Microsoft's stock has largely drifted sideways. The market is increasingly nervous, punishing the stock as Microsoft's core Azure cloud growth begins to cool.

Wall Street is realizing that Microsoft's capex looks like a margin-crushing transfer of wealth. Despite its near-monopoly in enterprise software, Microsoft is fundamentally renting its AI future. They are paying the "Nvidia tax" for expensive GPUs, and they rely heavily on OpenAI for their foundation models (which also account for a terrifying 45% of Microsoft's entire $625 billion cloud backlog).

Now look at Google. Over the exact same 12-month period, Google was aggressively ramping up its own capital intensity—a spending spree that recently culminated in a forecast to drop an eye-watering $185 billion on capex this year. Normally, this kind of structural pivot to heavy infrastructure would spook tech investors. Instead, the market expanded Google's multiple by 84%.

They did this because Wall Street doesn't seem to mind capital expenditures when you are paying yourself; they hate it when you are paying middlemen.

Investors saw Google's mounting capex bill and decided it wasn't a reckless wealth transfer—it was an impenetrable infrastructure moat. Google isn't just buying Nvidia chips; they are fabricating their own TPUs. They aren't renting a foundation model; they own Gemini.

Investors are paying 28-times forward earnings for an increasingly asset-heavy business because Alphabet is proving it is one of the only companies on Earth with the balance sheet, the custom silicon, and the proprietary language model to survive the AI transition without bleeding out to suppliers. The 70% surge over the last year is the market rewarding Google for vertical integration. They aren't just a software wrapper; they are the power plant, the power lines, and the grid.

More on Hyperscalers:

• Can hyperscalers justify their huge AI capex? (CNBC)
• Hyperscaler capex > $600 bn in 2026 a 36% increase over 2025 (IEEE)

On Our Radar

Our Intelligence Desk connects the dots across functions—from GTM to Operations—and delivers intelligence tailored for specific roles. Learn more about our bespoke streams.

AMD Backstops $300M Loan for AI Startup Crusoe to Drive Chip Sales

• The Story: AMD has agreed to provide a $300 million loan guarantee for AI cloud startup Crusoe, offering to lease back its own chips if Crusoe fails to find customers, which allowed the startup to secure a significantly lower interest rate. (Reuters)
• The Operations Implication: This deal signals AMD's entry into vendor financing to compete with Nvidia’s market-building tactics. By assuming the credit risk and residual value risk of its own hardware, AMD is operationally subsidizing the growth of neocloud providers to secure guaranteed order volume, effectively trading balance sheet liability for market share in the AI infrastructure race.

Salesforce Executes M&A Spree to Patch Agentforce Weaknesses

• The Story: Salesforce has completed ten acquisitions in six months, most recently acquiring conversational insights platform Momentum, as it moves to systematically "buy its way" out of Agentforce’s early stumbles in data quality, process visibility, and intent-driven discovery. (CX Today)
• The Product Implication: This aggressive M&A cycle confirms that the initial version of Agentforce lacked the industrial-grade data and behavioral intelligence layers required for complex enterprise automation. By acquiring domain specialists in process mining (Apromore) and intent intelligence (Cimulate), Salesforce is attempting to transform Agentforce from a conversational UI into a robust, autonomous execution engine, a necessary product pivot to maintain its "system of record" moat against general-purpose AI assistants.

P.S. Tracking these kinds of complex, cross-functional signals is what we do. If you have a specific intelligence challenge that goes beyond the headlines, get in touch to design your custom intelligence.

You received this message because you are subscribed to ARPU newsletter. If a friend forwarded you this message, sign up here to get it in your inbox.