The Alphabet equity raise that landed this week is the largest in tech history: 84.75 billion dollars in a single offering, and the entire pile is going to one thing, AI infrastructure. The part that made me sit up was not the headline number. It was who showed up to write a check. Berkshire Hathaway, Warren Buffett’s company, the outfit that has spent the better part of a decade buying back its own shares and famously sitting on a mountain of cash rather than chasing tech capex, put in 10 billion as the anchor investor. When Buffett’s people decide the smart move is funding data centers, something has shifted.
What actually happened
Quick timeline, because the speed of it tells you something. On June 1, Alphabet announced an 80 billion dollar raise. One day later, after the offering was oversubscribed, they upsized it to 84.75 billion. Demand blew past what they asked for, so they took more. That is not a company struggling to find buyers; that is a feeding frenzy for anything with AI exposure attached.
The structure is worth understanding, because it is not one simple stock sale. It is a stack of levers pulled at once: 18 billion in Class A and Class C common stock, 16.75 billion in mandatory convertible preferred stock, a 40 billion at-the-market program that drips shares into the open market over time starting in Q3, and the 10 billion Berkshire private placement at a negotiated discount. Class A priced around 355 dollars, Class C around 352. One detail I appreciated, mostly because of how loud the opposite has been everywhere else: there is zero crypto in this. No blockchain ventures, no token strategy, no digital-asset allocation. Every dollar is pointed at data centers and compute clusters. Refreshing.
The number behind the number
Here is the context that makes 84.75 billion make sense. Alphabet is guiding to between 180 and 190 billion dollars in capital expenditure for 2026 alone. For reference, they spent about 91 billion in all of 2025, so this is nearly double, in one year. Q1 2026 capex already hit 35.7 billion in a single quarter, more than twice the year-ago figure. And the CFO has already said 2027 will be “significantly” higher again.
So why does a company that prints cash, Alphabet generates enormous free cash flow, need to raise outside money at all? Because the spending has gotten so large it has outrun even Google’s cash generation. When your annual capex is closing in on 190 billion, the bank account alone does not cover it anymore. That alone should reframe how you think about the scale of the current AI buildout. This is the most profitable advertising machine ever constructed, and it still has to pass the hat.
The business read: why this move, and why now
This is the part I find most interesting, and where my day job watching how companies actually allocate capital makes me read it a little differently than a pure gadget take would. There are several signals stacked inside this one raise, and they do not all point the same direction.
Signal one: the constraint is supply, not demand. Sundar Pichai told analysts on the Q1 call, in plain words, “we are compute constrained in the near term,” and that cloud revenue “would have been higher if we were able to meet the demand.” Read that twice. Google is leaving cloud revenue on the table because it physically cannot build capacity fast enough. Google Cloud did 20 billion in Q1 revenue, up 63 percent year on year, with a backlog reported north of 460 billion. When you have a 460 billion dollar backlog and you are turning customers away for lack of compute, raising 85 billion to build faster is not reckless. It is the rational move. The money is chasing contracted demand that already exists, not a bet that demand might show up.
Signal two: equity, not debt, and that is a tell. This is the subtle one. Alphabet has plenty of access to debt markets; it did a 30 billion-plus bond issuance back in February. So why raise equity this time, which dilutes existing shareholders, instead of just borrowing more? Several analysts read it the same way I do: issuing equity is what you do when you think your stock is richly valued. If you believe your shares are expensive, selling them to fund spending is cheaper than taking on debt. Alphabet stock has more than doubled in the past year. Translation: management may quietly believe the market is paying a generous price for Alphabet right now, and they are happy to use that currency. That is not something they will ever say out loud, but the choice of instrument says it for them.
Signal three: the Berkshire stamp. Buffett’s involvement is being read as validation, and it mostly is. Berkshire reversing a decade-long pattern of buybacks and cash-hoarding to anchor an AI-infrastructure raise is a heavyweight vote of confidence that this spending will earn a return. But I would add a flatter, less romantic read too: Berkshire got a negotiated discount, it has been building an Alphabet position since 2025, and a 10 billion preferred-flavored stake in the most profitable company in the world is not exactly a wild gamble for them. It is validation, yes. It is also just a good deal for Berkshire. Both things are true.
Signal four, the one nobody at Alphabet will frame this way: it squeezes the competition. This is where it connects to everything else happening right now. By raising 85 billion in equity at the exact moment Anthropic has filed to go public and OpenAI is preparing its own listing, Alphabet is hoovering up a huge chunk of the available investor appetite for AI. There is a finite pool of capital chasing this theme, and Google just took a giant scoop of it. Analysts have flagged that this could raise the cost of capital for the pure-play AI labs coming to market behind it. It is the same dynamic I wrote about with the SpaceX and OpenAI mega-offerings: whoever reaches the trough first eats best. Alphabet did not need an IPO to get to the trough. It was already there.
The long-term view, and the part that should make you cautious
Zoom out and Alphabet is not alone in this. The four big hyperscalers, Amazon, Alphabet, Microsoft, Meta, are collectively on track to spend more than 700 billion dollars on infrastructure in 2026, roughly three quarters of it on AI. Amazon leads near 200 billion, Alphabet at 180 to 190, Microsoft and Meta each above 100. Some Wall Street estimates have total AI capex climbing past a trillion dollars in 2027. These are national-budget numbers, spent by a handful of companies, on the same bet. It is the demand side of the same story I traced in NVIDIA’s 91 billion dollar revenue forecast: every dollar of this capex lands as revenue on a chipmaker’s books somewhere down the line.
The long-term logic is coherent if you accept one premise: that AI compute is the new strategic chokepoint, the way oil reserves or fab capacity were in earlier eras, and that owning the most of it wins the decade. If that premise holds, then Alphabet diluting shareholders today to build the world’s largest compute network is a bargain, and the analysts slapping 445 and 450 dollar price targets on the stock are right. Google’s edge is real: it has the cloud backlog, the Gemini app reportedly near 900 million monthly users, and the cash flow to service all this. It is arguably the best-positioned of the bunch precisely because the AI spending is feeding businesses, Search and Cloud, that already make money. This is the through-line of how the data revolution and the AI revolution became one story: the infrastructure layer is where the value pools.
But here is where I keep my enthusiasm on a leash. An 84.75 billion dollar equity raise is dilutive by definition; every existing share is now a slightly smaller slice of the company. The market noticed: Alphabet shares slipped more than 4 percent on the week despite all the bullish framing, and faster depreciation from this capex is going to pressure free cash flow through 2026. The real risk is not whether AI demand is real; the Q1 numbers say it is. The risk is execution and timing. Spending 180 to 190 billion in a single year requires supply chains, power, construction pipelines, and talent that may not scale as smoothly as a spreadsheet assumes. And the deeper question, the one the whole industry is now quietly pricing, is whether the returns on infrastructure of this magnitude arrive fast enough to justify the spend, before the next cycle, the next architecture, or the next constraint changes the math.
That is the AI race in 2026: even Google has to pass the hat, and Warren Buffett is the one passing it back full.