Google’s Cloud backlog hit $460 billion last quarter. It couldn’t fill the orders. So it just raised $85 billion to buy the concrete, silicon, and fiber it needs to ship product to customers who have already signed.
The banker’s grim calculation mirrors Alphabet’s multi-currency debt and equity raise to fund AI infrastructure.
On Monday, Alphabet closed a $84.75 billion equity capital raise, the largest in the company’s history. The package includes a $10 billion private placement from Berkshire Hathaway, an upsized $34.75 billion public offering of stock and mandatory convertible preferred shares, and a newly established $40 billion at-the-market (ATM) program. The consensus reads this as a growth story. The structure of the raise betrays a deep anxiety about control. Alphabet is not raising capital to chase a future opportunity. It is buying the infrastructure necessary to honor existing contracts it cannot currently fulfill. The company is so supply-constrained that it had to dilute its shareholders just to deliver.
The Mechanism of Control
As hyperscalers clash in a margin-crushing price war, savvy players lock in long-term deals at rock-bottom prices.
The mechanics of the raise reveal a management team staring down a logistics crisis. The underwritten portion was oversubscribed and priced at roughly $35 billion, split between an $18 billion offering of Class A and Class C stock and $16.75 billion in depositary shares representing mandatory convertible preferred stock, according to an SEC filing. Berkshire Hathaway anchored the deal, buying $5 billion in Class A stock at $351.81 per share and $5 billion in Class C stock at $348.20 per share.
That is the headline. The fine print is the ATM program. Alphabet intends to drip $40 billion of stock into the open market beginning in the third quarter of 2026. The company explicitly states that approximately $30 billion of those ATM proceeds will be used to meet the 2026 calendar year tax obligations related to employee equity awards. This is the critical detail. Alphabet is issuing new equity to pay the tax bill on old equity. It is a dilutive spiral that only makes sense if management believes two things: first, that the stock is currently valued highly enough to make equity issuance cheaper than the alternative of letting employees sell into the market to cover their own tax bills, and second, that the AI revenue curve will accelerate fast enough to absorb the dilution before it craters earnings per share.
The company has already levered up. Alphabet ended the first quarter of 2026 with $81 billion in debt. Since the quarter closed, it issued approximately $20 billion in additional debt across six major currencies, bringing its pro-forma debt balance to just over $100 billion, according to company disclosures. This follows a $30 billion global bond issuance in February and an $11 billion raise in sterling and Swiss francs in European markets. The balance sheet is not distressed. The company holds $127 billion in cash and marketable securities and generated $174 billion in trailing twelve-month operating cash flow. But the direction of travel is clear. Alphabet is pulling every capital markets lever at once, equity and debt, public and private, to fund a CapEx budget that has exploded from $31 billion in 2022 to $91 billion in 2025, and is now guided to $180 to $190 billion for 2026.
The Supply Constraint Is Real
Sundar Pichai stated the problem without ambiguity in the company’s June investor presentation: demand for AI solutions from enterprises and consumers is “meaningfully exceeding” available supply. The numbers bear this out. Cloud revenue grew 63 percent in the first quarter, exceeding $20 billion for the first time. The backlog nearly doubled quarter-over-quarter to over $460 billion. Gemini Enterprise paid monthly active users grew 40 percent sequentially. Alphabet cannot build data centers fast enough.
The company is betting its custom silicon can solve the supply problem on the unit-economics side. At Cloud Next, Alphabet introduced its eighth-generation TPUs. The TPU 8t offers three times the processing power of Ironwood, the prior generation. The TPU 8i delivers 80 percent better performance per dollar for inference workloads. If those efficiency gains translate to real-world cost curves, Alphabet can serve its backlog at margins its competitors cannot match. That is the bet.
This capacity build-out, however, sets the stage for a broader industry collision.
The Frontier Take: A Bifurcation Is Coming
Here is what I think happens next. Within 18 months, Alphabet’s capacity build-out will force a bifurcation in the cloud market.
The hyperscalers are already spending at a rate that makes the 2022 numbers look like a rounding error. Alphabet, Microsoft, Meta, and Amazon are expected to pour more than $700 billion combined into capital expenditures in 2026, and Wall Street analysts estimate total AI CapEx could climb above $1 trillion in 2027. That is not sustainable without a return. But returns require utilization, and utilization requires price signals that clear the market.
Alphabet is building capacity to serve a backlog that exists today. When that capacity comes online, the company will have every incentive to fill it, even at thin margins, because the alternative is depreciation on idle silicon. AWS and Azure will be forced to match on terms and capacity. The result is a margin-crushing price war on inference workloads, the commodity layer of the AI stack.
The winner will be the enterprise customer who negotiates a multi-year AI commitment now, locking in rates that will look unsustainably low in hindsight. The losers will be the undifferentiated AI startups. A company that has built its entire business on a thin wrapper around someone else’s model cannot compete with near-zero-margin inference served directly by the hyperscaler that owns the data center. The $1 trillion CapEx bubble of 2027 will force an industry-wide reckoning on returns, and the startups that raised on the assumption of persistent pricing power will not survive it.
What to Do Now
Enterprise buyers should be negotiating aggressively. The hyperscalers need to fill capacity to justify the spend, and the window for locking in favorable multi-year terms on inference is open now. Investors should watch for margin compression at AWS and Azure as they respond to Alphabet’s capacity surge, and they should be extremely skeptical of any AI startup whose core value proposition is access to a model rather than a proprietary data moat or workflow integration that cannot be replicated by a cloud provider’s API. Alphabet employees should understand what the $30 billion ATM tax-withholding allocation really means: the company is printing stock to keep you whole because it believes the alternative—letting dilution hit the share price unchecked—is worse.
The $460 billion backlog is not a victory lap. It is a ticking clock. Alphabet just bought itself the runway to deliver on contracts it has already signed. The real test is whether those contracts renew at profitable rates once the price war begins.
