Alphabet’s Record Profit Hides Its First AI Cash Burn

Alphabet did two things in the same quarter it had never done together: it booked the largest profit in its history and it spent itself into negative free cash flow. Both trace to the same place — the money Google is pouring into AI.

For the three months ended June 30, 2026, Alphabet reported $119.8 billion in revenue, up 24% from a year earlier, and net income of $112.1 billion, according to its second-quarter results. That profit number will lead most coverage. It shouldn’t. Around $99 billion of it came from unrealized gains on Alphabet’s stakes in Anthropic and SpaceX — paper mark-ups that rise and fall with private-market valuations, not cash the business earned selling ads or cloud capacity. Operating income, the part that reflects Google actually operating, was $40.8 billion at a 34% margin. Once the investment accounting is set aside, the quarter’s defining figure is not the record profit. It is the $44.9 billion Alphabet spent on capital expenditure, and the roughly $5.9 billion cash hole that opened underneath it.

The number that actually moved

Free cash flow — the cash left after operating costs and capital spending — went negative for the first time on Alphabet’s record, at about -$5.9 billion for the quarter. This is a single-quarter event rather than a solvency problem: trailing-twelve-month free cash flow was still roughly $53 billion, and the company holds one of the strongest balance sheets in the market. What changed is the direction, and how fast it changed.

A year ago, in the comparable quarter, Google lifted its 2025 capital-spending plan to about $85 billion, and investors already treated that as an aggressive commitment to data centers and chips. For 2026, management just raised the range again — to $195–205 billion, up from a prior $180–190 billion. Capital spending that was a large line two years ago is now approaching a quarter-trillion-dollar annual habit, and it has finally grown faster than the cash the business throws off.

Why the record headline is a distraction

A 298% jump in net income reads like a blowout. The composition tells a quieter story. The bulk of the increase is a fair-value adjustment on Alphabet’s private holdings, booked because Anthropic and SpaceX carry higher valuations after recent funding activity. That is genuine accounting, but it is a mark-to-market gain on other companies, and it can reverse just as cleanly if those valuations cool.

The operating business grew, and grew well — but at a normal pace, not a triple-digit one. Google Services still funds the company, advertising remains the cash engine, and the $40.8 billion of operating income is the figure to anchor on when judging whether the AI spending is affordable. Reading the $112 billion as proof that AI is already paying for itself confuses an investment gain with an operating return.

The distinction is not academic. Anthropic and SpaceX are private, so their carrying values move in step whenever a new funding round reprices them, and the resulting swings land in Alphabet’s reported net income as non-cash gains or losses that have nothing to do with how many ads or cloud seats Google sold. A quarter like this one flatters earnings per share; a down round somewhere in the AI or space complex would do the reverse, and neither outcome would change the cash Alphabet is actually spending on servers. Investors who let the reported profit set their expectations are anchoring on the single most volatile line in the report.

Alphabet is spending against a backlog, not a hope

The reason management keeps raising the capex number sits in the cloud segment. Google Cloud revenue grew 82% to $24.8 billion, and its contracted backlog — revenue signed but not yet recognized — reached $514 billion. That backlog is why the spending is not speculative. Chief financial officer Anat Ashkenazi told analysts on the earnings call that “demand continues to outpace the rate of investment,” which is a polite way of saying Google is capacity-constrained and losing business it could otherwise book.

Put those two facts together and the negative free cash flow stops looking like a stumble. Alphabet is choosing to spend ahead of demand it has already contracted, because the cost of being short on compute is a customer moving to a competitor’s cloud. In that framing, the cash deficit is a deliberate purchase of market position, not a loss of control over costs.

The acceleration is what makes the choice defensible. Cloud growth of 82% is faster than the segment posted in the prior several quarters, and a backlog above half a trillion dollars means the revenue to fill the new capacity is already under contract rather than merely hoped for. A company adding capex into decelerating demand would be a warning; a company adding capex because it keeps running out of capacity to sell is a different story, even when the near-term cash math looks alarming.

The bill arrives twice

Capital spending hits the accounts in two stages, and only the first has landed. The $44.9 billion of cash out the door this quarter is the visible part. The second wave is depreciation: those data centers and chips are assets that will be written down over the years ahead, and at roughly $200 billion of annual additions the depreciation line eventually becomes a heavy, recurring charge against operating income — the same operating income that currently makes the spending look affordable.

That is the part of the AI-capex trade the market tends to underweight. A company can fund a cash-negative quarter easily when its balance sheet is this deep. Absorbing a depreciation schedule built on a quarter-trillion dollars a year, while advertising grows in the low-to-mid twenties, is a slower and more permanent squeeze. The question for 2027 and beyond is not whether Alphabet can pay for the build — it plainly can — but whether cloud and AI revenue ramps quickly enough to outrun the depreciation its own ambition is creating.

What a cash-negative Google means for the AI economy

When the most cash-generative company in technology crosses into negative free cash flow to fund AI capacity, it resets the floor for everyone downstream. The demand signal beneath Nvidia’s accelerators and Google’s own tensor chips just got a very large, very public confirmation: the buyer with the deepest pockets is willing to run a cash deficit rather than slow down.

That confidence is spreading through the same supply chain. Amazon is pacing its own multibillion-dollar Trainium commitments to lock in custom-silicon capacity, and the pattern across hyperscalers is to treat compute like a scarce input to be secured years in advance. Microsoft and Meta have each walked their capital budgets higher through 2026 for the same reason, which turns Alphabet’s cash-negative quarter into something less like a company-specific misstep and more like the leading edge of an industry-wide shift. When four of the largest technology firms are simultaneously spending down their cash cushions to buy AI capacity, the aggregate order book underneath chipmakers, memory suppliers, power utilities, and data-center builders stops looking cyclical and starts looking structural. It is also why the spending is increasingly financed rather than self-funded. Alphabet’s century-dated bond sale earlier this year was an early signal that even a balance sheet this strong reaches for outside capital once the number gets big enough — and a quarter of negative free cash flow makes that reach look less like opportunism and more like plumbing.

Someone still has to earn the return

Roughly $200 billion a year of new capacity has to be paid back, and search advertising alone will not do it. The return has to come from selling AI — cloud infrastructure, enterprise Gemini seats, and the model business itself. Alphabet gave investors the usage numbers to argue the demand is real: Gemini models now process about 22 billion API tokens per minute, the Gemini app has around 950 million monthly users, and close to 90% of the Fortune 100 are using Gemini Enterprise. Sundar Pichai said the company is moving to a near-monthly model-release cadence and described “the most ambitious Gemini 4 pre-training to date.”

Usage is not yet the same as pricing power, and that gap is the whole investment case. The capital is being spent now; the pricing that justifies it has to arrive later, through cloud contracts and per-seat AI revenue rather than ad impressions. That is the pressure a specialized inference economy is built to exploit, and it is the same AI bill that recently turned Tesla’s record quarter into a margin squeeze. The mechanism is identical across very different companies: AI shows up as spending long before it shows up as profit.

For now, Alphabet is asking investors to accept a cash deficit today as the price of owning AI capacity tomorrow. Shares slid about 5% after the report, which is the market extending that credit while it waits — willing to fund the build, but expecting the payoff to eventually appear in cloud revenue rather than in the fair value of a stake in Anthropic.