Second in the series. Google.
Google has been taking a lot of heat lately. Gemini 3.5 Pro slipped several months, coding performance didn't land, and new models from OpenAI and Meta pulled ahead on code generation — the stock dropped 4.4% the day that news broke in mid-July. Last week Jeff Dean announced he's leaving after 27 years. Sanjay Ghemawat, Oriol Vinyals and Quoc Le are going too. And this week the company priced $25B of debt across 10 tranches, 2 to 40 years.
Should you panic when the big names walk? I look at two things in a company: the business model and the culture. Neither has much to do with who's sitting in which chair. You want to be clear on that before you invest.
Back to the filing.
【Revenue mix】
Search $63.27B, +16.8%. That single line falsifies the "generative AI is killing search" story. It didn't shrink — it grew faster than total advertising (+14.4%). Management says AI features are driving query volume up. Gemini App is at 950M monthly actives.
Cloud $24.77B, +81.8%, operating income $8.81B. This is the one that matters, because it answers the actual question: is all that capex (capital expenditure — land, data centers, chips, the long-lived stuff) converting into profit acceleration.
That 82% needs a haircut. Wiz consolidated in March, bought for $32B, with ARR (annual recurring revenue — the subscription book annualized) just over $1B at close. Call it ~$250M for a quarter, against a $24.8B base, so a little over one percentage point. The rest is organic — grown in-house, not bought.
Here's the better number: cloud backlog of $514B, up more than $50B in a single quarter, above the $488.1B consensus. That's remaining performance obligations — contracts signed but not yet recognized as revenue. It's five times the current annualized run rate, and over half converts within 24 months. These are signed contracts.
On the same call the CFO said two things. Supply is still short — that's several quarters running now. And in Q3 they'll buy more third-party compute as a bridge, with an explicit warning that it compresses cloud margins. So that 35.6% probably comes down next quarter.
Google Network $7.30B, −0.7%, still negative. This is Google placing ads on third-party sites and splitting the revenue — not its own search or YouTube inventory. Q1'25 −2%, Q2'25 −1%, Q1'26 −4%. It hasn't stopped. It's the one line directly exposed to the shrinking open web.
Waymo and the rest don't move the financials yet. Skip them.
【Earnings quality】
There's a trick in this quarter's profit.
Net income $112.19B, of which $77.1B is securities gains on the Anthropic and SpaceX stakes — the holdings were marked up in value and the accounting runs it through income. No cash came in. The company discloses it in the filing, $6.26 per share.
Strip it out properly and Q2 normalized EPS (earnings per share) is $2.85 against $2.91 consensus.
Real earnings power came in below expectations.
【Margins】
Q2 operating margin 34.0%, versus 32.4% a year ago. TTM (trailing twelve months) 33.1%, a five-year high. Looks great.
Net PP&E (property, plant & equipment — the fixed asset base, mostly data centers and servers, after accumulated depreciation) went from $148.4B in 2023 to $321.2B. Depreciation over the same stretch went from $11.9B to $25.2B. The ratio barely moved, so on the surface it tracks.
The problem is this year. 2026 capex alone is $200B, equal to 62% of the current net PP&E base. Those assets just came online and have barely started depreciating. The pressure shows up over the next two to four years.
Today's 34% was earned before depreciation fully hit the income statement.
【Cash flow】
Quarterly free cash flow (operating cash minus capex — what's actually left for shareholders) went negative. Capex $44.9B against operating cash flow $39.1B. Spent more than it earned.
This is what the market cares about most, and where opinion is most split. Nobody agrees on when Google's free cash flow starts growing again. That's what the post-earnings selloff was about. Not the results.
【Capex and debt】
Full-year guidance went from $180–190B in April to $195–205B in July. Midpoint up $15B. That's restrained, and the company keeps saying compute is short. The demand is real.
But the money has to come from somewhere. Net debt issuance of $56.2B in the first half, $49.6B of equity raised in June, and this week another $25B of bonds with peak orders around $115B. The four hyperscalers have issued roughly $194B of debt through July, up 79% year over year.
【Buybacks】
Over $60B a year in 2023 and 2024. Zero in the first half of 2026.
I don't like this. As a shareholder you can't hand a company an unlimited spending budget. Spend, fine — but show me the return. Buybacks at zero, plus equity issuance, plus debt, means shareholders are fronting the money and the return date is unknown.
【Wrap】
The demand side isn't in question. Search held, and cloud has $514B of contracts sitting there.
Here's what I'm watching from here: how much of the profit is actually earned by operating the business, when depreciation catches up, which quarter free cash flow turns positive again, and how far third-party compute pushes cloud margins down.
Gemini's delay and Jeff Dean leaving don't concern me much. What I watch is return on capital.
Holdings and historical data 👉 https://t.co/co0JG43NK0
Microsoft next.
Sources: Alphabet Q2 2026 results and earnings call (07-22), Jeff Dean departure (08-05), bond issuance (08-06). Not investment advice.
$GOOGL