Google’s Search Growth Fell to 12% While Its Cloud Doubled: the 4 Engineering Bets I Changed That Night

I do not normally read quarterly results from US companies with a notebook open. I made an exception on Wednesday night, because one line in Alphabet’s third quarter tells you something about the next two years of web engineering that no conference talk has said out loud. Search and other revenue grew 12 per cent. One quarter earlier it grew 17 per cent. In the same three months, the Cloud segment more than doubled. The company that owns the channel most Singapore businesses still depend on is quietly telling you where it expects its own growth to come from, and it is not the channel you optimise for.

What Was Actually Reported, in Its Parts

Alphabet published its third quarter results on 28 October 2026. Revenue was US$127.58 billion, up 25 per cent from US$102.35 billion a year earlier. That is the number most coverage led with, and it is genuinely strong. The segment detail is where the story is. According to the breakdown of the release, Google Cloud revenue reached US$30.71 billion, up 103 per cent year on year, having grown 82 per cent in the second quarter. Search and other revenue reached US$63.35 billion, up 12 per cent, having grown 17 per cent in the second quarter.

Hold those two trajectories next to each other. Cloud growth accelerated by 21 percentage points in a quarter. Search growth decelerated by 5. Search is still more than twice the size of Cloud in absolute revenue, so this is not a story about one replacing the other. It is a story about which line the company’s capital is now chasing, and the capital numbers say the same thing. Capital expenditure was US$51.36 billion in the quarter and US$132.0 billion across the first nine months. Free cash flow was negative US$9.07 billion, against positive US$24.46 billion in the same quarter last year. Alphabet’s finance chief told the earnings call that capital expenditure is expected to increase significantly in 2027.

One correction before anyone forwards you a panicked summary. Diluted earnings per share were US$2.12, down 26 per cent, and net income for common shareholders was US$26.17 billion, down 25 per cent from US$34.98 billion. That decline was not the operating business weakening. Alphabet attributed it to an US$8.7 billion loss on equity securities, the stakes it holds in other companies, against a US$10.7 billion gain a year earlier, and stated the effect cut earnings per share by US$0.55. If you saw a different earnings per share figure that week, it is because several outlets, CNBC among them, led with an adjusted measure rather than the reported diluted one. Class A shares traded around 9.4 per cent below the previous close in after-hours dealing.

One figure I am deliberately not using: the cloud backlog. A number has been circulating this week that I cannot tie to the release. What is verifiable is the US$514 billion figure given for the end of June on the second quarter call, with the third quarter figure due in the quarterly filing. When a number is doing a lot of work in an argument, it is worth knowing whether it has been published yet.

Bet 1: I Stopped Treating Organic Traffic as One Number

The first change cost nothing and was the most uncomfortable. We split the single organic traffic line in our reporting into three: visits that arrive with a query we can see, visits that arrive from a search surface with no query attached, and visits with no referrer at all. Most teams report the sum of those three and call it organic. The sum is the one number guaranteed to hide a change in mix.

This matters because the mix is what moves first. If answers are increasingly assembled before the click, the volume of sessions can hold steady for a quarter or two while their composition changes underneath, and the first visible symptom is not a traffic fall. It is a conversion rate that drifts because the visitors who still click are at a different stage of their decision. A team reporting one organic number experiences that as a mysterious conversion problem and sends a designer to fix it.

If you are building the measurement side of this properly, it is a data engineering job rather than an analytics configuration job, which is why I would put it next to the rest of your pipeline work rather than inside the marketing stack. The practical skills overlap heavily with what you would look for when you hire a data engineer in Singapore, and very little with what a growth generalist brings.

Bet 2: The Requisition Changed From Front End to Content Systems

The second bet is a hiring one. We had an open role written as a front end engineer with some search engine optimisation exposure. I rewrote it as a content systems engineer, and the shortlist changed completely.

The distinction is not cosmetic. A front end engineer with optimisation exposure optimises pages. A content systems engineer owns the machinery that decides what a machine reading your site at scale can reliably extract from it: whether structured data is generated from the same source of truth as the visible page or maintained separately and therefore wrong within a month, whether rendering is consistent for a client that does not execute your JavaScript, whether your feeds and sitemaps reflect what is actually published, and whether a thousand pages stay internally consistent when the template changes.

That work has always mattered and has usually been nobody’s job. When acquisition came predominantly from ranked lists of links, you could survive doing it badly, because a human would click through and forgive the inconsistency. When a growing share of acquisition is mediated by something that reads your site once and summarises it, inconsistency is not forgiven, it is propagated. The useful interview question is not about tags. It is to ask a candidate how they would detect that structured data had silently diverged from the page content across two thousand URLs, and to listen for whether the answer involves a monitor or a spreadsheet.

Bet 3: We Costed the Acquisition Channel Like Infrastructure

The third change came straight out of the capital expenditure line. If the companies that deliver your audience are funding that delivery with an amount of capital that puts their own free cash flow below zero, the cost of reaching an audience through them is unlikely to fall. That is not a prediction about advertising rates. It is an observation about who is paying for the capacity, and capacity that is being financed gets priced to repay.

So we moved acquisition out of a marketing budget line and into the same monthly review as our cloud and model spend, with the same question asked of all three: what did we spend, on what, and what is the cost per unit of the thing the business counts. For a business whose economics depend on a predictable cost of acquisition, that one reporting change does more than a campaign. If you want to sanity check the people side of the same model, the team cost calculator is a reasonable starting point.

The hiring consequence is specific. A requisition that says the hire will improve the site competes with every other project in the company. A requisition that names a cost per qualified lead, states where that number currently sits and what the hire is expected to do to it, is arguing in the language your finance team already uses. The same logic is why an ecommerce platform build in Singapore should have its acquisition instrumentation specified in the first sprint rather than the last, which is where it usually ends up.

Bet 4: I Stopped Hiring Against the Channel as It Was

The fourth bet is the one I would keep if I could only keep one, and it is a judgement rather than a process. I stopped making multi-year hiring commitments that assume the shape of the acquisition channel stays fixed, and started preferring people who have migrated a measurement system at least once.

The reason sits in the two growth rates. A line growing at 12 per cent, decelerating, while its parent company redirects historic amounts of capital elsewhere, is a line in transition. Nobody outside Alphabet knows what it looks like in three years, and the honest position is that the people inside are hedging too, which is what US$51 billion in a quarter represents. In that situation, hiring deep specialists in the current surface is a bet that the surface persists. Hiring people who are good at instrumenting and re-instrumenting is a bet that it changes. The second bet is cheaper to be wrong about.

Concretely, that moved two things in our screening. We now ask every candidate for acquisition-adjacent roles to describe a time the definition of a metric they owned had to change, and what they did about the historical data. And we weight people who have run a migration where the old and new numbers disagreed and both were defensible, because that is the exact situation a changing search surface creates, and the wrong instinct in that moment, picking the flattering number, is very hard to train out.

What This News Is Not

It is not the death of search as an acquisition channel. A line that grew 12 per cent to US$63.35 billion in a quarter is not dying, and people who spent this week saying otherwise were mostly saying it before the results came out. Deceleration from a high rate is the most common thing that happens to large businesses and it usually continues for years without anything dramatic occurring.

It is not a reason to pause hiring either. Alphabet is spending more, not less, and the engineering work downstream of that spending, integration, measurement, data plumbing, cost control, grows with it. What changes is which requisitions are easy to defend. A generic optimisation role gets harder to justify through 2027. A role attached to a number that appears in a board pack gets easier.

And it is not an argument that earnings per share fell because the business is weak. It fell because of an US$8.7 billion mark on equity holdings, an effect Alphabet itself quantified at US$0.55 per share. If you take one transferable habit from this quarter, let it be the discipline of asking which number moved and why before building a plan on it. That habit is also the one I now screen for.

FAQ: Alphabet’s Third Quarter and Singapore Web Engineering