Google struggling to make AI Mode searches pay

Alphabet’s latest earnings call was a wake‑up call for anyone who thinks AI‑powered search is a cash‑cow. The company’s shares slumped almost ten percent after chief business officer Philipp Schindler admitted that the new “AI Mode” is actually dragging down monetisation. That’s a big deal when you consider Google is still pouring more than $51 billion a quarter into servers, networking gear and new datacentres, yet the AI‑driven front‑end is still earning less per query than the legacy ad‑stack that has powered the bulk of its cash flow for the past two decades.

AI Mode’s growth is real – but the money isn’t

Alphabet reported that year‑over‑year search query growth is at its highest level in five years, a surge largely credited to users migrating to AI Mode. Search revenue rose 12 percent year‑over‑year to $63.25 billion, a respectable figure but noticeably below the 14.5 percent growth seen in the same quarter last year. Schindler bluntly linked the slowdown to “lower monetisation we were seeing in AI Mode.” In other words, more eyes on the new interface aren’t translating into more dollars.

The gap matters because AI Mode currently sits on the fringe of Google’s revenue engine. The traditional search business still delivers just under half of Alphabet’s total quarterly revenue of $125.8 billion, while AI Mode is still being tested for ad insertion. The company has only just begun to sprinkle ads alongside AI Mode results globally after a US‑only trial, signalling that the monetisation model is still very much in beta.

The infrastructure bill is massive – and it’s eating profit

Google’s capital outlay is staggering. In Q3 alone, $51.4 billion left the company’s coffers, with roughly two‑thirds earmarked for servers and networking equipment and the remainder for new datacentre builds. That level of spend is a double‑edged sword: it underpins the massive scale needed for AI workloads, but it also drags down the bottom line. Net income fell to $26.17 billion this quarter, down from $35 billion a year ago when revenue was $102.35 billion.

For independent hosting providers, this is a cautionary tale. The hyperscalers’ appetite for ever‑larger silicon and endless rack space forces them to keep expanding, and the cost structure they reveal shows why marginal pricing on AI services can quickly become unsustainable. If you’re running your own servers, you can’t afford to chase the same scale without a clear path to revenue.

Google Cloud’s capacity crunch – a warning for the rest of us

Even Google Cloud, the second‑largest revenue generator at $24.75 billion for the quarter, is feeling the squeeze. The company admitted it does not have enough capacity to serve all its customers and is offloading workloads to third‑party providers. That move will “dent its future margins,” according to Thomas Kurian, head of Google Cloud.

The fact that a behemoth like Google needs to rent compute from others is a stark reminder that the AI hype train is not a free ride. When you factor in the cost of specialised hardware like TPUs, the economics become even tighter. Kurian noted that over 100 companies are already using Google’s TPU silicon under committed contracts, which will generate some revenue to offset capex, but the underlying capacity issue remains.

TPU sales as a stop‑gap, not a solution

Google’s strategy to monetise its custom silicon by selling TPU capacity to third parties is a classic “sell‑what‑you‑don’t‑need” play. While it will bring in cash, it’s a stop‑gap rather than a long‑term fix for the capacity shortfall. The company expects to book revenue from TPU sales in coming quarters, but the bulk of the spend on those chips still sits on the balance sheet as capital expense.

For smaller providers, the lesson is clear: building your own specialised AI hardware is a massive gamble unless you have deep pockets and a guaranteed pipeline of high‑margin contracts. Most of us are better off leveraging existing cloud offerings while keeping a close eye on the economics of those services.

The Gemini app – a user‑growth story, not a profit story

Sundar Pichai highlighted the Gemini app’s rapid adoption – a billion users faster than any previous Google product. That’s an impressive headline, but user count alone does not equal revenue. The app’s success is a testament to the appetite for AI‑driven experiences, yet the monetisation model is still in flux. Pichai’s optimism about “agentic AI tools” aimed at business customers reflects a shift toward higher‑value contracts, but the timeline and pricing remain vague.

From a founder’s perspective, the Gemini rollout shows the power of network effects: once you have a massive user base, you can experiment with premium features, data licensing, or enterprise tiers. However, the path from user acquisition to sustainable cash flow is fraught with regulatory, security and governance hurdles – exactly the concerns Pichai flagged for business customers.

Capital‑expenditure outlook – more spending ahead

Alphabet stuck to its $195‑$205 billion capital‑expenditure forecast for the current year and signalled an intention to spend “significantly” more in 2027. That commitment underscores the scale of the infrastructure race and the pressure to keep the AI pipelines humming. For anyone watching the market, it signals that the hyperscaler model will continue to demand ever‑larger capex, which inevitably squeezes margins unless new revenue streams can offset the outlay.

Independent hosting firms must ask: can we survive in a world where the giants are pouring trillions into silicon and still struggling to monetise AI queries? The answer lies in focusing on niche workloads, offering transparent pricing, and avoiding the “best practice” trap of over‑engineering for speculative AI workloads.

Actionable takeaways for founders and providers

First, don’t chase AI Mode ad revenue until you have a proven, high‑margin use case. Google’s own struggle shows that “upper‑funnel” queries may look promising but still lag in cash generation. Second, keep your infrastructure spend disciplined. The $51 billion quarterly outlay is a reminder that hardware costs can quickly outpace revenue if you’re not careful.

Third, consider hybrid models: use public cloud for burst AI workloads but keep core services on‑premises to control costs. Fourth, watch the TPU market closely – while selling custom silicon can bring in cash, it’s not a panacea for capacity constraints. Finally, focus on building strong, contract‑backed enterprise relationships. As Kurian noted, committed contracts are what turn inventory into revenue.

In short, Google’s earnings call is a reality check for anyone banking on the AI hype. The infrastructure bill is real, the monetisation gap is widening, and the only way to stay afloat is to be brutally pragmatic about where you spend and how you price. Keep your servers humming, your margins tight, and your customers happy, and you’ll survive the hyperscaler tidal wave.

— Allan Ali, Founder

This article was produced with AI-assisted research and editorial support. Reporting is based on the source material cited below. Sources: The Register; theregister.com; Global1.News (31 October 2026).

By Allan Ali, Global1.News