The ClickedOn AI Wrap: Week of 1 Nov 2026

October 2026 will be remembered as the month investors stopped cheering AI ambition and started demanding proof of returns. A brutal end-of-month selloff erased roughly $2 trillion from the Nasdaq-100 as earnings reports from the four biggest AI spenders landed in quick succession. The Nasdaq itself finished about 10% off its early-October peak.

The trigger was not a single miss but a cumulative realisation. Amazon, Alphabet, Microsoft and Meta together reported roughly $200 billion in property and equipment spending for the quarter, and investors questioned whether AI monetisation can keep pace. Meta posted 26% revenue growth to $64.6b but operating profit collapsed 37% year-on-year as the company's AI capital expenditure plan reached up to $145b for fiscal 2026. That made it the company's first negative free-cash-flow quarter since 2012. Alphabet's revenue rose 25% to $127.6b, though Search advertising growth decelerated to 12% year-on-year, down from 17% in Q2. Microsoft was the cleanest story, reporting $92.3b in revenue up 19% and profit of $36.6b up 32%, but it also signalled heavy ongoing capital commitments.

Beyond earnings, OpenAI cut its fundraising target by $10b, moving from a $30b round at a $1.4 trillion valuation to $20b at $1.25 trillion, according to the Financial Times. The reduction reflected both the market rout and competition from Anthropic, which filed its S-1 on 26 October targeting a mid-November Nasdaq debut at a valuation of at least $1.5 trillion. Anthropic reported $16.3b in first-half revenue alongside an $80.7b net loss, most of which came from noncash accounting charges. Internal documents cited by analysts suggest OpenAI faces a cumulative free cash flow gap of roughly $278b between 2026 and 2030.

One nuance worth watching: AI-services firms held up well while hardware and infrastructure names fell hardest. Accenture gained nearly 19% over two weeks while chip-adjacent names like CoreWeave and SanDisk dropped sharply. The market is not saying AI is over. It is saying the easy money phase of infrastructure betting is over, and the businesses that monetise AI output are now more interesting than those that build the pipes.

For Australian businesses watching their own AI budgets, the message is the same one ClickedOn has been saying all year: the question is not how much you spend on AI, it is what you can measure from it.

GitHub's Octoverse 2026 report, released on 28 October at the company's Universe conference, put a concrete number on the agentic coding shift: coding agents started more than 40% of pull requests in September 2026, up from just 17% in March. That translated to approximately 110 million agent-started PRs in a single month. Total commits across the platform hit 34.3 billion for the year, nearly triple the prior year's 12.9 billion. GitHub describes the agent-started share as a floor because it only counts agents it can detect.

The breakdown by language is telling for anyone building on the web: AI involvement in merged pull requests ran at 44.1% for TypeScript, 41.1% for Python and 40.4% for Rust. The data also shows 90% of agent-started PRs merged within two weeks, slightly ahead of the 86% rate for human-only PRs. Code review capacity is now the main constraint, and GitHub's own AI-assisted review and Advanced Security products grew more than 90% to address it.

Alongside the Octoverse data, GitHub announced Copilot Fusion as generally available. Rather than asking developers to pick a model for each task, Fusion automatically selects and coordinates across models including Anthropic's Claude Opus 5.5 and Sonnet 5.5 and OpenAI's GPT-6 Astra and GPT-6 Sol. The product offers Standard, High and Max effort tiers. Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code running Opus 5.5 Max, at roughly half the average API cost per task and using about half as many tokens. A 30% discount on model costs applies through 30 November 2026. Fusion is live across VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs.

On 28 October, Google confirmed it is expanding ads inside AI Mode to ten international markets, with Australia included in the first wave alongside Canada, India, Kenya, Malaysia, New Zealand, Nigeria, Pakistan, the Philippines and Singapore. A second wave covering Brazil, Mexico and the UK is planned for later in Q4. At the same time, Google launched multiturn ads in the United States, a format that lets sponsored content update as a conversation progresses through follow-up queries.

The announcement came during Alphabet's Q3 earnings call, where the company reported $127.6b in total revenue but disclosed that Search advertising growth had decelerated to 12% year-on-year, down from 19% in Q1. That context matters: Google is pushing AI Mode ads internationally partly to arrest the monetisation gap that comes with AI-generated answers serving queries that previously drove traditional clicks. Google's own research shows queries in AI Mode are, on average, three times longer than traditional searches, which is both an opportunity and a challenge for attribution.

For advertisers, the practical implication is immediate: existing Search, Shopping and Performance Max campaigns are automatically eligible to serve inside AI Overviews and AI Mode without setting up separate budgets or campaigns. No new campaign type is required, though Google has not disclosed click-through or conversion benchmarks for the Australian market yet. Ads will be clearly labelled as Sponsored. Gemini also helped advertisers create over 1.5 billion campaign assets in Q3, up more than 7x year-on-year. Publishers, on the other hand, are watching closely. European publishers have already told the European Commission that AI Mode sends fewer than 5% of queries to external websites, a sign of the traffic diversion that comes with AI-answered search.

If you run Google Ads in Australia, check your AI Mode impression data now. The surface is live, and the early data on which query types convert will be valuable. Our team at ClickedOn is already monitoring this across client accounts.