The AI industry's late-October numbers tell two stories at once: adoption is accelerating inside software development at a startling pace, while the financial plumbing behind AI infrastructure is starting to groan under its own weight.

Five companies, $3 trillion, and vanishing buybacks

The starkest data point of the week comes from corporate filings rather than a product launch. Alphabet alone has more than $1 trillion in spending lined up, as the race for AI data centres grows, with Alphabet, Meta, Microsoft, Amazon and Oracle committing over $3 trillion to data centres, leases and energy. That spending has consequences: Microsoft, Alphabet, Amazon, Meta and Oracle poured $556.9 billion into capital spending in the first nine months of 2026, and share buybacks have nearly disappeared as a result. Alphabet's situation is especially notable — capital spending reached $51.4 billion for the quarter against $24.0 billion a year ago, and free cash flow came in at negative $9.1 billion against positive $24.5 billion. Microsoft is the lone holdout among the group still funding its buildout from operations, though even there it recorded $18.6 billion in positive free cash flow in its latest quarter, with capex at $45.2 billion, less than the $50 billion it had projected, which it blamed on delays in capacity from third-party providers rather than an intentional pullback.

The financing behind all this is increasingly borrowed rather than earned. Microsoft, Alphabet, Amazon, Meta, and Oracle raised a combined $326 billion in debt in 2026, triple the amount raised last year, according to S&P Global Market Intelligence.

SpaceX joins the GPU arms race — and warns rivals won't survive it

That debt anxiety got a vivid illustration from an unexpected corner: Elon Musk's rocket company. In a regulatory filing, SpaceX said it signed a binding commitment with an Apollo-managed entity for roughly $40 billion in GPU financing, chips that are expected to arrive in the first half of 2027. Musk framed the deal as proof of SpaceX's financial strength compared with pure-play AI cloud providers, warning that debt markets had all but shut for neoclouds depending solely on revenue streams from AI data centers, predicting the 200-plus neoclouds are going to see guppy-level death and consolidation, with only some making it through. Analysts noted the price of that confidence: SpaceX is borrowing $40 billion to buy AI chips at a reported 7.75% to 8.3%, more than a point above where its existing bonds trade.

Coding agents cross a tipping point

Away from the balance sheets, GitHub's annual Octoverse report shows just how normalized autonomous coding has become. As of September, developers start more than 40% of pull requests with an agent, up from 17% in March, amounting to 110 million agent-started pull requests in September 2026, growth that was sixfold in six months. Agent-authored work also looks different in practice: agent-started pull requests also more often change multiple files — about 87%, compared with 79% of pull requests without detected agent involvement, and they merge fast, with 90% of agent-started pull requests merging within two weeks in September, compared with about 86% of pull requests without detected agent involvement.

The throughline

Together, these stories capture an industry simultaneously proving its utility and testing the limits of its financing. Agents are already doing a large share of the world's coding work, even as the hardware underneath that work is increasingly bought on credit — a tension likely to define the AI narrative through the rest of 2026.

Sources

AI InfrastructureArtificial IntelligenceBig TechDaily AI UpdateGitHub CopilotSpaceX