The cloud race has entered its cash-flow phase. In today’s AI news today brief, Amazon supplied the clearest proof yet that demand and economics can pull in opposite directions: AWS posted its fastest growth in nearly five years, while Amazon’s trailing free cash flow swung to a $42.4 billion outflow as infrastructure spending surged. That does not mean the AI buildout is failing. It means investors are beginning to ask a more useful question than “Who is spending most?”—namely, “Who can turn that spending into durable returns?”

That tension runs through the rest of the October 29 news day. Intel’s data-center business accelerated as PCs weakened. Canada put public money behind a sovereign-AI champion. OpenAI rewrote its behavioral north star for more capable systems. And Washington and Helsinki sketched a broad technology alliance whose ambitions now need projects, budgets, and deadlines. Here is what matters, without the victory-lap confetti.

AI news today: Amazon’s AWS growth meets the cash-flow bill

Amazon reported third-quarter net sales of $201.9 billion and operating income of $28.1 billion. The more consequential number was AWS: revenue rose 39.3% year over year to $46.0 billion, its fastest growth rate in almost five years, while AWS operating income reached $17.8 billion.

Then comes the bill. Trailing-12-month free cash flow was a $42.4 billion outflow, versus a $14.8 billion inflow a year earlier. Amazon attributed the reversal mainly to a $94.8 billion increase in purchases of property and equipment, “primarily reflecting investments in artificial intelligence.” Management says AWS growth should exceed 40% in both 2027 and 2028; that is a company forecast, not a booked result.

For builders, the demand signal is excellent. For leaders and investors, the lesson is less comfortable: cloud growth no longer settles the capital-efficiency argument by itself. This is the same pressure visible in Microsoft’s latest Azure numbers. Watch whether Amazon can keep utilization rising faster than depreciation, power, and data-center commitments.

Intel’s AI servers surge while the PC market retreats

Intel’s third-quarter release showed revenue up 21% to $16.5 billion. Its Data Center and AI unit grew 78% to $7.3 billion—the company called it its strongest data-center growth in 15 years—while the client-computing unit fell 5% to $8.1 billion. Intel expects the PC slowdown to continue through at least the first half of 2027 and guided fourth-quarter revenue to $15.0–$16.0 billion.

The split matters because AI infrastructure is rescuing parts of the old CPU business even as consumer demand softens. But execution risk has not vanished: Reuters reported that Intel shares fell 18% as investors focused on the outlook and supply constraints. Next comes proof that extraordinary Xeon demand can become sustained margin improvement, not merely a helpful quarter.

Canada takes an equity stake in Cohere’s sovereign-AI bet

The Canadian government committed to invest up to C$1.6 billion in Cohere as part of a financing round that also includes Nvidia, Bell, and Schwarz Group. According to the official announcement, the staged investment comes with commitments for at least 1,000 new Canadian technical jobs and C$1.7 billion in additional Canadian computing infrastructure by 2030, plus continued Canadian incorporation and headquarters.

This is industrial policy wearing a venture-capital jacket. Ottawa gets equity exposure and accountability milestones; Cohere gets capital to compete in an expensive frontier-model market. The public release does not disclose the government’s ownership percentage or all investment terms, so taxpayers cannot yet judge the price of the risk. Still, the strategic message is unambiguous: countries increasingly see model capacity, data control, and compute as infrastructure worth owning, not simply software worth buying.

OpenAI rewrites the Model Spec around five values

OpenAI updated its Model Spec, the public document describing how it wants its systems to behave. The revision organizes guidance around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. It also adds guidance on context, memory, personalization, oversight, and side effects while retiring much of the earlier detailed style advice.

The important caveat comes from OpenAI itself: the Spec describes target behavior, and current models do not yet fully reflect it. In other words, this is a design constitution, not a compliance certificate. Developers should watch for corresponding changes in evaluations, product controls, and observed behavior. The broader capital story matters here too: as frontier labs pursue more autonomy and more consequential actions—one reason AI companies keep exploring public-market capital—behavioral promises need measurable evidence.

The U.S. and Finland put AI, quantum, 6G, and chips in one pact

Washington and Helsinki signed a Technology Prosperity Deal covering AI and “super intelligence,” 5G and 6G, quantum technology, semiconductors, biotechnology, energy, space, and Arctic systems. Finland’s foreign ministry described the goal as faster research, commercialization, standards coordination, and trusted supply chains.

That breadth is both the attraction and the weakness. The memorandum records intentions; it is not evidence that factories, export financing, or joint research programs are already funded. For technology companies, however, standards work and public financing can shape markets long before products ship. The next useful update will be narrower: named projects, accountable agencies, money, and dates.

Watch & Learn

Video: ChatGPT Data Analysis for Beginners (Full Guide) by AI Foundations.

This roughly 14-minute walkthrough shows how to upload, clean, question, and visualize structured data in ChatGPT. It is useful for operators who want to test an earnings dataset without writing code. The interface shown is older, but the core workflow—inspect, define metrics, calculate, then verify—still maps cleanly to today’s product.

AI, Translated: inference economics

Inference economics is the cost-and-revenue math of running a trained AI model for real users. A chatbot answer consumes chips, memory, networking, electricity, and data-center capacity; shorter prompts, better batching, or a smaller model can reduce that cost. If a service earns ten cents per request but spends twelve cents delivering it, impressive usage growth still destroys value. You should care because the winners in AI may be decided as much by efficient delivery as by benchmark scores.

Try This Today: audit an AI earnings claim with ChatGPT

Goal: turn an earnings table into a small, checkable trend analysis in about ten minutes.

1. Copy two to four quarters of revenue, capex, cloud growth, and free cash flow into a CSV, with one row per quarter.
2. Upload it to ChatGPT and ask it to restate every metric definition before calculating.
3. Request a table plus one chart, then inspect the generated code or calculation details before using the result.

Copy-ready prompt: “Audit this dataset for missing values and inconsistent units. Then calculate quarter-over-quarter changes for each metric, chart cloud growth against free cash flow, and identify what the data supports versus what would still be speculation. Show your formulas and cite the source row for every number.”

OpenAI’s current data-analysis guide confirms support for uploaded spreadsheets, code-backed calculations, tables, and charts. Availability and file limits can vary by plan and workspace settings.

One thing to remember

AI demand is no longer the disputed point. The harder contest is who can finance, govern, and operate that demand without letting cost or control outrun the value created.