Amazon is using this week's AWS re:Invent conference to assert itself in an AI race that suddenly looks more competitive.

Why it matters: Last week's news cycle was dominated by Google staking a claim that it has pulled ahead of OpenAI.

• Amazon now wants to signal that it belongs in that same tier, with its own models and chips and the world's largest cloud.

The big picture: Garman tells Axios that AWS is increasingly the cloud where customers are putting real production workloads due to its combination of capabilities and cost-effectiveness.

• "A year ago, there were questions about whether we'd missed the wave, but now most people are building their production systems in AWS because of what we've built over the past couple of years," Garman told Axios. "People are now realizing that Amazon has a great platform for AI."
• Garman's comments come as the company opens its Las Vegas conference, where it's expected to unveil new AI models and infrastructure.

What they're saying: The industry itself is at an inflection point, Garman said, moving from summarization and content creation to transforming broader workflows by taking on repetitive tasks.

• "It's not slowing down anytime soon. I think there was fear a year ago that maybe the model capabilities were plateauing," Garman said. "I think that is not the case anymore."

Between the lines: AWS is touting a trio of strengths to convince customers — and Wall Street — that it's at the AI frontier:

• Amazon hosts Anthropic, Meta, Mistral, Cohere, plus Amazon's own models — giving enterprises an array of choices that rivals sometimes lack.
• Trainium and Inferentia — Amazon's custom chips — are designed to help AWS compete on cost.
• Garman also pointed to AWS's deep integration with enterprise systems, security policies and compliance requirements.

Yes, but: AWS is often still missing from the conversations around the latest and greatest AI.

• Microsoft remains the default AI cloud for many CIOs because of its OpenAI partnership and early Copilot momentum.
• Although Amazon has been beefing up its internal models, it lacks a flagship frontier model directly comparable to GPT-5 or Gemini 3 Pro.
• The success of Trainium and other Amazon-designed chips depends on convincing customers to switch from Nvidia.

By the numbers: While AWS remains the leading name in the broader cloud computing race, its rivals are growing faster.

• Last quarter AWS saw its business grow 20%. Compare that with 34% for Google Cloud and 40% for Microsoft's Azure.

The bottom line: AWS dominates cloud, but is still working to prove its position at the AI frontier.

Apple AI chief John Giannandrea is retiring, the company said yesterday. Amar Subramanya, a prominent AI researcher and former employee of Microsoft and Google, will join the company to lead its work in the field.

Why it matters: Apple outlined a bold strategy for Apple Intelligence but has struggled to deliver on key components, including modernizing Siri.

Driving the news: Apple said Giannandrea is stepping down from his role as senior VP of machine learning and AI strategy and will serve as an adviser before formally retiring next spring.

• At the same time, Apple announced the hiring of Subramanya, a former Google researcher who joined Microsoft only four months ago, according to his LinkedIn profile.
• Subramanya will lead Apple's research, model development and safety work and report to Craig Federighi, while the rest of Giannandrea's organization will report to two other executives — Sabih Khan and Eddy Cue.

What they're saying: Apple CEO Tim Cook thanked Giannandrea for his role in a statement.

• "AI has long been central to Apple's strategy," Cook said.

AI copilots are reshaping how bankers source, analyze and act on opportunities.

• The result is sharper execution and faster decision-making.

Inside the room: AI is reducing tedious work that consumes junior analysts' time, primarily across deal origination, diligence and research.

• It can automatically update buyer lists and personalize client outreach, summarize dense investment memos, and mark up both NDAs and term sheets in minutes.
• Bankers use copilots to query models or past presentations, eliminating hours of searching through archived reports.

Between the lines: Each task might save only a few hours a week, but across hundreds of deals, those gains compound into real returns.

By the numbers: Evident, which monitors AI adoption at the world's largest banks, says ROI is finally starting to show up in its most recent AI Index.

• Of the 50 banks it tracks, 32 now disclose AI use cases that deliver a financial business impact.
• In some cases, the savings are in the low hundreds of millions of dollars. For one bank, it's around $2 billion.

• The OECD warned that the AI bubble bursting is a "key downside risk" to a U.S. economy where growth is already slowing. (Axios)
• OpenAI will take an ownership stake in Thrive Holdings, a unit of Josh Kushner's Thrive Capital, as part of a broader deal that will give the firm greater access to OpenAI models. (NYT)
• OpenAI also announced an expanded deal with Accenture to give thousands of the consulting giant's employees access to the enterprise version of ChatGPT. (Business Insider)
• Exclusive: Claude's "Skills" tool can be weaponized to execute ransomware, Cato Networks found. (Axios)
• Runway released an updated version of its AI video model. (CNBC)

Want to run a newspaper without torching millions? News Tower is a simulation game that offers that chance.