Big Tech Commits Over $300 Billion to AI Capex for 2025

Pivot 5
May 05, 2025

Pivot 5 is the must-read daily AI briefing for 500,000+ CEOs and business leaders who need signal, not noise.

1. Big Tech Commits Over $300 Billion to AI Capex for 2025

Alphabet, Microsoft, Meta and Amazon told investors they will pour more than $300 billion into AI-focused infrastructure this year. The pledges surfaced in their Q1 2025 earnings calls and show no appetite to slow spending despite tariff uncertainty.
Microsoft and Alphabet reaffirmed their capex forecasts, Meta lifted its 2025 target to as much as $72 billion, and Amazon has already spent $24.3 billion in Q1 on the way to roughly $100 billion for the year. Investors now assign outsized weight to hyperscaler cash flow as the fuel that powers every other AI player.
Such sums effectively gate access to advanced compute behind a handful of cash-rich giants. AI competition is shifting from algorithmic breakthroughs to whoever can keep the power and steel flowing.

Read more here

Anthropic unveiled “Integrations” that allow Claude to draft PayPal invoices and pull live sales numbers from Square directly in the chat. The launch also introduced an Advanced Research mode that breaks complex asks into multi-step workflows.
Both features sit on the company’s open Model Context Protocol, which now links Claude to Atlassian, Zapier, Cloudflare and Intercom alongside the payment apps. Anthropic says the toolkit will soon connect to “hundreds” of data sources.
By embedding payments and operational data, Claude graduates from conversational aide to lightweight back-office hub. The player that owns that real-time transaction stream will shape the next training loop in enterprise AI.

Read more here

Amazon CEO Andy Jassy revealed that Alexa+, the company’s generative-AI assistant, now serves more than 100,000 users barely three months after its debut. The service is expanding in waves across a base of roughly 600 million Alexa devices worldwide.
Jassy called Alexa+ one of the first action-oriented AI agents for consumers but admitted current multi-step accuracy sits between 30% and 60%. He set a public goal of 90% accuracy and acknowledged that demoed capabilities like third-party app control are still absent.
The candor underscores how far agentic AI has to travel from lab demo to living-room reality. Amazon is effectively beta-testing the category at scale before Apple and Google can reboot their own voice assistants.

Read more here

BBC Maestro will stream a crime-writing course “taught” by Agatha Christie, reconstructed with AI from archival audio, letters and licensed images. Produced with BBC Studios and the Christie estate, the videos use actor Vivien Keene and visual-effects artists to recreate the author’s voice and likeness.
Scholars curated Christie’s own advice on story structure, plot twists and suspense for the subscription-based lessons, which her family has fully endorsed. The team says the project pieced together her words from interviews and correspondence, even as some authors question the ethics of AI revivals.
Animating a century-old literary brand as an interactive instructor blurs the line between archive and new production. If the format succeeds, dormant IP catalogs could turn into perpetual content factories.

Read more here

Author Katie Parrott describes building a workflow app with Cursor by simply chatting commands, echoing Andrej Karpathy’s “vibe coding” credo of forgetting the code and following the vibes. AI translated her plain-language prompt into working software during a Saturday side project.
Parrott points to marketers, operators and hobbyists who have spun up content generators, meal planners and board-game advisors, drawing parallels to the creative booms of blogging and smartphone photography. Fast feedback loops and natural-language interfaces are lowering the psychological and technical barriers that once fenced off software creation.
When domain experts can conjure custom apps without calling engineering, new products start emerging from unexpected corners. Scarcity of coding talent stops being the choke point once context and creativity drive what gets built.

Read the opinion essay from Kate Parrot here