🚨 Big Tech to spend $650B on AI in 2026

The AI race just hit a new level, with Big Tech planning an unprecedented $650B in spending, shifting the battle from software features to raw infrastructure, energy, and compute power. It shows how fast AI is moving from hype into heavy real world buildout. Let’s dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: Four US tech giants plan to spend a combined $650B this year as the AI race turns into a full scale infrastructure battle.

Most of this spending goes into massive AI data centers packed with advanced processors from companies like NVIDIA. Building these sites means land, power, cooling systems, and skilled labor at a scale that is starting to strain supply chains and local energy grids. The strategy behind it is that generative AI systems, including tools like OpenAI’s ChatGPT, will become core digital infrastructure. Training and running these models requires thousands of high end chips that cost tens of thousands of dollars each. Owning the compute is seen as the path to long term dominance.

The $650B number is big, but the bigger shift is what it forces Big Tech to become: infrastructure companies, not just software companies. This spend is turning AI into a race for power, permits, chips, and construction crews, and those are already the choke points in both the US and Europe. At the same time, the IEA (International Energy Agency) expects data center electricity use to keep climbing fast through 2030, which is why grids and energy prices are suddenly part of the AI conversation. And if these capex plans keep outrunning cash flow, you will see more debt, more investor pushback, and a very real “prove it” moment where AI has to show revenue, not just demos.

Evolving AI: The question isn’t if you’ll deploy voice AI. It’s how.

This handbook from ElevenLabs explores the modern approaches to building scalable, humanlike voice agents from full in-house builds to hybrid orchestration frameworks.

Learn from real-world enterprise examples and find the model that fits your business and technical goals.

Free Download →

Evolving AI: Anthropic’s latest Claude upgrade shook markets as investors questioned how safe traditional enterprise software is in an AI-first world.

The market reaction started earlier this week when Anthropic rolled out industry focused plug ins for its Claude Cowork product. What was framed as a product update quickly turned into a market event. Investors began pricing in the risk that AI systems could take over knowledge work tasks that many enterprise tools are built around. Financial data providers were hit hardest. Shares of FactSet fell around 10%, while S&P Global, Moody’s, and Nasdaq also declined. Investors focused on AI’s growing ability to handle research, screening, due diligence support, and market analysis. These are core activities across financial and enterprise workflows. The pressure increased as Anthropic expanded further into business use cases beyond coding, while OpenAI also pushed deeper into productivity tasks like documentation and presentations. Together, this signaled to markets that general purpose AI is moving directly into areas long served by specialized software.

Wild part is a Claude update helped spark a roughly $1T wipeout in software and data stocks in days. Markets are now pricing AI agents as a revenue threat to SaaS and research businesses, not a nice add on. Who’s next?

Evolving AI: Waymo is using tech from Google DeepMind to simulate rare road events instead of waiting years to capture them in real life.

Self driving systems used to learn mostly from real miles. Waymo has logged over 200 million on real roads, plus billions virtually. But rare events like unusual snow, strange road behavior, or edge cases barely show up in normal data. The Waymo World Model fills those gaps. Engineers can take real driving footage and change weather, time of day, road layout, or traffic, and add unexpected objects or vehicles. Genie 3 is not a game engine but a world model trained on video that learns how reality behaves and predicts what should happen next frame by frame. That means Waymo can create situations that never happened but still follow realistic motion and cause and effect. The model keeps scenes consistent for minutes at a time, which is a big step for world models.

Waymo is basically saying “we’re done waiting for rare stuff to happen in the real world.” With its new Waymo World Model built on DeepMind’s Genie 3, they can remix real drives into snow, weird road layouts, and edge cases, while also generating matching lidar and camera data so the car learns the full scene. That’s a big shift as Waymo pushes into tougher markets and more cities, where weather and road chaos are the norm and regulation can be the bigger blocker than tech.

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

QUICK HITS

🤖 ChatGPT’s new internet browser can run 80% of a 1-person business — no tech skills required.

🥋 Robots practice kung fu with monks at the Shaolin Temple in China.

🎬 Kuaishou Technology has launched Kling AI 3.0, a new version of its AI video and image generation tool.

📱 Google’s Gemini app has surpassed 750M monthly active users.

🏥 Fangzhou wins Tencent Health AI-powered Future Award as MaaS platform drives AI chronic care.

⚖️ US companies accused of AI washing in citing artificial intelligence for job losses.

🧬 ByteDance releases Protenix-v1, a new open-source model achieving AF3-Level performance in biomolecular structure prediction.

 *partner link