Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• Google Earth AI

• Why AI data centers are measured in gigawatts

• Claude adds new memory features

• OpenAI acquires AI startup

Let’s get right into it!

Sign Up | Advertise | Tools

BIG TECH

Google Earth AI

Google’s new Earth AI, launched in July, merges Google Earth with Gemini AI to transform massive amounts of satellite data into actionable environmental insights. At its core is AlphaEarth Foundations, an AI model that converts terabytes of Earth imagery into detailed historical data layers that track climate and ecosystem changes—such as rising sea levels, surface temperature trends, and pollution shifts.

Now, Google has added chatbot-style querying, allowing users to ask natural-language questions like “find algae blooms” to analyze data across weather, population, and imagery layers. This Gemini-powered geospatial reasoning system can identify at-risk communities and forecast disaster impacts, like predicting which areas a hurricane will hit hardest. The new features are available through Google Earth Professional ($75/month) and Professional Advanced ($150/month) tiers, as part of Google’s broader push to apply AI toward climate monitoring and sustainability efforts.

RESEARCH

MIT: Why AI data centers are measured in gigawatts

Data centers are often measured in gigawatts (GW) because power capacity directly limits how much computing hardware and therefore how much AI processing they can run. Every server, cooling system, and networking device consumes electricity, and AI workloads in particular are extremely power-hungry due to the energy demands of GPUs and specialized AI chips like Nvidia’s H100 or AMD’s MI300. For example, a single AI training cluster might draw tens of megawatts enough to power a small town so describing a facility as “a 100 MW data center” quickly communicates how much computational load it can sustain.

Using energy input as the key metric is convenient because more power generally means more servers and higher compute capacity, even though efficiency (PUE), cooling, and design can affect actual output. As AI datacenters grow into the gigawatt scale, this reflects not physical size but electrical throughput a proxy for how many petaflops or exaflops of AI computation can run simultaneously. In short, gigawatt-scale data centers exist because AI chips convert enormous amounts of electricity into computational power, making total energy capacity the simplest, most comparable way to measure the scale of modern AI infrastructure.

STARTUPS

Claude adds new memory features

Anthropic’s new Claude memory feature lets paid users carry context across chats for weeks or months, view and edit what’s remembered, and maintain separate memories for different projects similar to how ChatGPT and Google Gemini already work.

ChatGPT has long-term memory across all tiers (including free) that can be managed or deleted in settings, while Gemini offers “Saved Info” and automatic recall of past chats tied to a user’s personal context. Claude’s main differentiators are its project-based memory separation, editable transparency, and data export/import option, all backed by safety testing to prevent misuse.

BIG TECH

OpenAI acquires AI startup

OpenAI has acquired Software Applications Inc., a startup founded in 2023 by former Apple engineers who helped build the iPhone’s Shortcuts app, as part of its effort to enhance how ChatGPT handles computer-based tasks. The acquisition will bring Software Applications’ roughly dozen-person team and its AI-powered Mac interface technology into ChatGPT, though financial terms were not disclosed.

This deal continues OpenAI’s aggressive acquisition streak following its $1.1 billion purchase of product-testing firm Statsig and a $6.5 billion all-stock acquisition of an AI device company co-founded by former Apple design chief Jony Ive, contributing to its recent $500 billion valuation.