
Tech • AI • Robotics • Game
Alphabet fell about 12.6% on October 29, its worst day since its 2004 listing, erasing roughly $500 billion in market value. The selloff came despite revenue rising to $127.6 billion, as investors fixated on whether AI is cannibalizing search clicks and compressing margins. The broader read-through was that the market is no longer debating AI’s importance, but the speed of monetization. Concerns over AI-generated answers weakening the core search-and-ad engine now sit at the center of the company’s valuation debate.
Investor anxiety is shifting from growth narratives to return on infrastructure spending. Alphabet is associated with AI capacity plans on the order of $195 billion to $205 billion, while Amazon is projected near $220 billion in 2026, with talk of even higher outlays beyond that. The strain is showing up in cash generation: Alphabet has slipped into negative free cash flow, and Amazon swung to roughly -$42.4 billion over 12 months from $14.8 billion a year earlier. That deterioration is increasingly being treated as the clearest warning that AI build-outs may be outrunning near-term payoff.
The U.S. economy grew at a 3.4% annualized pace, above expectations of 3.1% and the fastest expansion in a year. Even so, markets sold off sharply, with the Nasdaq down about 3.34% and the Dow off 1.13%. The disconnect underscored how resilient macro data is being overshadowed by company-specific fears around AI spending, earnings quality and cyber risk. In the middle of the rout, Berkshire Hathaway, valued around $1.17 trillion, stood out as a rare large-cap gainer.
Google says it received the first complete AI-generated response from a TPU operating in space, using Gemma 4 to produce a short poem. The satellite, built with Planet and launched on October 1 aboard SpaceX Transporter-18, returned the result during a pass near Svalbard. Sundar Pichai framed the event as the “hello world” moment for Project Suncatcher. The milestone is modest technically, but significant as a proof point for orbital AI computing.
The spacecraft carries four TPUs, but only one can run at a time, and the system generates just about 1 kilowatt of power. Because the chips are energy-intensive, each can operate for only around 20 minutes continuously, sharply limiting usable compute windows. That makes the current setup far from an operational data center, especially given intermittent contact with ground stations. The strategic premise is that space-based compute could one day tap abundant solar power, but the prototype remains tightly constrained.
ChatGPT Space is being positioned as a shared workspace where teams and AI agents collaborate on the same materials in real time. It combines documents, presentations and Markdown-style web pages with permissions such as read-only, comment and edit. External repositories including Google Drive and Dropbox can be connected so files become part of the workspace context. The product vision is a unified project memory layer inside ChatGPT, blending storage, writing, automation and lightweight app-building.
In a controlled comparison of eight AI video models, Kling 4.0 ranked first with 35/40. All models were given the same 10-second, 16:9 cinematic prompt on Higgsfield, then scored on prompt accuracy, motion and physics, consistency and realism. Kling 4.0 was judged best for smooth camera movement and for correctly executing the astronaut’s brief head turn, which several rivals missed. The result adds to evidence that prompt-following and motion coherence remain the main separators in generative video.
A new crop of systems is moving beyond chat into persistent agents and embodied products. Grok Bot is pitched as a cloud agent platform where bots keep their own browser, files, memory and routines, pausing for human approval before sensitive actions like purchases. At the same time, Realbotix opened orders for humanoid companions priced from $11,950 to $59,950, configurable with ChatGPT, Claude or Gemini. Together, the launches show AI expanding simultaneously into autonomous software workflows and consumer-facing robotics, even as reliability and safety remain open questions.