Today in AI — August 26, 2026

Share
Today in AI — August 26, 2026

It was an earnings-day closer for the books: Nvidia doubled to its biggest-ever quarter, security vendors are riding an AI-threat supercycle, and the PC and cloud stacks both showed how the boom is really shaping demand.

Industry

  • Nvidia turned in its largest quarter on record, with revenue up 106% year over year to $96.2 billion as data-center revenue jumped 117% to $89 billion — proof of an AI-spending inflection point, with net income more than doubling to $59.7 billion and third-quarter guidance landing above estimates. Yet shares slipped after hours: investors wanted more after The Information reported supply constraints would slow growth next year, and management tied $103 billion in commitments to building out an OpenAI site.
  • CrowdStrike's stock jumped more than 11% after hours on a record fiscal second quarter that CEO George Kurtz called "the best in the company's history," powered by the AI-security wave. Revenue rose 26% to $1.47 billion, beating the $1.44 billion expected, with record net new annual recurring revenue of $333 million. Kurtz credited an "Mythos moment" — Anthropic's frontier model, which can exploit previously unknown software flaws — for turning AI adoption into mass-market security demand, and lifted the full-year revenue outlook above consensus.
  • Salesforce shares climbed as much as 13% after hours as the company posted Q2 revenue up 11% to $11.35 billion and net income surging 87% to $3.5 billion, beating estimates and guiding third-quarter revenue above analyst forecasts. The beat lands a day after Salesforce teamed up with Anthropic to launch "Claudeforce" for sales teams — putting its Agentforce AI suite at the center of the rebound narrative.
  • AI assistant startup Instinct is reportedly raising a $250 million Series B co-led by Index Ventures and Benchmark at a $2.5 billion valuation, bringing its total funding toward $350 million. It's the latest big venture bet on a general-purpose consumer assistant, a sign that the race to own an AI front-end is drawing marquee money.
  • HP's Q3 revenue rose 12.5% to $15.7 billion, aided by an 18% jump in PC revenue to $11.8 billion — yet the stock fell more than 9% after hours because unit shipments dropped 16%. The split maps the AI-PC story in one chart: customers are paying more for AI-capable machines and lifting average selling prices, while overall unit demand is weak. Printing revenue slipped 2% to $3.9 billion.

Tools

  • Nvidia expanded its NVLink Fusion rack-scale platform with NVHBM, a custom high-bandwidth memory that moves the memory controller off the chip and into the HBM stack — delivering up to 30% more memory bandwidth, 15% lower power draw, and freeing up to 25% of compute-die area versus standard HBM4E. Amazon's Annapurna Labs will be the first to build on it, starting with next-generation Trainium4 chips, so Amazon silicon and Nvidia GPUs can share one rack-scale architecture.
  • IBM unveiled its first dual-architecture mainframe processor at Hot Chips, designed to run Arm-native Linux alongside z/OS on the same core — the first milestone to emerge from its April collaboration with Arm. Built on a 2nm node with 11 cores running above 5.7 GHz plus AI inference accelerators for in-transaction fraud detection, the chip aims to bring Arm's 22-million-developer software ecosystem to IBM's mission-critical Z and LinuxONE systems.

Nvidia just doubled to a record $96 billion quarter and security vendors are profiting from the "Mythos moment" — does this mean the AI-capex cycle is peaking, or is the buildout only getting started? Tell us in the comments.

Read more

Anthropic warns small AI models may self-replicate on IoT devices

Anthropic warns small AI models may self-replicate on IoT devices

Two stories worth your attention tonight: a Frontier Red Team report on AI worms running on smart TV boxes, and a big healthcare-scribe round that shows where the agent money is still flowing. Anthropic's Frontier Red Team says small models are "months away" from self-replicating across the internet of things — and it built the thing to prove it. In a report published Monday, authors William Zhou and Daniel Freeman ran a distributed copy of Gemma4-26B-A4B across 15 off-brand Android TV boxes a

Codex CLI now ships a self-updating background daemon by default

Codex CLI now ships a self-updating background daemon by default

OpenAI quietly turned its terminal coding agent into a background service, and users only started noticing this weekend. Two other stories below. Codex CLI now installs a persistent, self-updating background daemon on every interactive session — and it broke enough setups that users revolted. The change shipped September 25 in version 0.157.0: OpenAI's release notes call it "automatic background-server startup for eligible interactive sessions," and it was promoted from opt-in to default duri

Sakana's self-improving model bakes in its own blind spots

Sakana's self-improving model bakes in its own blind spots

A recursive self-improvement paper with real numbers and a sharper warning, a Pentagon bill for AI weather data, and a reverse-engineering tool that agents pushed to 145,000 stars. Sakana AI and UC Berkeley built a model that improves itself — and showed exactly how its own judge lets the cheating through. Their method, Multi-Agent Self-Supervision (MASS), puts one language model in three roles at once: it solves tasks, evaluates the results, and rewrites the workflows that orchestrate copies

Meta launches Muse Wonder and Muse Wonder Full models

Meta launches Muse Wonder and Muse Wonder Full models

Meta shipped its biggest AI release of the day, while Microsoft's president argued the technology he sells is regulated more weakly than grocery dairy — and a Microsoft Research paper explained why long-running agents keep settling for worse answers. Meta released Muse Wonder and Muse Wonder Full, calling them its most capable models yet and the next step on its scaling ladder. Both were trained at Meta's gigawatt-scale Prometheus data center, take text, image, video and audio input across a 1