#1OpenAI Kills Sora, Disney Walks on $1B Deal
Six months after launch and three months after inking a character-licensing deal with Disney, OpenAI is shutting down its Sora video app — downloads had already cratered 75% from their November peak. Disney is pulling its pledged $1 billion investment, and no money ever changed hands. OpenAI says compute is too precious to burn on Sora when enterprise and infrastructure demand the resources.
*Source: [Axios](https://www.axios.com/2026/03/24/openai-discontinue-sora-video-app) | [Variety](https://variety.com/2026/digital/news/openai-shutting-down-sora-video-disney-1236698277/)*
---
#2Google's TurboQuant: 6x LLM Compression, Zero Accuracy Loss
Google released TurboQuant, a KV-cache compression algorithm that squeezes LLMs to 3 bits per value with no measurable accuracy drop and up to 8x speedup on H100 GPUs. The algorithm combines polar coordinate quantization (PolarQuant) with a sign-bit residual step (QJL) — no retraining required. This is the kind of quiet infrastructure win that could put powerful models on consumer hardware at scale.
*Source: [Tom's Hardware](https://www.tomshardware.com/tech-industry/artificial-intelligence/googles-turboquant-compresses-llm-kv-caches-to-3-bits-with-no-accuracy-loss) | [TechCrunch](https://techcrunch.com/2026/03/25/google-turboquant-ai-memory-compression-silicon-valley-pied-piper/)*
---
#3Anthropic Launches Claude "Computer Use" and Dispatch
Anthropic introduced agentic capabilities that let Claude autonomously operate across a user's system — acting as a meta-agent that can orchestrate other AI tools. The Dispatch feature allows task delegation across agents. This is a direct shot at OpenAI's Operator and marks Anthropic's most aggressive push into practical, on-device agentic deployment.
*Source: [The Neuron](https://www.theneuron.ai/ai-news-digests/around-the-horn-digest-everything-that-happened-in-ai-on-wednesday-march-25-2026-/)*
---
#4OpenAI Seeks $10B at $730B Valuation With Guaranteed Returns
OpenAI is raising $10 billion in new capital, dangling a 17.5% guaranteed return to lure private equity — an unusual structure that signals the company needs cash badly enough to promise downside protection. This comes after the confirmed $110B Series G earlier this year and amid a fierce arms race for compute.
*Source: [CryptoIntegrat / AI News March 25](https://www.cryptointegrat.com/p/ai-news-march-25-2026)*
---
#5Google Gemini Personal Intelligence Goes Live for All U.S. Users
Google is rolling out Gemini's Personal Intelligence feature nationally, allowing the assistant to tap Gmail, Photos, YouTube, and other connected apps for context-aware responses. Google also confirmed it's leaving the door open to serving ads inside Gemini — a move that would reshape the economics of AI assistants and further squeeze traditional search.
*Source: [FinancialContent](https://markets.financialcontent.com/stocks/article/marketminute-2026-3-25-the-ai-era-20-how-agentic-systems-and-gigawatt-scale-infrastructure-are-redefining-market-leadership-in-2026)*
---
#6CFOs Expect AI to Shrink Headcounts in 2026
A Wall Street Journal survey of chief financial officers finds that finance leaders are now building workforce reductions into plans for 2026, with administrative and office roles most exposed. This is the signal shifting from think-piece speculation to actual budget decisions — companies are planning around automation, not just experimenting with it.
*Source: [TechStartups](https://techstartups.com/2026/03/25/top-tech-news-today-march-25-2026/)*
---
#7Meta Hit With $375M Verdict Over Child Safety Failures
A New Mexico jury ordered Meta to pay $375 million, finding the company liable for misleading users about platform safety and enabling the exploitation of minors on Facebook and Instagram. The case argued Meta knew about the risks and prioritized growth over child protection. Expect this verdict to accelerate legislative pressure on AI platforms over similar liability questions.
*Source: [TechStartups](https://techstartups.com/2026/03/25/top-tech-news-today-march-25-2026/)*
---
**Big Picture**
Today's news tells a story of triage. The AI industry is running out of runway to treat every experiment as a viable product — Sora's death is a clean example of compute economics forcing hard cuts. Meanwhile, the real bets are consolidating around agentic systems (Anthropic, OpenAI's "Spud"), on-device efficiency (TurboQuant), and deep platform integration (Gemini Personal Intelligence). The frothy "demo era" is over. What's left standing is what can generate revenue, reduce costs, or survive on a constrained compute budget. The CFO survey is a reminder that for most of the economy, AI isn't a research question anymore — it's an org chart question.
