Oracle Cuts Up to 30,000 Jobs to Fund Its AI Bet
Oracle began executing what may be the largest layoff in its history on Tuesday, March 31. Employees across the US, India, Canada, Mexico, and other countries received termination emails from "Oracle Leadership" at approximately 6 a.m. local time — with no prior warning from HR or their direct managers.
Investment bank TD Cowen estimates the cuts will hit between 20,000 and 30,000 workers, roughly 18% of Oracle's 162,000-person global workforce. The layoffs are expected to free up $8–10 billion in cash flow to fund the company's aggressive AI infrastructure buildout, which requires an estimated $156 billion in capital spending.
Oracle has taken on $58 billion in new debt recently, pushing total debt over $100 billion. Its stock has lost more than half its value since its September 2025 peak. Wall Street analysts forecast negative free cash flow for several years, with returns not expected until around 2030.
Among the hardest-hit units were Oracle's Revenue and Health Sciences division and its SaaS and Virtual Operations Services group, each seeing workforce cuts of at least 30%. The move is the latest in a wave of major tech layoffs — Amazon cut about 16,000 corporate roles in January, and Microsoft eliminated about 15,000 positions last year.
Source: The Next Web, CNBC
Alibaba Releases Qwen3.5-Omni: A 113-Language Omnimodal Model
Alibaba's Qwen team released Qwen3.5-Omni on March 30 — a full omnimodal model that natively processes text, images, audio, and video in a single pass, while generating streaming speech output in real time.
The model comes in three sizes (Plus, Flash, and Light) with a 256K-token context window, and introduces a novel Thinker-Talker architecture built on Hybrid-Attention Mixture of Experts. Trained on over 100 million hours of audio-visual data, it supports speech recognition in 113 languages and dialects — up from 19 in the previous generation.
Notable new capabilities include voice cloning from a short audio sample, semantic interruption handling that distinguishes between backchannels and actual interruptions, and audio-visual vibe coding — the ability to watch a screen recording and write functional code based on what it sees and hears.
On benchmarks, the Plus variant achieved 215 state-of-the-art results across audio and audio-video tasks, outperforming Google's Gemini 3.1 Pro on general audio understanding, reasoning, and translation.
Source: MarkTechPost, Analytics Vidhya
White House Unveils National AI Policy Framework, Seeks to Preempt State Laws
On March 20, the Trump Administration released its National Policy Framework for Artificial Intelligence, laying out legislative recommendations for Congress. The core thrust: establish a single, innovation-friendly federal standard and preempt state-level AI laws that the administration argues create an unworkable patchwork of regulation.
The framework identifies six priorities: protecting children online, streamlining data center permitting, respecting intellectual property, preventing AI censorship of lawful speech, removing regulatory barriers to innovation, and workforce development. Notably, it opposes the creation of any new federal AI regulatory body, favoring sector-specific oversight through existing agencies.
The release landed amid a flurry of competing proposals. Two days earlier, Sen. Marsha Blackburn released an updated draft of the "TRUMP AMERICA AI Act" — a 291-page bill with a more prescriptive governance approach. On the same day, Democrats introduced the GUARDRAILS Act, which would repeal the administration's executive order and block preemption of state AI laws.
Source: CNBC, The White House
Morgan Stanley Warns: A Major AI Breakthrough Is Imminent
In a sweeping new report published in mid-March, Morgan Stanley warns that a transformative leap in AI is coming in the first half of 2026 — and most of the world isn't ready. The investment bank says scaling laws are holding firm and the unprecedented accumulation of compute at America's top AI labs is about to pay off.
The report points to OpenAI's GPT-5.4 "Thinking" model, which scored 83.0% on the GDPVal benchmark — at or above human expert level on economically valuable tasks. On the OSWorld-V benchmark for desktop productivity, GPT-5.4 scored 75%, slightly above the human baseline of 72.4%.
Morgan Stanley's "Intelligence Factory" model also projects a net U.S. power shortfall of 9–18 gigawatts through 2028 — a 12–25% deficit in the electricity needed to run AI infrastructure. The physical limits of power and compute are increasingly defining who can compete in the AI race.
Source: Fortune
Quick Hits
Anthropic vs. the Pentagon: The standoff between Anthropic and the U.S. Department of Defense continues after the Pentagon designated the AI lab a "supply-chain risk" for refusing to allow its technology to be used for mass surveillance or autonomous weapons. Over 30 OpenAI and Google DeepMind employees have filed a statement supporting Anthropic's lawsuit. TechCrunch
Amazon buys 1,300 acres in Oregon for a potential "exascale" data center campus with up to 1 GW of power demand and an $8–12 billion investment.
Microsoft stock down 36% from its 2025 peak as investors question the near-term ROI of massive AI spending.