Top Stories
Jensen Huang Declares ‘We Have Achieved AGI’
NVIDIA CEO Jensen Huang made a bold claim on Lex Fridman’s podcast Monday: “I think we’ve achieved AGI.”
The statement comes with a significant caveat about definitions. Fridman proposed defining AGI as an AI capable of building and running a billion-dollar tech company. Huang’s threshold doesn’t require permanence — in his view, an AI that builds a viral app, earns a billion dollars, and subsequently shuts down still qualifies, drawing a comparison to dot-com era companies.
Most AI researchers consider this a redefinition rather than a genuine milestone. The classical definition of AGI — human-level reasoning across all domains — remains unmet. But Huang’s comments reflect how quickly the goalposts are moving as AI capabilities expand. Whether you call it AGI or not, autonomous agents can now do things that seemed like science fiction two years ago.
Source: Yahoo Finance
ARM Ships First Data Center Chip in 35-Year History
ARM announced its first in-house production silicon for data centers — the AGI CPU, built on TSMC’s 3nm process with up to 136 Neoverse V3 cores. The move marks a fundamental shift: ARM is no longer just licensing designs to other chipmakers.
The chip packs serious specs: up to 3.7 GHz boost, 12 channels of DDR5 memory delivering over 800 GB/s bandwidth, all within a 300-watt power envelope. ARM claims it delivers more than twice the performance per rack compared to the latest x86 platforms.
Meta served as the lead development partner and plans to deploy the AGI CPU alongside its custom MTIA accelerators. Other committed customers include OpenAI, Cerebras, Cloudflare, and SAP. A liquid-cooled 200kW rack configuration from Supermicro can house over 45,000 cores.
The timing isn’t coincidental — ARM is positioning itself at the center of the agentic AI infrastructure buildout, where CPU-side orchestration matters as much as raw accelerator compute.
Source: Tom’s Hardware
Anthropic Awaits Court Ruling as Judge Calls Pentagon Ban ‘Troubling’
Judge Rita Lin is expected to issue a ruling in the coming days on Anthropic’s request to halt the Pentagon’s blacklisting of Claude. The company had asked for a decision by March 26, though the court isn’t bound by that date.
At Tuesday’s hearing, Lin called the Pentagon’s action “troubling” and questioned whether Anthropic is being “punished for criticizing the government’s contracting position in the press.” She noted that the ban “looks like an attempt to cripple” the company.
The core dispute: Anthropic signed a $200 million Pentagon contract last July and was the first AI lab deployed on classified networks. But the company refused to allow Claude for fully autonomous lethal weapons or mass surveillance of Americans. The Trump administration responded by designating Anthropic a “supply chain risk.”
Court filings revealed the Pentagon told Anthropic the two sides were “nearly aligned” just a week before Trump publicly severed the relationship. The outcome could set precedent for whether the government can label AI companies as security threats based on policy disagreements rather than actual risks.
Source: CNBC
Quick Hits
-
Open-source web agent: Allen Institute for AI released MolmoWeb, an 8B parameter browser agent that interprets screenshots and performs actions. It outperforms GPT-4o on web navigation tasks and runs locally.
-
OpenAI kills Sora, preps Spud: OpenAI is shutting down Sora to free compute for its new model codenamed “Spud.” Sam Altman told employees to expect a “very strong model” in a few weeks that can “really accelerate the economy.”
-
Sam Altman’s infrastructure pivot: OpenAI’s CEO has stepped back from safety oversight to focus on fundraising, supply chains, and “building datacenters at unprecedented scale.”
-
China’s brain chip first: China approved the world’s first commercial brain-computer interface for paralysis patients. The NEO device from Neuracle can translate neural signals into hand movements without penetrating brain tissue.
-
Meta layoffs continue: Bloomberg reports Meta is cutting several hundred jobs across sales, recruiting, and Reality Labs — part of ongoing restructuring as AI capex hits $115-135 billion for 2026.
-
2026 tech layoffs surge: Over 59,000 tech jobs cut in 2026 so far, with more than 9,200 explicitly attributed to AI and automation. Amazon, Meta, Block, and Atlassian among the largest cuts.
Worth Watching
The Anthropic ruling could come today or in the next few days. A decision favoring the company would limit the executive branch’s ability to weaponize supply chain designations against AI firms that refuse military applications.
ARM’s entry into silicon production represents a fundamental shift in the chip industry power structure. If the AGI CPU delivers on performance claims, it could accelerate the already-rapid shift away from x86 in data centers.
OpenAI’s Spud model — possibly GPT-6 or an interim release — is expected within weeks. The shutdown of Sora to free compute suggests the company is prioritizing reasoning and productivity over creative tools.
Jensen Huang’s AGI claim, whatever you think of the definition, signals that the industry narrative is shifting. The question is no longer whether AI can do economically valuable work, but how quickly that capability scales.