Jensen Huang Says 'We’ve Achieved AGI' – A Battle Over Defining Artificial General Intelligence

Jensen Huang’s AGI Claim

On the Lex Fridman podcast in March 2026, NVIDIA CEO Jensen Huang declared, “I think we have achieved AGI,” and defined AGI as an AI system that can “basically do your job” by building and operating a company worth more than $10 billion. He answered the host’s timeline question with “Now.”

Evidence from OpenClaw

Huang pointed to OpenClaw, an open‑source AI‑agent platform, as proof of the claim. OpenClaw’s viral adoption shows agents being used for personal scheduling, code generation, and social‑media account management. He suggested that if someone creates a digital influencer or a pet‑care app that suddenly goes viral, it would not surprise him. He later qualified his statement, noting that the hype may fade after a few months and that “the probability that 100 000 such agents could build a NVIDIA is—zero.”

Industry Leaders’ Reaction

Many tech executives are distancing themselves from the term “AGI,” coining vaguer, tech‑bro alternatives they consider less hype‑prone and less likely to cause public panic. Huang’s proclamation, made as NVIDIA’s market cap surpassed $4 trillion, appears aimed at reinforcing the company’s GPU dominance, since AGI development depends on massive compute power.

Anthropic vs. the U.S. Department of Defense

In the same week, Anthropic sued the U.S. Department of Defense after being placed on a “supply‑chain risk” list for refusing to develop AI for large‑scale domestic surveillance and fully autonomous weapons. The lawsuit alleges retaliation for Anthropic’s safety‑first stance, marking the first U.S. case where an AI firm confronts the government over ethical positions. The GSA terminated Anthropic’s “OneGov” contract, affecting all three government branches. Microsoft, which works with both the DoD and Anthropic, said it will continue collaborating with Anthropic while ensuring no DoD‑related projects involve the company.

Google’s TurboQuant Algorithm

Google announced TurboQuant, an algorithm that reorganizes large‑language‑model data to compress memory usage by at least six times without any loss of precision. By reducing the memory footprint, the same GPU can run models that are three to six times larger, potentially transforming the economics of AI inference.

Weekly AI Snapshot

Jensen Huang claims AGI is achieved – defines AGI as an AI that can run a $10 billion company.

Anthropic sues the U.S. Department of Defense – over being listed as a supply‑chain risk for refusing AI‑enabled mass surveillance.

OpenAI talks Helion nuclear fusion acquisition – Sam Altman steps down from Helion’s board while talks continue.

Google TurboQuant – memory compression of LLMs by ≥6× with zero accuracy loss.

Intel Arc Pro B70 GPU – 32 GB memory, AI‑focused, priced at $949.

Meta AI glasses face EU regulatory hurdles – EU battery‑removal rules affect Ray‑Ban Display rollout.