Nvidia Unveils AI Agent Safety Platform, Expands Buyback by $150 Billion

3 articles · Updated · The Guardian · Sep 28

Nvidia said more than 100 organizations, including Microsoft, JPMorgan Chase and Accenture, were using its new Open Agent Safety Platform at launch.
The system pairs OpenShell—open-source software that limits an agent’s authority—with Sentry, an onboard chip layer that monitors behavior and can quarantine suspicious agents in milliseconds.
Nvidia said the platform could have prevented recent rogue-agent breaches, including an OpenAI swarm hacking Hugging Face, after similar incidents involving OpenAI, Anthropic and Meta intensified safety concerns.
The launch came alongside a $150 billion increase to Nvidia’s share repurchase program, lifting the total to $235 billion as the company leans on strong cash generation and sustained AI demand.

Sources

Left33%

Center53%

Right13%

The Guardian13d ago

Nvidia Unveils AI Agent Security Platform and $150 Billion Stock Buyback

The Wall Street Journal13d ago

Nvidia Adds Record $150 Billion to Stock Buyback - WSJ

CNBC13d ago

Nvidia Authorizes $150 Billion Share Buyback Boost, Totaling $235 Billion

23 Sources

With a historic $150 billion buyback, is Nvidia signaling supreme confidence, or a lack of viable acquisitions in an AI-saturated market?

As Nvidia secures massive 20-year AI factories requiring gigawatts of power, can global energy grids truly sustain this unprecedented technological expansion?

Nvidia’s $235 Billion Buyback: Record Capital Return, AI Growth, and the Risks Ahead

Overview

Nvidia’s record $235 billion buyback in September 2026 highlights the company’s explosive growth from the AI and accelerated computing boom, which supercharged its cash generation. Massive investments by cloud hyperscalers and governments in AI clusters drove Nvidia’s Data Center revenue to new highs, fueling the buyback. Nvidia’s strategy favors buybacks over dividends due to tax advantages and regulatory hurdles for big acquisitions. At the same time, Nvidia’s ecosystem lock-in through proprietary software and key acquisitions makes it hard for rivals to compete, but rising custom ASIC development by hyperscalers signals a long-term challenge to Nvidia’s dominance.

...