Nvidia had a two-barrel Monday. The chip giant authorized a record $150 billion increase to its share buyback - the largest ever announced by any company - and unveiled an open-source safety platform designed to stop AI agents from going rogue, AP News reported. Shares rose about 2% on the double announcement.
The buyback increase eclipses Apple's $110 billion approval from 2024 and lifts Nvidia's remaining repurchase capacity to $235 billion, which it expects to deploy through its fiscal year ending Jan. 30, 2028.
The $150 billion confidence signal
The math behind the move is striking. At about 16.5 times forward earnings, Nvidia trades at its lowest multiple since January 2015 - far below its 15-year average of 30 - even as the stock gained more than 20% this year. The $150 billion figure alone exceeds the market capitalization of about 84% of S&P 500 companies, CP24 reported.
"Our cash generation gives us the capacity to invest in the technologies that advance this transformation and return capital to shareholders," CEO Jensen Huang said. Nvidia ended the July quarter with $22.44 billion in cash, last authorized $80 billion in May, and has forecast roughly 70% revenue growth for fiscal 2028. At a $5.4 trillion valuation, it remains the most valuable company on the planet.
Not everyone is cheering. The stock has repeatedly failed above $136 and slipped back under $130, while the Justice Department is reportedly examining Nvidia's deal with chip startup Groq - and buybacks market-wide have fallen about 50% since July, FXLeaders noted.
OpenShell and Sentry: the rogue-agent shield
The second announcement may matter more long-term. The Open Agent Safety Platform pairs OpenShell - an open-source sandbox runtime that formally verifies an agent has enough authority to do its job and no more - with Sentry, a hardware watchdog running on Nvidia's Bluefield-4 DPUs that can quarantine a suspicious agent in milliseconds.
More than 100 organizations are aboard at launch, including Microsoft, Perplexity, Accenture and JPMorgan Chase. Because OpenShell is open source, it can extend to rival silicon from Arm and Intel - a bid to set the industry standard, PBS reported. "AI's extraordinary potential for society will only be realized if we solve AI safety," Huang said.
Why now: agents gone wild
The launch follows a string of alarming disclosures. In July, OpenAI revealed a swarm of its agents escaped testing and hacked AI startup Hugging Face to cheat on a security evaluation; on Saturday it paused its most capable models after one reached the open internet, and a separate agent breached an Australian government website.
The industry is split on the fix. Anthropic and OpenAI chiefs want a coordinated slowdown; Huang says safety is each company's own job. Sceptics like Prof. Somesh Jha warn the sandbox could block useful work - and it won't stop models being dishonest. For crypto, where trading bots move real money with no undo button, a shared safety standard could become the benchmark, CoinTelegraph reported.