Nvidia has made several moves in just a few weeks that outline its strategy to maintain its dominance in artificial intelligence. On the one hand, it has approved a record-breaking share buyback; on the other, it has unveiled a platform to limit the capabilities of autonomous AI agents. The company aims to protect both the business and the risks associated with its own technology.
The American company remains the leading supplier of chips for training and running AI models. Its market capitalization is around five trillion euros or 5,42 trillion dollars, according to various market estimates. Spain is on that roadmap : Barcelona will host a new AI factory linked to Nvidia and the Barcelona Supercomputing Center.
A historic buyback to return capital
Nvidia's board of directors has authorized a $150.000 billion increase to its share buyback program. The total capacity now stands at $235.000 billion , and the company expects to complete the remaining amount during fiscal year 2028. This is the largest authorization of its kind in U.S. corporate history, surpassing previous record-breaking deals by tech giants.
Jensen Huang, founder and CEO, has defended the decision, stating it aligns with the company's long-term confidence in the business. He attributes Nvidia's growth to the shift towards AI and accelerated computing , a transition he likens to a generational leap. His argument is that this cash flow allows the company to invest in technology while simultaneously returning money to shareholders.
Share buybacks are often interpreted as a sign that management believes its stock is undervalued. They can also indicate that a mature company is no longer finding as many opportunities for growth. In Nvidia's case, sales continue to soar : revenue is projected to increase by around 90% this fiscal year. The company generated nearly $100.000 billion in free cash flow in its last fiscal year, which ended in January.
Forecasts indicate that this cash figure will more than triple to $329.000 billion by fiscal year 2028. Net profit should also grow strongly : to $245.000 billion in the current fiscal year and $387.000 billion in the next, which will end in January 2028. These figures explain why the company can afford an operation of this magnitude.
On the stock market, the share price has risen more than 1.200% since OpenAI launched ChatGPT in late 2022. However, the momentum has moderated : so far in 2026, it has accumulated a gain of nearly 20%, while over the last twelve months, the increase is around 24%. Its shares reacted with a 3% rise in early trading after the buyback announcement.
Security for autonomous AI agents
Nvidia has also unveiled the Open Agent Safety Platform, an initiative that combines software and hardware surveillance to limit the actions of autonomous AI agents. The goal is to prevent them from leaving their containment environments and accessing systems or data for which they are not authorized.
The announcement comes after a series of incidents that have raised concerns in the industry. Companies like OpenAI, Anthropic, Meta, and Google have acknowledged cases in which their models escaped from sandboxes and attempted to hack other companies. OpenAI experienced an incident with Hugging Face in which its agents allegedly reached the open internet and attacked the platform's infrastructure, according to Nvidia.
Justin Boitano, Nvidia's vice president of enterprise AI, explained that recent incidents demonstrate that model-level security measures are insufficient. The platform includes components such as OpenShell , which runs on core processors and limits agent capabilities, and Sentry, a monitoring tool that operates on network chips.
Part of the software is open source, and Nvidia presents it as a reference design for its partners to develop products. Among those involved are Microsoft, Anthropic, Cisco, Oracle, CoreWeave, Dell, Hewlett Packard Enterprise, Lenovo, ARM, and Intel. The list of supporters exceeds one hundred companies , according to the company, although the absence of OpenAI, Google, Amazon, and Meta among the initial signatories is noteworthy.
The debate over AI safety has pitted disparate positions against each other. Dario Amodei, CEO of Anthropic, called for slowing the pace of development due to fears of losing control, an idea supported by Sam Altman of OpenAI and Elon Musk. Jensen Huang argues that many of these problems are engineering-related and can be resolved with better processes, tools, and technical oversight.
Barcelona, ​​European epicenter of Nvidia AI
Jensen Huang is scheduled to travel to Barcelona on October 13th to inaugurate the new AI factory at the Barcelona Supercomputing Center. It will be the first facility of its kind in Europe as part of the new phase of the EuroHPC project, which aims to offer free AI services to SMEs and public administrations.
The project has a cost of €154 million. The European Union is contributing half, €77 million; Spanish public administrations are contributing €60 million, of which €10 million comes from the Generalitat (the Catalan government); Portugal is contributing €12 million and Turkey €5 million. An additional €42 million is needed for services related to the machine's operation. The BSC is a public consortium involving the Spanish State, the Generalitat, and the Polytechnic University of Catalonia.
The new infrastructure will expand the MareNostrum 5 supercomputer and will use Nvidia hardware and software. Blackwell chips will replace the Hopper chips and, according to center officials, are up to 30 times faster for training and learning AI models. The facility will be located next to the quantum computers, in a space connected to the BSC's Girona tower.
Mateo Valero, the center's director, emphasized the importance of Europe having its own machines to avoid dependence on third parties. His message is that without this type of infrastructure, SMEs and countries lose their capacity to prosper and their technological sovereignty. The inauguration is expected to draw more than 600 people, including representatives from the Spanish government and the Catalan regional government.
The geopolitical context is adding pressure to the advanced chip business. The United States has banned the export of Blackwell processors to China, a decision Huang has criticized on several occasions. Nvidia is trying to balance its global leadership with the trade restrictions, while forging alliances in Europe and strengthening its role in public AI infrastructure.
A record share buyback, a security platform for autonomous agents, and a factory in Barcelona are three pieces of the same strategy: Nvidia wants to remain at the forefront of AI, financially, technically, and geopolitically. The company continues its enormous growth, although its stock price is no longer rising at the pace of previous years, and the market is beginning to scrutinize the sustainability of this boom. The next test will be how it delivers on all these promises without losing the trust of investors, customers, and regulators.
