The AI boom
Nvidia longterm CEO Jensen Huang resigned as CEO on 12 Nov 2026, after a week of reports about his personal behavior. 1 His resignation is a major transition for the company and the entire AI industry, as he has shaped it for more than three decades.
Nvidia remains the key supplier to the AI boom. Its chips power both the training of models (inference) and the usage of AI systems at scale. Demand has surpassed supply, giving the company high pricing power. Nvidia’s gross margin has recently been around 74–75%. 2
A major wave of AI infrastructure spending is also happening. The hyperscalers are expected to spend more than $700 billion on CapEx in 2026 for AI infrastructure. 3 Nvidia, meanwhile, also has invested in AI companies and AI infrastructure providers. In exchange, some have agreed to buy and use Nvidia chips. Those circular investments have created a connected financing system. Some analysts call Nvidia nothing less than the “Central Bank of AI.”
After the recent events, Nvidia’s shares fell 12.78% over the week ending 13 Nov 2026.
The risks behind the spending for AI infrastructure
Investors now take the risks begind the AI infrastructure buildout into focus:
Key-person risk for Nvidia: Jensen Huang led Nvidia for more than three decades and was known for an wide span of control in his leadership, with 60 direct reports. 4 Finding a successor who can lead a company like this will be a challenge.
Competition gets harder: AMD and custom chips from hyperscalers, for instance Google, create more competition. Nvidia’s dominant market share is not assured.
AI Infrastructure buildout could slow down: Local opposition, environmental concerns, and potential shifts in US regulation on the datacenter buildout could slow datacenter construction and, as result, decelerate demand for AI hardware.
Financing risk: The AI datacenter buildout increasingly relies on debt and intransparent financing arrangements. If access to capital weakens, that high debts could become a source of stress.
Unproven returns: AI companies still have to show that their capital spending can produce sustainable returns. That proof has not yet been delivered.
Containment before deployment
The departure of the Nvidia CEO is a critical moment for Nvidia and a loss for the AI industry. He has promoted the idea that companies, who develope or use AI are responsible for the safety of AI. For instance, he has argued that AI agents need containment and monitoring. In a Bloomberg interview, he said developers who cannot contain and monitor their AI systems should not test or deploy them. 5 Derived from that, there two aspects to make the use of AI agents in enterprises safe:
Containment: Apply least-privilege access. Models should only have access to the systems and data they need for a specific task. That results in narrow, specialized agents rather than general-purpose agents with broad access to a wide set of tools.
Monitoring: Log and review agent actions, and establish clear ways to revoke access if an agent behaves unexpectedly.
My Take: Why specialized agents have the future
A safe architecture for AI agents in enterprises is still yet to be found. Companies need to experiment in their own business processes and workflows with specialized agents to make AI for their use cases safe rather than applying general-purpose agents.
If I had to bet on where the market lands...
It would be on contained specialized agents and a smart orchestration layer that coordinates them.
Further viewing: Interviews on AI and leadership.
Bloomberg Originals
BBC News:
Footnotes
[1] The New York Times, 12 Nov 2026
[2] Nvidia Newsroom, 26 Aug 2026
[3] Forbes, 07 Oct 2026
[4] Fortune, 14 Nov 2026
[5] Bloomberg Originals on Youtube, 12 Nov 2026

Top comments (0)