After Nvidia founder Jensen Huang stepped down as CEO, the stock fell as much as 10% in premarket trading. We believe investors are worried about more than the founder’s departure: governance, succession and regulatory pressure behind AI’s high-growth trajectory are all rising at once.
What This Is
Nvidia announced that Huang will step down as president, CEO and board member, with Chief Financial Officer Colette Kress serving as interim CEO. At the same time, the US Department of Justice is pursuing multiple investigations involving Nvidia, including an antitrust investigation. The company remains a critical supplier of AI computing infrastructure, but the founder’s personal influence, continuity of governance and regulatory exposure are now being assessed together in the valuation.
On the other side of the industry, Baidu declared at its 2026 World Conference that “the next chapter of artificial intelligence is the intelligent organization.” It launched the large-model trading platform “Ci Suo,” previewed ERNIE 5.5 Flash and upgraded four agents. Agents are AI programs capable of invoking models, tools and business systems to complete tasks. They once depended on people to operate them step by step; now they are beginning to coordinate work across applications.
Compute investment is still accelerating. SpaceX plans to invest $74 billion in supercomputing facilities in Arizona and Texas, backed by gas-fired power generation. Doubao Work is connecting with third-party models such as Kimi K3.1 Pro. Taken together, the industry is advancing three tracks simultaneously: expanding compute supply, embedding agents in workflows, and building mechanisms for model trading and collaboration.
Industry View
We believe agents are moving from demonstrations into warehousing, production scheduling and office environments, showing that enterprise priorities are shifting from whether models can answer questions to whether they can complete tasks reliably. Zhejiang has issued guidelines for building industrial agent platforms. SMEs will no longer need to develop the entire technology stack themselves, and deployment barriers are likely to fall.
The risks, however, are equally clear. Microsoft’s latest round of workforce adjustments, involving nearly 60,000 positions, makes AI’s impact on jobs feel more immediate. If agents are granted excessive permissions, they may access customer information without authorization or breach sandbox boundaries—the isolated runtime environments that prevent uncontrolled access. Connecting Doubao Work to multiple models also raises questions about data flows, cost control and accountability. The more autonomous an agent becomes, the more clearly enterprises must define permissions, audit trails and human-intervention mechanisms before deployment.
Huang’s handover is even more cause for concern. The stock fell as much as 10% at one point, indicating that investors are beginning to distinguish between Nvidia’s technological advantages and its dependence on Huang personally. The antitrust investigation may not change compute demand, but it could increase the cost of adjusting the business. It is notable that $74 billion-scale compute projects and workforce cuts involving nearly 60,000 positions are occurring at the same time. AI investment returns will be judged not only on efficiency, but also against pressure from energy, employment and capital expenditure.
Impact on Regular People
For enterprise IT: As agents enter offices and business processes, IT departments will shift their focus from deploying models to managing permissions, governing data and tracing failures. During procurement, teams should first test task-success rates and human-intervention mechanisms rather than judging systems only by their demos.
For individual careers: Standardized documentation, information retrieval and scheduling work are more likely to be broken into agent-collaboration workflows. The valuable skills will increasingly be business judgment, result review and cross-system coordination—not merely prompt techniques for a particular model.
For the consumer market: Model subscriptions and agent services will become further integrated into office software and e-commerce platforms, with consumers paying more often by API call allowance or completed task. Lower prices will not necessarily mean lower total spending, while frequent automated execution may introduce new privacy and subscription costs.