Nvidia CEO Jensen Huang has offered a counter-narrative to the growing anxiety around artificial intelligence and job losses, arguing that the AI boom is set to create a wave of high-paying jobs far from traditional office settings.

Speaking at the World Economic Forum in Davos on Wednesday, Huang said the global rush to build AI infrastructure is already driving demand — and wages — for skilled trades, with salaries reaching into six figures for those helping to construct and maintain what he described as “AI factories.”

At the heart of Huang’s argument is scale. The global race to deploy artificial intelligence is no longer only about algorithms and software models. It is about factories, power systems, cooling plants, fiber networks, and specialized facilities that can house and operate vast amounts of computing equipment. Huang described this moment as the largest infrastructure buildout ever undertaken, measured not only in dollars but in geographic reach and industrial complexity.

That buildout, he said, is already changing who benefits from the AI economy. Chip fabrication plants, data centers, and so-called AI factories require armies of electricians, plumbers, steelworkers, construction crews, and network technicians. These are roles that cannot be easily automated and cannot be filled overnight. As demand surges faster than supply, wages are rising sharply.

Huang told the Davos audience that salaries in some of these trades are nearly doubling, pushing total compensation into six-figure territory for workers involved in building and maintaining AI infrastructure. The implication is that while AI threatens to compress pay and reduce headcount in parts of the white-collar economy, it is creating scarcity — and pricing power — for skilled manual labor.

That view adds an important layer to a week dominated by warnings. Consulting firm Challenger, Gray & Christmas has linked nearly 55,000 U.S. layoffs in 2025 to AI adoption, with companies such as Amazon, Salesforce, Accenture, and Lufthansa pointing to automation and efficiency drives. IMF managing director Kristalina Georgieva captured the prevailing mood when she said AI is hitting the labor market “like a tsunami,” leaving governments and companies unprepared.

Huang does not deny the disruption. Instead, he is arguing that the focus has been too narrow. The AI economy, in his telling, is not a purely digital phenomenon. It is an industrial one, tied to energy systems, land use, supply chains, and construction capacity. Every new model trained and deployed increases demand for physical assets that must be built, installed, and serviced by people.

That argument finds support in Microsoft research released in 2025, which examined how often workers rely on AI tools to complete their tasks. Analyzing about 200,000 Bing Copilot conversations, Microsoft found that roles involving physical work with people or machines showed the lowest reliance on AI assistance. Jobs ranging from painters and plasterers to ship engineers and healthcare support workers were among the least exposed to automation pressure.

In practical terms, that means AI is not flattening the labor market evenly. It is accelerating a long-running divergence. Office roles built around routine information processing are becoming easier to automate or augment, while hands-on technical work tied to complex physical systems is becoming more valuable.

European policymakers at Davos acknowledged the implications. Roxana Mînzatu, the European Commission’s executive vice president for social rights and skills, said the semiconductor industry alone is searching for tens of thousands of vocationally trained workers. Her comments underline a growing concern across Europe and North America: the bottleneck in the AI era may not be software talent, but electricians, technicians, and engineers who can physically deliver projects.

The energy dimension adds another layer of pressure. AI infrastructure is energy-hungry, and regions with high power costs or limited grid capacity face constraints on how quickly they can build. That reality links Huang’s labor optimism to a broader policy challenge. Training workers is only part of the equation. Governments must also expand energy supply, streamline permitting, and modernize grids if they want to capture the industrial upside of AI.

There are social implications as well. For years, political leaders have encouraged university education as the primary path to economic security, even as tuition costs rose and returns became less certain. In the United States, the annual cost of attending a four-year public college increased by about 30% between 2011 and 2023, according to CNBC Make It calculations. Over that period, enrollment fell by roughly 2 million students.

At the same time, skilled trades have gained appeal, particularly among younger workers. Data from the Department of Labor and payroll firm Gusto show that Gen Z now accounts for a growing share of new hires in trade roles, outpacing their representation in the overall workforce. For many, the appeal is straightforward: lower training costs, faster entry into paid work, and wages that now rival or exceed many graduate-level office jobs.

Huang’s message taps directly into that shift. “You don’t need to have a PhD in computer science to make a great living,” he said, framing AI as a force that could rebalance opportunity rather than concentrate it further.

Still, the transition is not frictionless. Scaling vocational training fast enough to meet demand will test education systems that have spent decades prioritizing academic pathways. Labor shortages could delay projects, inflate costs, and slow the rollout of AI infrastructure. And while trade jobs may be safer from automation, they are not immune to economic cycles or policy shocks.

Even so, Huang’s intervention in Davos reframed the AI debate in a way that many executives and policymakers have avoided. The question is no longer only how many jobs AI will eliminate, but which kinds of work it will elevate. If the AI race continues at its current pace, the winners may include not just chip designers and software engineers, but the people wiring, cooling, and powering the factories that make the technology possible.

U.S. Treasury yields held in a narrow range on Thursday, as investors parsed a steady flow of economic data while keeping one eye on an increasingly complex mix of trade signals, geopolitics, and political developments in Washington.

The benchmark 10-year Treasury yield edged higher by just over one basis point to 4.265%, a move that signaled caution rather than conviction. The 30-year yield was largely unchanged at 4.873%, while the policy-sensitive 2-year note climbed by more than a basis point to 3.61%. The muted shifts pointed to a market struggling to reconcile strong domestic fundamentals with rising uncertainty around policy direction.

Early trading leaned slightly bearish for bonds after new labor market data reinforced the view that the U.S. economy remains on a firm footing. Initial jobless claims for the week ended January 17 fell to 200,000, below the 208,000 forecast by economists surveyed by Dow Jones. Claims were marginally higher than the prior week after an upward revision, but the broader trend continues to show limited stress in the labor market.

For bond investors, the implication was straightforward: an economy still generating jobs at a solid pace reduces the urgency for aggressive interest-rate cuts. Chris Rupkey, chief economist at FWDBONDS, said the data fit a pattern of economic growth running well above 3%, with labor indicators pointing to a potentially strong January employment report.

In his view, the resilience of growth and hiring suggests that the economy is performing well without additional intervention. That assessment matters for Treasuries, as expectations around Federal Reserve policy remain one of the dominant drivers of yields, particularly at the front end of the curve.

Later in the session, attention shifted to inflation, where the Commerce Department reported that both headline and core personal consumption expenditures price index readings for November matched expectations. The PCE index is closely followed by Federal Reserve officials, and the absence of any upside surprise helped calm fears that inflation pressures were re-emerging after months of gradual easing.

Still, the data did little to move yields decisively. Investors appear increasingly sensitive not just to economic releases, but to how those figures interact with political signals coming from the White House and the Federal Reserve.

One such signal came from President Donald Trump, who said in an interview with CNBC at the World Economic Forum in Davos that he may have already selected his nominee to replace Jerome Powell as Fed chair. Trump said the shortlist had narrowed sharply, suggesting that a final decision was effectively settled in his mind, though he declined to identify the candidate.

The comment injected an additional layer of uncertainty into markets. Even though Powell’s term still has time remaining, early speculation about the next Fed chair raises questions about the future policy stance of the central bank, particularly at a moment when investors are debating how long rates will remain elevated.

That uncertainty is compounded by scrutiny surrounding Powell himself. The Fed chair recently said he was under investigation by the Department of Justice in connection with the $2.5 billion renovation of the Federal Reserve’s headquarters in Washington. While the issue has not altered policy operations, it has added political noise at a time when markets are highly sensitive to any hint of instability at the central bank.

Geopolitical developments also played a role in shaping sentiment. Trump said he had reached what he described as a “concept of a deal” on Greenland, following earlier remarks that he would not pursue control of the Danish territory through military means. In a subsequent post on Truth Social, he said tariffs announced last week on eight European countries opposing a U.S. takeover would not take effect.

The decision eased immediate trade-related anxiety that had weighed on markets earlier in the week. Tariff threats had contributed to sharp swings across equities, currencies, and bonds, with Treasuries initially benefiting from a flight to safety. Thursday’s pause in escalation removed one source of pressure, leaving yields to drift rather than trend.

The combination of firm economic data, steady inflation readings, and shifting political signals left the Treasury market in a holding pattern. Shorter-dated yields continued to show sensitivity to labor strength and Fed policy expectations, while longer-dated yields suggested investors remain uncertain about the long-term path of growth, inflation, and fiscal policy.

Currently, the bond market appears to be waiting for a clearer catalyst. A stronger-than-expected jobs report could push yields higher by reinforcing the case for rates to stay elevated. On the other hand, renewed trade tensions or sharper political uncertainty could revive demand for safe-haven assets. Until one of those forces asserts itself, Treasury yields look set to remain range-bound, even as the list of risks facing investors continues to grow.

Big technology companies and a growing crop of startups are racing to use generative artificial intelligence to build software and hardware for children.

Most of those efforts, however, still lean heavily on text or voice-based interfaces — formats that often struggle to hold a child’s attention. Three former Google employees believe that gap is precisely where their new startup, Sparkli, has an opening, according to TechCrunch.

Founded last year by Lax Poojary, Lucie Marchand, and Myn Kang, Sparkli is an AI-powered interactive learning app designed to turn children’s questions into immersive, multimedia “expeditions.” The founders say the idea emerged from a practical frustration they experienced as parents.

“Kids, by definition, are very curious,” Poojary said in an interview. “My son would ask me questions about how cars work or how it rains. I would try using tools like ChatGPT or Gemini to explain these concepts to a six-year-old, but that’s still a wall of text. What kids want is an interactive experience.”

The team behind Sparkli brings deep experience inside Google’s ecosystem. Before this venture, Poojary and Kang co-founded Touring Bird, a travel aggregator, and Shoploop, a video-focused social commerce app, both developed within Google’s Area 120 internal incubator. Poojary later worked on shopping products across Google and YouTube. Marchand, now Sparkli’s chief technology officer, also co-founded Shoploop and went on to work at Google.

That background shapes Sparkli’s core pitch: generative AI should not just answer questions, but create experiences. Poojary explains the evolution this way: years ago, a child curious about Mars might have been shown a picture; later, a video. Sparkli aims to let children explore and interact with what Mars might feel like, rather than passively consuming information.

At a time when many education systems struggle to keep pace with rapid technological change, Sparkli is positioning itself as a supplement rather than a replacement for classrooms. The app focuses on topics that are often underrepresented in traditional curricula, including financial literacy, entrepreneurship, and design skills. Each topic becomes an AI-generated learning journey, built on demand.

Children can choose from predefined subjects or ask their own questions, which the system then turns into a structured learning path. Each topic is broken into chapters that combine audio narration, text, images, video clips, quizzes, and games. The app also features daily highlighted topics to encourage regular exploration, as well as “choose-your-own-adventure” style paths that remove the pressure of right-or-wrong answers.

Under the hood, Sparkli relies heavily on generative AI to produce its content in real time. The company says it can generate a complete learning experience within about two minutes of a child asking a question, and that it is working to shorten that turnaround further. This on-the-fly approach allows the app to adapt to a wide range of interests without relying on a fixed content library.

The founders are careful to draw a distinction between Sparkli and general-purpose AI assistants. While chatbots can explain concepts, Poojary argues they are not designed with children’s cognitive development in mind. To address that, Sparkli’s first hires included a PhD-trained specialist in educational science and AI, as well as a classroom teacher. The goal, the company says, is to ensure that pedagogy — not just technology — shapes how content is delivered.

Safety is another central concern, particularly as AI tools for children face growing scrutiny. OpenAI and Character.ai are among the companies facing lawsuits from parents who allege their products encouraged harmful behavior. Sparkli says it has taken a more restrictive approach. Certain topics, such as sexual content, are entirely blocked. When children ask about sensitive issues like self-harm, the app shifts toward teaching emotional intelligence and encourages conversations with parents, rather than attempting to handle the issue autonomously.

So far, Sparkli’s early traction has come through schools. The company is piloting the app with an educational institute that serves a network of schools reaching more than 100,000 students. Its current target audience is children aged five to twelve, and it tested the product in more than 20 schools last year.

To support classroom use, Sparkli has built a teacher module that allows educators to assign content, track progress, and set homework. Teachers can use the app to introduce a topic at the start of a lesson, then transition into discussion, or to extend learning after class. According to Poojary, feedback from these pilots has been encouraging, with teachers using Sparkli both as a teaching aid and as a way to gauge student understanding.

The app borrows engagement mechanics from consumer platforms such as Duolingo, including streaks, rewards, and personalised avatars. Children earn quest cards linked to their avatars as they complete lessons, an approach the company hopes will make learning feel closer to play than to homework.

For now, Sparkli plans to focus on partnerships with schools globally. Consumer access, allowing parents to download the app directly, is expected to follow by mid-2026.

The startup recently raised $5 million in pre-seed funding led by Swiss venture firm Founderful, marking the firm’s first investment focused purely on education technology. Founderful’s founding partner, Lukas Wender, said the decision was driven by both the team’s technical background and the perceived gap in what children are taught.

“As a father of two kids in school, I see them learning interesting things, but not topics like financial literacy or innovation in technology,” Wender said. “From a product point of view, Sparkli gets them away from video games and lets them learn in an immersive way.”

Sparkli’s bet is that the future of learning for children will depend less on answers and more on experiences — and that making AI engaging, safe, and pedagogically sound may be the real challenge ahead.