Listen to the audio version of this article (generated by AI).
The AI industry’s biggest names descended on Davos this week with predictions that range from inspiring to alarming. Elon Musk told the World Economic Forum that AI will be smarter than any individual human by the end of 2026. Jensen Huang countered that we’re witnessing “the largest infrastructure buildout in human history.” Meanwhile, back in Silicon Valley, Mira Murati’s Thinking Machines Lab is hemorrhaging talent back to OpenAI, and Amazon just dropped an agentic AI doctor into every One Medical member’s pocket. Here’s everything that happened in AI on January 22-23, 2026.
Musk’s Davos Debut: Robots, Energy, and AGI Timelines
Elon Musk—a longtime Davos skeptic—made his first-ever appearance at the World Economic Forum on January 22, sitting down with BlackRock CEO Larry Fink in front of a packed auditorium. The interview covered everything from humanoid robots to extraterrestrial life, but the AI predictions grabbed headlines.
Musk’s timeline for artificial general intelligence has compressed dramatically. “AI smarter than any human is coming by next year,” he told Fink, predicting that artificial intelligence will surpass the collective intelligence of humanity within five years. For context, that’s 2031—a date that would have seemed absurd in mainstream AI discourse just two years ago.
But the more actionable insight wasn’t about timelines—it was about constraints. “The limiting factor for AI is electrical power,” Musk said. The chip production problem is largely solved; NVIDIA and competitors are scaling manufacturing capacity aggressively. The new bottleneck is energy. “Very soon, we’ll be producing more chips than we can power,” he warned.
China, according to Musk, doesn’t face the same constraint. “China’s growth in electricity is tremendous,” he noted, criticizing U.S. tariff barriers on solar that make domestic renewable energy more expensive. It’s a stark geopolitical framing: the AI race may ultimately be won not by whoever has the best models, but by whoever can power the data centers running them.
On the robotics front, Musk confirmed that Tesla’s Optimus humanoid robots are currently performing “simple tasks in the factory” and predicted public sales by the end of 2027. “My prediction is there will be more robots than people,” he said—a statement that lands differently depending on whether you’re a factory owner or a factory worker.
Huang’s $85 Trillion Bet: AI as Infrastructure
Jensen Huang’s Davos appearance on January 21 painted an even more expansive picture. The NVIDIA CEO pushed back against AI bubble concerns by reframing the spending entirely: this isn’t speculative investment, it’s infrastructure.
“AI is the foundation of the largest infrastructure buildout in human history,” Huang said, describing a five-layer stack that spans energy, chips, cloud data centers, AI models, and applications. While hundreds of billions have already been invested, Huang argued far more capital is needed. His estimate: over the next 15 years, roughly $85 trillion of global R&D and operating expenses will be “augmented with artificial intelligence.”
The infrastructure framing carries policy implications. “Every country should treat AI like electricity or roads,” Huang argued. For Europe specifically, he called AI robotics a “once-in-a-generation” opportunity—but only if the continent “gets serious” about energy supply.
On jobs, Huang was notably more optimistic than typical doom narratives suggest. He pointed to radiology as a case study: AI has become a key tool in the field, yet there are now more radiologists than ever. The technology is handling administrative burden and expanding diagnostic capacity, not replacing practitioners. Similar dynamics, he argued, apply to plumbers, electricians, and construction workers who are in high demand for the physical infrastructure buildout.
This message echoes themes from earlier Davos 2026 sessions, but with a key distinction: Huang is betting that AI creates more work than it destroys—at least for now.
The Davos Disconnect: Worker Anxiety Meets Executive Optimism
The billionaire optimism at Davos stands in stark contrast to worker sentiment on the ground. According to recent analysis from TechAIMint, 95% of employers expect business growth from AI—but only 51% of employees share that optimism. Nearly half of workers believe AI benefits employers more than workers.
The numbers paint a picture of growing unease: over 55,000 U.S. job cuts in 2025 were linked to AI adoption. Job postings for “AI agent” skills have risen 16-fold, yet 21% of employees expect no changes to their roles whatsoever—a cognitive dissonance that suggests either mass denial or a genuine disconnect between executive AI strategies and frontline realities.
Perhaps most telling: 40% of employees now have or are considering second jobs—nearly double the rate from 2024. Traditional career paths feel obsolete; the “portfolio career” with multiple income streams is becoming the new default. Only 50% of workers still seek manager advice, with AI increasingly filling that guidance role.
When Musk predicts robots outnumbering humans and Huang promises $85 trillion in infrastructure spending, the response from the workforce isn’t excitement—it’s anxiety about who gets left behind.
Thinking Machines Lab: The Startup That’s Bleeding Talent Back to OpenAI
While Davos dominated headlines, a quieter drama unfolded in Silicon Valley. Thinking Machines Lab—the AI startup founded by former OpenAI CTO Mira Murati—is losing key talent back to the company they left.
The departures are significant: Barret Zoph (co-founder and CTO), Luke Metz (co-founder), and Sam Schoenholz have all returned to OpenAI. According to TechCrunch, Murati terminated Zoph’s employment citing “unethical conduct”—a claim OpenAI disputes. OpenAI’s Fidji Simo stated that Zoph “told [Murati] on Monday that he was considering leaving and she fired him today,” adding that OpenAI does “not share these concerns.”
The context makes the exodus more striking. Thinking Machines raised a $2 billion seed round at a $12 billion valuation—one of the largest in Silicon Valley history. The company reportedly aimed to boost that valuation to $50 billion but has struggled to secure additional funding.
The product strategy remains murky. Thinking Machines’ only released product, Tinker, is an API for fine-tuning open-source AI models. The company has yet to train a major foundation model—the core competency its founders were known for at OpenAI. With co-founders departing and employees reportedly in discussions to follow, the startup that was supposed to challenge OpenAI may instead be serving as a talent pipeline back to it.
Amazon’s Health AI: An Agentic Doctor in Your Pocket
Amazon launched “Health AI” on January 21, rolling out an agentic AI health assistant to all One Medical members. This isn’t a simple chatbot—it’s designed to take action: booking appointments, reading lab results, managing medications, and providing 24/7 personalized health guidance based on your medical records.
The technical stack runs on Amazon Bedrock’s large language models, with clinical protocols built in to identify when symptoms require escalation to a provider. Amazon claims Health AI can “analyze images,” though specifics on whether that means medical imaging or user-uploaded photos remain unclear.
According to Amazon, the key differentiator from competitors like ChatGPT Health and Claude for Healthcare is integration: Health AI accesses your existing One Medical records without requiring document uploads or external app connections. Conversations aren’t added to medical records, and Amazon says it follows HIPAA-compliant practices.
The timing is notable. Patient safety nonprofit ECRI just named misuse of AI chatbots as the top health technology hazard for 2026, citing cases where tools have suggested incorrect diagnoses, recommended unnecessary testing, and—in one memorable example—invented body parts. Amazon’s clinical protocol layer is clearly designed to address these concerns, but the risk of AI hallucinations in healthcare contexts remains a genuine safety question.
For One Medical members paying $99-199 annually (or $9/month as an Amazon Prime add-on), Health AI represents a significant value addition. For the healthcare industry, it’s another signal that big tech’s AI ambitions now extend directly into patient care.
The Bigger Picture: AI’s Agentic Moment
A thread connects all of this week’s AI news: the shift from models that answer questions to agents that take action. Amazon’s Health AI books your appointments. Claude’s Agent SDK lets developers build autonomous research tools. Musk’s Optimus robots perform factory tasks. Even Huang’s infrastructure framing assumes AI that operates independently across complex workflows.
As we covered in our Claude Agent SDK tutorial, the technical infrastructure for agentic AI is now accessible to individual developers. The question is no longer whether AI can act autonomously—it’s who benefits when it does.
The executives at Davos see infrastructure buildouts and productivity gains. The workers in survey data see job cuts and career uncertainty. The truth is probably somewhere in between: AI will create new categories of work while eliminating others, and the transition period will be messy.
What’s clear is that 2026 is the year AI moves from hype to deployment. Whether that deployment serves broad prosperity or concentrated gains depends on choices being made right now—in boardrooms, in policy discussions, and in the code being written by developers building the agentic future.
Get the Daily Pulse
Sharp analysis on what's actually moving in AI. No hype, no filler, no weekly digest.
