This week, AI hit an inflection point. Public audiences started booing the hype. Infrastructure spending accelerated past anything we've seen. And the race to own the next technical frontier — world models — just received its clearest signal yet. The euphoria phase is ending. What follows is realer, messier, and more consequential.
This week in AI (May 29, 2026) was defined by three converging forces: a genuine public backlash at commencement speeches and in polling, Runway raising $315M at a $5.3B valuation to build world models rather than improve video generation, and ByteDance planning up to $70B in compute spending while JPMorgan formalised AI as core infrastructure. The gap between public sentiment and private capital deployment has never been wider — and it's reshaping how AI will be built, regulated, and received over the next two years.
Here's what happened this week — and what it means.
1 The AI Backlash Is Real — and It's Affecting Infrastructure
This week, something shifted in public perception of AI. At university commencement speeches across the U.S., former Google CEO Eric Schmidt and other tech executives praised AI's potential — and were met with clear, organised booing from graduating classes. This wasn't scattered discontent. It was a signal.
The data backs it up. Gallup and Economist/YouGov polling found low optimism among younger people and bipartisan concern about AI's speed, job impacts, environmental costs, and concentration of wealth. These are not fringe positions. They represent the mainstream mood among the people who will actually build, use, and live with AI over the next decade.
What People Are Actually Worried About
- Jobs disappearing in real time. Companies are using AI to cut headcount. The announcements are happening publicly and frequently.
- Compute eating power. Data-centre expansion is straining electrical grids. Communities that have to live near this infrastructure are resisting it.
- Wealth concentrating upward. A handful of labs control frontier models. A handful of companies control compute. Benefits flow up; costs flow down.
- Speed with no clarity on destination. Even optimists can't articulate what AI looks like in three years. That uncertainty is unsettling, not exciting.
The backlash is already affecting infrastructure expansion. Communities resisting data-centre construction could constrain compute capacity. If public sentiment hardens against AI companies, regulation accelerates. The growth narrative is no longer unstoppable — this is the moment where capability and trust need to converge, not diverge further.
2 Runway's $5.3B Bet — From Video to World Models
Runway, the AI video generation company, raised $315 million in a Series E round at a $5.3 billion valuation — nearly double its prior valuation. The critical detail isn't the money. It's what they're raising it for.
Runway's CEO Cristóbal Valenzuela explicitly called AI video a "prequel" to world models. The company announced it is moving beyond short video clips toward persistent, interactive simulated worlds. They're not improving what they already have — they're building the next layer entirely.
What Are World Models?
World models are AI systems designed to construct internal representations of environments so they can simulate and plan for future scenarios. They understand physics. They understand causality. They can reason about what happens when objects move through space — something LLMs fundamentally cannot do.
Robotics that can plan movements and predict consequences. Gaming with worlds that respond realistically. Virtual production for film studios simulating scenes before shooting. Engineering that simulates designs before building. Medicine that models how drugs interact with biological systems. The labs that build capable world models will own the next generation of automation — not just software.
Runway has earned credibility here — they've outperformed Google and OpenAI on several video generation benchmarks. The $5.3B valuation reflects that technical lead, and their ambition to capture early-stage value in the world-model stack before it becomes a crowded race.
Companies that can model reality itself are being priced as strategic infrastructure rather than tools.
— NeeAr Ventures Editorial Analysis, May 2026
3 Enterprise AI: From Experiment to Core Business
JPMorgan Chase made a significant structural move this week: the bank formally reclassified its AI investments from experimental R&D to core infrastructure. Their 2026 technology budget is approximately $19.8 billion, with 2,000 staff dedicated to AI development across three focus areas — boosting internal productivity through AI agents, hardening cybersecurity defences, and personalising retail banking.
This is significant. JPMorgan isn't piloting AI anymore. They're betting their infrastructure on it. When the world's largest bank by market cap makes that reclassification, it sends an unambiguous message to the market: AI is foundational, not experimental. Deployment is the next frontier. Scale requires specialisation.
Meanwhile: Telegram Embeds AI into Messaging
At the consumer level, Telegram introduced assistant bots capable of reading, filtering, and replying to messages based on user-defined permissions. The platform is no longer positioning AI as a chatbot — it's an assistance layer directly integrated into everyday conversations.
Every major messaging platform — WhatsApp, Signal, Discord — is moving the same direction. AI isn't a separate tool you open in a different tab. It's a layer inside the tools people already live in. The shift from "AI app" to "AI layer" is accelerating across both enterprise and consumer contexts simultaneously.
4 The Great Infrastructure Race: ByteDance, the U.S., and Policy Signals
While U.S. AI infrastructure expansion faces resistance from environmental concerns, power grid constraints, and community opposition, China is doubling down at scale. ByteDance is discussing capital expenditure of up to $70 billion in 2026 to build out data centres and AI infrastructure — underwritten by approximately $50 billion in profit from 2025.
To be clear about what that number means: a single Chinese company is planning to spend more on AI infrastructure in one year than most countries invest in their entire technology sectors. And they're using their own profits to do it, not capital raises.
If ByteDance succeeds in building efficient, low-cost AI systems on domestic hardware, they won't need to licence GPT or Claude. They'll have their own frontier models, trained on their own compute, using their own chips. The U.S. is constrained by power grids and regulation. China is building at industrial velocity. The compute divide is widening, not narrowing.
Quick Hits: Three More Signals
Beyond the four main stories, three smaller developments caught our attention this week:
- Clawdmeter goes mainstream. Developers are using an open-source gadget that displays Claude Code usage in real time. Token consumption is becoming a performance metric — a sign that AI has shifted from being a tool to being a work environment.
- AI judges face legal liability. A landmark Northern District of California ruling found that when a platform's AI exercises "ultimate authority" over assembled ad content, the platform may be liable for fraudulent statements under securities law. AI systems that make decisions now carry legal weight.
- Apple's AI-enabled AirPods. Apple is approaching early mass-production testing for new AirPods equipped with low-resolution cameras for visual context. Always-on, context-aware AI is coming to your ear — not just your screen.
On the policy front: the Trump administration is preparing an executive order focused on cybersecurity and frontier-model oversight. The emphasis is on security and governance of the most capable systems — not innovation incentives. That's a meaningful pivot in framing.
What This Week Tells Us
Three signals that matter heading into the next chapter of AI:
- The euphoria phase is ending. Audiences are skeptical. Communities are resisting data centres. The "AI is the future" narrative now hits real friction. That's not a reason to slow down — it's a reason to build more carefully, with more transparency, and with genuine attention to who bears the costs.
- World models are the next frontier. After 18 months of LLM dominance, the race has shifted toward systems that understand physical space. Runway's valuation jump shows where capital is moving — and where the next generation of automation capability will come from.
- The infrastructure divide is widening. The U.S. is constrained by power grids, environmental concerns, and emerging regulation. China is spending $70B to own compute. The implications for long-term AI leadership are enormous and under-discussed.
AI didn't get easier this week. It got realer. The hype faded. The infrastructure race intensified. The backlash began. This is what happens when a technology crosses from experimental to essential — and the world has to decide what kind of essential it wants it to be.
Stay curious. See you next week.
At U.S. university commencement speeches in May 2026, former Google CEO Eric Schmidt and other tech executives praised AI's potential and were met with organised booing from graduating classes. Gallup and Economist/YouGov polling show low optimism among younger people and bipartisan concern about AI's speed, job impacts, environmental costs, and concentration of wealth. The backlash reflects a growing disconnect between the pace of AI deployment and the lived experience of people worried about job losses, strained power grids, and wealth concentration among a handful of labs and companies.
World models are AI systems designed to construct internal representations of environments so they can simulate and plan for future scenarios — understanding physics, causality, and what comes next in physical space. Unlike LLMs, which understand language, world models can reason about 3D environments, enabling robotics, gaming, virtual production, engineering simulation, and medical modelling. Runway raised $315 million at a $5.3 billion valuation in May 2026 specifically to build world models, calling AI video generation a "prequel" to this next frontier. Industry analysts increasingly see world models as the next multi-billion-dollar layer beyond LLMs.
JPMorgan Chase formally reclassified its AI investments from experimental R&D to core infrastructure in 2026, committing a $19.8 billion technology budget and 2,000 dedicated AI staff. The bank is focusing on three areas: boosting internal productivity through AI agents, hardening cybersecurity defences, and personalising retail banking. This move signals that AI is no longer optional for large enterprises — it is foundational, and deployment at scale is now the frontier, not just model research.
ByteDance, the company behind TikTok, is discussing capital expenditure of up to $70 billion in 2026 to build out data centres and AI infrastructure, funded by its roughly $50 billion in profit from 2025. This dwarfs most Western AI infrastructure commitments and reflects China's strategy to own compute at industrial scale. If ByteDance succeeds in building efficient, low-cost AI systems on domestic hardware, it would not need to licence frontier models from U.S. labs, fundamentally changing the competitive landscape.
The Trump administration is preparing an executive order focused on cybersecurity and frontier-model oversight following growing concerns about highly capable AI systems. The emphasis is on security and governance of advanced models, not innovation incentives. This signals a shift in U.S. AI policy from primarily growth-oriented framing toward risk management for the most powerful frontier systems.