On Monday morning in Davos, nearly 3,000 participants from 130 countries gathered to discuss what the World Economic Forum calls “The Spirit of Dialogue.” OpenAI’s Chief Global Affairs Officer Chris Lehane took the stage to warn about a “capability overhang” – the massive gap between what AI can do and what the world has figured out how to extract from it. The message was urgent, the setting was spectacular, and the champagne was excellent.
But here’s the uncomfortable truth nobody in the Swiss Alps wants to admit: the cavalry isn’t coming. Governments won’t regulate AI in time. Tech companies are already too big to fail. And I’ll say the quiet part out loud: we’re in a de facto “AI Cold War” where each side must “win” (however “winning” is being defined). If AI governance happens at all, it will happen in the one place where both technology companies and political leaders must still ask permission: the corporate boardroom.
The Math That Broke Government Regulation
Consider the scale of what we’re dealing with. According to CSIS analysis, five U.S. companies – Alphabet, Amazon, Meta, Microsoft and Oracle – will spend more than $450 billion on AI-specific capital expenditures in 2026 alone. That’s compared to the $326 billion (inflation-adjusted) the U.S. government spent on the entire Apollo program over thirteen years. OpenAI, Anthropic and xAI will add hundreds of billions more.
When Sam Altman publicly stated OpenAI could commit $1.4 trillion in AI spending, and then briefly suggested the government might need to act as “insurer of last resort,” he revealed the structural reality: these companies have become systemically important. As AI frontier labs like Anthropic and OpenAI prepare to go public (with speculation that OpenAI’s IPO could target a trillion-dollar valuation), critics worry we are witnessing the rise of “too big to fail” AI monopolies.
The IMF confirmed this systemic risk today. In its January 2026 World Economic Outlook, the NGO warned that “global growth has been impressively resilient amid trade disruptions, but this masks underlying fragilities tied to the concentration of investment in the tech sector.” The numbers tell the story: IT investment as a share of U.S. economic output has surged to its highest level since 2001. U.S. market capitalization relative to GDP has climbed from 132 percent during the dot-com peak to 226 percent today, a 71 percent increase that makes the current AI boom potentially more consequential than its predecessor.
McKinsey estimates over $6.7 trillion may be spent on AI and computing infrastructure by 2030. That’s an investment scale usually seen in wartime. No regulatory framework in history has successfully constrained capital deployment at this magnitude while it was happening.
The AI Cold War Nobody Can Win (But Everyone Must Fight)
The geopolitical dimension makes government action even more unlikely. Both Washington and Beijing treat AI as an existential competition, often comparing it to the Cold War space race given its military applications and capacity to transform nearly every domain of national policy.
The Atlantic Council identifies eight ways AI will shape geopolitics in 2026, noting that countries are now choosing between American and Chinese AI tech stacks. China’s “Delete A” project aims to remove American technology from Chinese supply chains entirely. Meanwhile, U.S. export controls on advanced chips have created a bifurcated global technology ecosystem.
The result is what I call an “AI Cold War” characterized by “two parallel digital ecosystems” that may never be reconciled. In this environment, meaningful international AI regulation becomes impossible. The European Union pushes a rights-based regulatory model. The United States favors voluntary standards to preserve innovation and security flexibility. China promotes cooperation while defending state control over data and AI deployment.
Each bloc fears that regulation means unilateral disarmament. So none of them regulate meaningfully.
The Human Cost Nobody Wants To Quantify
While the Davos echo chamber is full of panels, dinners and selfies, the WEF’s own data reveals what’s actually coming. The Future of Jobs Report projects 92 million existing jobs face displacement by 2030, with 170 million new roles emerging resulting in a net positive of 78 million jobs, but only for those who can navigate the transition.
70% of the skills needed for the average job are expected to change by 2030. Yet 55% of employees received no workplace training in the past year. Nearly 40% of current workforce skills will become obsolete within five years. We are building the plane while it’s crashing.
The IMF’s warning reinforces this urgency. “AI’s uneven impact on workers is another important consideration,” the fund noted. “While innovation drives growth, it risks displacing jobs and depressing wages for certain segments of the workforce.” This isn’t World Economic Forum speculation, but rather the institution responsible for global financial stability acknowledging that concentrated AI investment creates systemic workforce risks.
The WEF’s “Four Futures for Jobs” report maps what happens next. In the “Age of Displacement” scenario, exponential AI advancement outpaces workforce adaptation. Businesses race to automate in lieu of scarce talent. Unemployment spikes. Consumer confidence erodes. Governments face mounting societal instability. This isn’t speculation. It’s one of four plausible scenarios developed by the organization hosting the very summit that claims to address it.
Why A Tech Correction Would Ripple Everywhere
The IMF’s analysis reveals why boards can’t treat AI governance as a technology issue alone. In fact, it may be a financial stability issue. The fund identified three reasons why AI-related market vulnerability may exceed the dot-com era despite more modest price-earnings ratios.
1) Rising stock prices have been driven predominantly by technology stocks, particularly AI-related names, making a narrow group the major driver of index performance.
2) Many critical AI firms remain private, meaning their debt borrowings could create consequences unseen during the dot-com crash.
3) With market capitalization now at 226 percent of GDP versus 132 percent in 2001, even a more modest correction would have substantially larger effects on consumption and economic activity.
The IMF’s October 2025 scenario analysis showed that a moderate correction in AI stock valuations, combined with tightening financial conditions, would reduce global growth by 0.4 percent. “Given the decade-long increase in foreign ownership of US equities,” the fund warned, “this sharp correction could also trigger sizable wealth losses outside the United States.” Every board approving AI investments is now operating in this interconnected risk environment.
Why The Boardroom Becomes The Last Line Of Defense
Here’s what Davos conversations consistently miss: every technology company must eventually sell to enterprises. Every enterprise has a board. And boards have fiduciary duties that transcend quarterly earnings, including the duty of care and the duty of loyalty to shareholders and stakeholders.
When tech leaders speak at Davos about the promise and potential of AI and the need for adoption, many are essentially making a sales pitch dressed as policy discussion. When political leaders attend, they need campaign donations and the support of business. In China, government control substitutes for independent board governance. In the West, the boardroom represents the last institution where technology deployment requires formal approval from humans with legal accountability.
Directors and executives who approve and deploy AI that undermines workforces, violate emerging regulations, or create systemic risks face personal liability. Unlike legislators who face voters every few years, board members face shareholders every day and plaintiffs every time something goes wrong. That’s not a perfect governance mechanism, but it’s the only one with teeth.
Where The Real Conversations Will Happen
The WEF’s Borge Brende captured the problem perfectly: “The pace of technological change is accelerating faster than the institutions that are meant to govern it.”
While Davos generates headlines, the substantive work on AI governance increasingly happens elsewhere. In late February, the International Association for Safe and Ethical AI will convene its second annual conference at UNESCO House in Paris. IASEAI’26, chaired by one of the “Godfathers of AI” Dr. Stuart Russell and featuring an interdisciplinary community of researchers, policymakers and practitioners, anticipates 1,000 to 1,500 attendees focused on ensuring AI technologies are safe, ethical and beneficial.
Unlike Davos, IASEAI explicitly addresses agentic and multi-agent safety, AI in warfare, control and containment mechanisms, and the future of work and education. It’s invitation-only, but designed to produce actionable outcomes rather than photo opportunities. The organization emerged from the 2023 AI Safety Summit in the United Kingdom with a steering committee including Yoshua Bengio, Max Tegmark and Kate Crawford, all researchers with no “products” to sell.
The 90-Day Boardroom Agenda
For directors and executives ready to act rather than wait, here’s what the next quarter should look like.
Days 1-30: Establish AI as a standing board agenda item. Request a comprehensive inventory of all AI systems currently deployed or under consideration. Identify which decisions are being made by algorithms rather than humans, and who bears accountability when those decisions cause harm. The IMF’s warning about “underlying fragilities” applies to your company specifically – you need to know where they are.
Days 31-60: Demand workforce impact assessments before approving any AI deployment. The 78 million net new jobs won’t create themselves. For every efficiency proposal, require a parallel reskilling plan. Calculate not just the cost savings from automation, but the customer base erosion when displaced workers stop buying. Companies that automate without reskilling are destroying demand alongside costs – basic economics that quarterly earnings calls ignore.
Days 61-90: Engage with emerging governance frameworks before they’re imposed. The EU’s AI Act is already in force. State regulations are proliferating. International standards are forming. Assign board-level responsibility for tracking these developments and participating in shaping them. Companies that help write the rules fare better than those who discover them through enforcement actions.
The Davos Paradox
The irony of Davos 2026 is that it brings together the precise people with the power to address AI governance, yet gives them a format designed to avoid binding commitments. Panels replace policies. Pledges replace accountability. And another year passes while the capability gap widens, the talent gap deepens, and the governance gap becomes unbridgeable.
At CES, AMD’s Lisa Su said active AI users will hit 5 billion by 2030, up from ~1 billion today. This means that AI will touch more than half the world in the next few years. AI has already achieved escape velocity. The question isn’t whether to govern AI, it’s whether governance can happen before the decisions become irreversible.
Governments won’t do it in time. International bodies can’t do it at all. That leaves the boardroom: imperfect, conflicted, but ultimately accountable to someone.
The conversations that matter in 2026 won’t happen in Davos. They’ll happen in your next board meeting.
The only question is whether you’ll be ready to have them.
