A $2 trillion question hangs in the air: Can a company racing to build the world's most powerful intelligence survive a public that increasingly wants it stopped?
Seventy-five percent. That's the number that should terrify every AI executive, every venture capitalist, and every institutional investor preparing for what could be the largest technology IPO in history. In just twelve months, the percentage of Americans who oppose the construction of new AI data centers in their communities has surged from 42 percent to 75 percent, according to Heatmap Pro's latest tracking data. The Gallup numbers are no less damning: 71 percent of American adults now expect AI to eliminate jobs in their industries.
I've spent two decades watching markets react to sentiment shifts, but I've never seen a disconnect quite like this one. The public narrative around AI has curdled in real-time, and yet Anthropic—the safety-obsessed AI lab founded by former OpenAI researchers—is preparing to go public at a valuation approaching one trillion dollars, with early whispers suggesting a potential $2 trillion market debut.
This isn't a public relations problem. This is a balance sheet problem. And it's about to become the most visible test of whether "public sentiment" belongs in the same sentence as "risk factors" in an S-1 filing.
The market has spent years pricing AI as a pure technology story—compute, models, benchmarks, adoption curves. But the next phase of the AI trade is going to be priced on something far less quantifiable: whether ordinary people, in ordinary towns, will let the infrastructure get built at all.
The Infrastructure Backlash Nobody Priced In
Let me take you to Pennsylvania, where a governor's executive order has quietly become a template for a new kind of AI regulation. It doesn't target model outputs, content moderation, or algorithmic bias. It targets something far more fundamental: the physical footprint of AI itself.
Pennsylvania's executive order requires data center developers to submit comprehensive environmental impact assessments, prove energy efficiency standards, and demonstrate community benefit commitments before breaking ground. New York followed with similar measures. These aren't hostile bills designed to kill the industry—they're the opposite. They're designed to make AI infrastructure politically survivable. But the message to the market is unmistakable: data centers are no longer purely economic development projects; they're now contested public goods.
For Anthropic, this is not an abstract policy concern. The company doesn't own its own data centers. It relies on third-party cloud providers—AWS, Google Cloud, and others—to train and serve its Claude models. That means its entire commercial future depends on somebody else's ability to build and expand compute infrastructure in communities that increasingly don't want it.
I've been in the rooms where infrastructure deals get done, and I can tell you with absolute confidence: when local opposition becomes organized, costs rise exponentially. Not just in legal fees and delays, but in the soft costs—the community benefit agreements, the green energy premiums, the workforce development promises. Every one of those concessions is a line item that ultimately gets priced into AI tokens, API calls, and enterprise contracts.
The investors lining up for Anthropic's IPO understand this. The questions they're asking during roadshow presentations aren't about model architecture or benchmark scores. They're asking about data center construction delays, energy procurement contracts, and what happens when a third-party cloud provider can't get a permit in Ohio.
The "Safety" Paradox
Here's where the story gets genuinely interesting, and where I think most analysts are missing the plot.
Anthropic has built its entire brand identity around AI safety. The company was founded on a principle that the most powerful AI systems need to be developed with careful attention to alignment, interpretability, and harm reduction. Its Constitutional AI approach is designed to make Claude models more helpful, harmless, and honest than the alternatives.
You would think this would make Anthropic the natural choice for cautious institutions, for regulators, for a public that's worried about AI running amok. And in some ways, it does. But there's a dark side to this positioning that nobody on the sell-side seems willing to discuss publicly.
The more Anthropic emphasizes how powerful and potentially dangerous AI systems are, the more it validates the public's worst fears.
Think about it. If you're an ordinary person in rural Pennsylvania, and you read that a company is building a facility to support an AI system so powerful that its own creators are terrified of it—so powerful that they need a dedicated safety research team to keep it in check—are you more likely to welcome that facility into your community, or less?
The safety narrative that works beautifully for enterprise sales and institutional investors becomes a liability when it reaches Main Street. The polls are telling us something uncomfortable: the public doesn't distinguish between "AI that could destroy the world" and "AI that could take my job." They just know they don't want it in their backyard.
This creates a fundamental tension at the heart of Anthropic's IPO. The company's entire value proposition to investors is that it's the "responsible AI" bet—the one you can feel good about buying. But that same positioning may be amplifying the very sentiment that threatens its growth trajectory.
The Competitive Asymmetry
Let me take you through the competitive landscape, because the anti-AI sentiment doesn't hit every player equally.
Anthropic is structurally more exposed than its peers. It's a pure-play AI company with no cloud infrastructure of its own, no hardware division, no consumer ecosystem to diversify its revenue base. Everything flows through third-party compute. That's an elegant asset-light model during boom times, but it becomes a straitjacket when the physical infrastructure gets constrained.
OpenAI, by contrast, has Microsoft Azure's massive global footprint and its own dedicated compute agreements. When Azure runs into permitting issues in one region, there's always another region—another country, another cloud zone—to pick up the slack. The same applies to Google, which has both its own cloud infrastructure and proprietary TPUs that can be deployed in diverse locations.
Meta is perhaps the most interesting case. Its open-source Llama models can be deployed locally, on-premises, or through any cloud provider. The company doesn't need to build massive new data centers to serve its AI products to the world—at least not to the same degree as its closed-source competitors. That's a structural hedge against the "not in my backyard" problem.
The anti-AI sentiment, in other words, may accelerate the consolidation of AI compute around a handful of players with diversified infrastructure footprints. It could also push more developers toward open-source models that can run on existing hardware. Neither trend favors Anthropic's current strategy.
I've seen this movie before. During the early days of cloud computing, the companies that owned their own data centers—Amazon, Google, Microsoft—ultimately crushed the pure-play software companies that rented capacity from them. The economics were similar: the infrastructure owners could squeeze margins, control availability, and dictate terms. The AI industry is heading toward the same dynamic, and the anti-AI sentiment is only accelerating the timeline.
What This Means for the IPO
Let's get concrete about the numbers. Anthropic's annualized revenue run rate has reportedly crossed $650 billion—a staggering figure that suggests genuine product-market fit. At a trillion-dollar valuation, that implies a price-to-sales ratio of roughly 15x. In the current AI environment, that's not insane. But it leaves very little room for error.
Here's what the market is actually pricing: continued hypergrowth, expanding enterprise adoption, and—critically—the ability to scale compute capacity without catastrophic cost increases. The anti-AI sentiment threatens all three.
If data center construction slows down, Anthropic's ability to train larger models and serve more customers becomes constrained. That's not just a growth problem—it's a competitive problem. A model that can't be trained on enough compute becomes a worse model. A company that can't serve enough inference requests loses customers to competitors with spare capacity.
If regulatory costs rise, Anthropic's margin profile deteriorates. Every dollar spent on community benefits, green energy premiums, and compliance overhead is a dollar that doesn't go into R&D or customer acquisition. The company's "safety-first" positioning already carries a cost premium—its models are more expensive to train and run because of the rigorous evaluation and red-teaming processes. Adding an "emotion tax" on top of that is a double whammy.
And if public sentiment translates into slower enterprise adoption, the revenue growth story gets called into question. Companies that are worried about their customers' reactions to AI might delay deployments, reduce their AI budgets, or choose less controversial providers. The "safe" choice might become the "risky" choice in a different way.
The Case for Optimism
I'm not going to pretend this is a straightforward bearish story, because it isn't. There's a plausible path where Anthropic's "safety" brand becomes an asset rather than a liability—where the company successfully positions itself as the solution to anti-AI sentiment rather than a victim of it.
That path starts with a fundamental reframing. Instead of defending data center construction on economic grounds, Anthropic could lead the charge on sustainable AI infrastructure. Publish annual environmental impact reports. Commit to renewable energy sourcing. Partner with local communities rather than negotiating with them. Build a template for "good neighbor" AI infrastructure that other companies would be forced to follow.
There's also a technical angle that doesn't get enough attention. If Anthropic can achieve meaningful breakthroughs in model efficiency—training stronger models with fewer FLOPs, or deploying quantization and distillation techniques that reduce inference costs—it could partially decouple its growth trajectory from data center construction. A company that can deliver frontier AI with 30% less compute is a company that's 30% less exposed to the permitting problem.
The contrarian investment thesis is that anti-AI sentiment becomes a moat. If it's genuinely harder to build AI infrastructure in the future, then the companies that already have compute access—locked in through long-term contracts with cloud providers—have a durable advantage over newcomers. Anthropic's existing agreements with AWS and Google Cloud might be worth more than anyone realizes.
The Real Question
Here's what I'm watching as this IPO takes shape. The S-1 filing will list "risk factors"—the mandatory disclosures that lawyers craft to protect companies from shareholder lawsuits. The question isn't whether anti-AI sentiment appears in that list. It will. The question is how it's described.
If Anthropic frames it as a manageable operational challenge—something that can be addressed through stakeholder engagement, green energy commitments, and community partnerships—that tells you the company has a real strategy. If it's buried in generic language about "regulatory changes" and "market conditions," that tells you they don't know what to do about it.
The other signal I'm tracking is what OpenAI, Google, and Meta say in their own filings and investor communications over the next few quarters. If they all start prominently discussing public sentiment as a risk factor, we're looking at an industry-wide problem. If it remains Anthropic-specific, that tells you something about the relative strength of their positioning.
The uncomfortable truth is that the AI industry has spent the last decade building technology without building consent. The public never agreed to this future. They were never asked whether they wanted AI systems that could write code, diagnose diseases, and replace entire job categories. The industry simply assumed that if the technology was impressive enough, the acceptance would follow automatically.
That assumption is now being tested at the most fundamental level—not in the pages of academic papers or the boardrooms of Fortune 500 companies, but in the zoning hearings, town halls, and public utility commissions where the physical future of AI will be decided.
I've seen companies go public with brilliant technology and terrible timing. I've seen valuations that made no sense except through the lens of collective mania. And I've seen what happens when an industry's foundational assumptions turn out to be wrong.
The anti-AI sentiment isn't a bug in the market's pricing mechanism. It's a signal about the social license to operate—the invisible but critical asset that every technology company needs to survive. Anthropic's IPO will be the first major test of whether that asset can be valued, managed, and defended.
The public has been telling us what they think for over a year now. The question is whether anyone in the AI industry was listening.
I'll be watching the S-1 with more attention than I've given any filing in my career. The answer is going to tell us not just about Anthropic's future, but about whether the AI trade—the biggest market story of this decade—has a sustainable foundation or is building on sand.