The tape doesn't lie. And right now, the tape is showing something ugly for Anthropic.
I've been sitting with the raw data from the last few weeks. The kind of data that doesn't make it into polished pitch decks. Gallup numbers showing anti-AI sentiment in America jumping from 42% opposition to 75% within a single year. Heatmap Pro's survey landing at 7 in 10 Americans opposing new AI data centers. We didn't see this coming at this velocity. Nobody did.
Now, I've watched the AI trade for years. I've audited the architecture. I've spoken with enough institutional allocators to know when they're nervous. And right now, the chatter isn't about the models. The chatter is about the pushback.
Anthropic wants to go public. Reports peg the valuation at nearly a trillion dollars. The IPO haul could reach two trillion. That's not a rounding error. That's the entire annual GDP of Australia. The annualized revenue run rate is already above $650 billion, according to recent numbers. The market is treating this like the next OpenAI, the next NVIDIA, the next everything.
But here's the thing about running fast: you don't always see the wall until you've hit it.
The wall here isn't another foundation model with better benchmarks. It's not a startup with a cooler interface. The wall is America itself, and what regular humans are feeling about the entire industry. It's not a technical problem. It's a sentiment problem. And that's a problem you can't fix with a 3am engineering sprint.
THE HUMAN PROBLEM
Let's get the basics out first. This public-facing number is genuinely astonishing. We're talking a $650B annualized revenue run rate for the company. That means their API business is enormous, likely B2B focused, and growing at a pace that would make any enterprise software company weep with envy. The public-facing Claude models are the tip of the spear, but the real money is in the API calls, the enterprise contracts, the government pilots, the startups building on their stack.
Then there's the other part of the narrative: the data centers. The absolute need for infrastructure at a scale we haven't seen before. The kind of infrastructure that raises property taxes, strains power grids, and makes residents nervous.
The problem isn't that the technology is bad. The problem is that the technology is visible. Real. It needs real-world real estate. It needs massive amounts of electricity, water, and community goodwill. And when the community doesn't want to be your neighbor, that's when the tape starts to look really different.
I've spent 24 years watching institutional narratives clash with individual realities. I've seen the "trust us, we're building the future" pitch fall flat when people realize the future includes a substation and cooling towers behind their kids' school. The NIMBY effect isn't just for suburban malls anymore. It's for the heart of the modern digital economy.
THE SENTIMENT SHIFT
What's happening is not a blip. Look at the Gallup data. 2022: 42% oppose. 2023: 75% oppose. That's not a natural fluctuation; that's a seismic shift in public consciousness. This is a five-alarm fire for every AI company.
The reasons are fairly obvious but often glossed over. The fear of job loss. Pew's surveys have consistently shown that over 70% of American adults believe AI will reduce employment opportunities. That's a huge, systemic worry. And it's directly hitting the basic human need for security. Not a niche worry.
Then there's the data center issue. Local communities aren't just worried about aesthetics. They're worried about the actual, real infrastructure and how it affects them. Water, electricity, noise, land use. That's why we're seeing local pushes against them. And that's where the recent executive orders in Pennsylvania and New York come in. These aren't just proposals. These are actual orders, actual policy shifts.
These are direct actions taken to limit the construction of new data centers. I've seen this pattern before with every major infrastructure project, from nuclear plants to pipelines. The public says no, the policy follows, and the project costs go through the roof.
That's a direct threat to the growth engine of every AI lab, but especially for Anthropic. Why? Because they're not a cloud company. They don't own the infrastructure in the same way Google does. They rely on third-party clouds—AWS, Azure, GCP—for their compute. This creates a critical vulnerability.
THE TECH BOX
Now, let's get technical. The public sentiment is not just a marketing issue. It's a technical roadblock. The technical demand is insatiable. Every frontier lab is in a brutal arms race to get more chips, more compute, more power.
The compute-power revenue link is direct. More compute means faster iterations. Faster iterations mean better models. Better models mean more customers. More customers mean more revenue. So if you can't build more data centers, that flywheel stops spinning.
Anthropic's Claude models, known for their long-context capabilities, are great, but they're notoriously compute-heavy. This means they need even more data center capacity to maintain their edge. If the data centers get blocked, the model improvements get blocked, and the revenue growth gets blocked.
It's a simple, brutal logic. And the market hasn't fully priced it in yet.
The public doesn't understand the tech, but they understand the cost. They know that these things need electricity. They know that the electricity is getting more expensive, and that their own bills are going up. They know that the water is being diverted. They know that the local grid is strained. And they don't like it.
THE OPPORTUNITY
Here's the counterintuitive bit. The anti-AI sentiment is actually a huge opportunity, not just a threat. It's an opportunity for the companies that are willing to adapt.
Anthropic, with its "Constitutional AI" safety narrative, has a shot at being the "responsible AI" champion. If they can position themselves as the ones who are building AI in a way that's sustainable, transparent, and community-focused, they can potentially differentiate themselves from the pack.
Think about it. They can become the company that says, "we're building the future, but we're not going to build it on your back." That's a massive brand position.
But there's a big "if" here. If they can't do that, then the "safety" narrative might become a liability. It might seem like they're admitting AI is dangerous, which is the exact thing people are afraid of. The public doesn't want a safer version of something they don't want. They want a different future altogether.
The companies that succeed are the ones that go around the NIMBY problem, not through it. They'll go to places with less regulation. They'll build more efficient models. They'll invest in better cooling tech. They'll partner with local communities and give them a tangible stake in the game.
But it's not just about tech. It's about narrative.
THE FINANCIAL MATH
Let's get the financial numbers straight. A trillion-dollar valuation on a $650 billion run-rate is a P/S ratio of roughly 15. In a high-growth tech market, that's not crazy. But it's demanding. It assumes this kind of growth continues for a long time without major interruptions.
But the recent executive orders are a direct counterweight to that assumption. It's a tangible policy risk that can increase costs, slow down deployment, and, if it gets bad enough, start to impact the growth rate. This is a physical, structural risk that isn't in the standard tech stock model.
Investors are asking the right questions. They're asking about data center slowdowns. They're asking about the impact of public sentiment. This is a sign that the market is starting to wake up to this risk. But it's not yet fully priced in.
This is a textbook "too big to fail, but too fragile to ignore" situation.
THE REGULATION GAME
The regulatory environment is a huge wildcard. The governor's orders in PA and NY are just the beginning. Expect to see more states follow. It's a local issue that goes viral and becomes national.
This is where things get dangerous. The initial response is often a complete stop. And we know how the "NIMBY" cycle works: new order, pushback, lawsuits, community meetings. The result is massive project delays. In some cases, it's completely stopped.
That's why the entire industry needs a better approach. They can't just fight it. They need to think about how to work with the local communities.
The smart play is to be proactive. Build a community benefit agreement from day one. Create a "golden share" for the local community. Give them a real financial stake in the project. This way, they're not just the neighbors. They're the shareholders.
THE DATA & THE DISCONNECT
The irony is that the public hates the data centers, but they love the AI services. They use AI assistants all the time. They don't realize that the AI is the same thing that's causing the data center to be built. They think they're separate.
This disconnect is the core of the problem. They want the benefits of the AI, but they don't want the physical cost. That's not a sustainable position. They're going to have to realize the price of the future. And the AI companies are going to have to show them what's in it for them, not just for the tech companies.
The public needs a direct answer to "what do I get?" Not "what do you get?" The answer can't just be "better AI." That's not good enough. It has to be more concrete. It has to be cheaper, better, faster health care, education, infrastructure, etc. That's the only way to turn the sentiment around.
THE REAL REASON FOR THE SLOWDOWN
And here's the thing that gets me, the thing that's not being talked about. The public is not just against the data centers. The public is against the idea that they're not in control. The "neighborhood effect" is just the trigger. The underlying cause is a feeling of powerlessness. A feeling that decisions are being made by big tech, in some distant city, that will fundamentally alter their lives, and they have no say.
That's a much bigger problem than any technical risk. It's the failure of trust.
So, the real question is: How do you build trust when the business model relies on the scale that scares people?
THE EXIT SIGNAL
For the IPO, the timing is interesting. Going public when sentiment is this negative is a bold move. It might signal confidence. Or it might signal a "sell into strength" move by early investors. The market will have to decide.
The "safety" angle could be the perfect story for a public company. It gives a sense of maturity. But the public's got a great radar for empty words. If they see the "safety" as just a marketing gimmick, while they're simultaneously fighting new data centers, the "safety" message will be met with extreme skepticism.
THE TAPE DOESN'T LIE
I keep saying it, but the tape doesn't lie. The tape is the market. The tape is the actual reactions of the people on the ground.
The tape is the 75% Gallup number. The tape is the governor's order in Pennsylvania. The tape is the investors' questions about the data center delays. The tape is the continuous stream of "next door" complaints from residents in the communities where they want to build.
This is the real story. And it's a story that can't be fixed by a code update.
The new currency in AI is no longer just the token count or the parameter count. It's the amount of public consent you can gather. It's the community's willingness to let you build. It's the political capital you have to overcome the objections. This is the "Human Compatibility" metric.
Anthropic's IPO is the test case for the entire industry. If they can pull it off, despite all this, it's a big win for the entire space. If they can't, it's going to be a bloodbath. And I'm not sure the market is fully prepared for that.
THE VERDICT
So, what's the real lesson here? The hard truth is that the future of AI is a physical infrastructure and a social contract. It's not just an algorithm.
It's a data center. It's a power grid. It's a water pipe. It's a set of building permits. It's the willingness of a community to say "yes."
We've been so busy building the virtual world that we forgot about the physical one. We have to build out the physical one, or the virtual one will never get off the ground. The physical world is the friction. The physical world is the bottleneck.
The market is finally waking up to the fact that the "AI race" is not just a race to a new model. It's a race to a new real estate. And the most expensive, hardest, and longest part of that process is the part that requires the consent of the people.
The "anti-AI" sentiment isn't a bug. It's a feature. It's the market saying, "We want this, but we want it done responsibly." It's a reality check. It's a demand for a different kind of innovation.
The next trillion-dollar companies aren't just the ones with the best code. They're the ones with the best relationships with the people who live next to their data centers.
THE TAKEAWAY
The question now isn't "can Anthropic beat OpenAI?" The question is "Can Anthropic get the approval of the local zoning board in rural Ohio?" The question is "Can they secure the power grid's capacity?" The question is "Can they make the case to the locals that this is a good thing?"
The model is already here. The risk is not in the model's intelligence. The risk is in the world around it.
We're going to see the next billion-dollar bets on the ability to build, not just the ability to compute.
The tape doesn't lie. The tape is the market. The tape is the people. And the people are asking, "What do I get out of this?"
And until we get a good answer, the tape is going to be scared.