Everything About AI
NextFin News - Anthropic is preparing for what could be the largest initial public offering in history, and its prospectus is expected to carry a warning unlike almost any in corporate history: the public may hate what the company sells enough to stop it from building the infrastructure it needs to survive.
The Claude creator, valued at $965 billion after a $65 billion funding round in May, confidentially filed IPO paperwork with the Securities and Exchange Commission in June, beating rival OpenAI to the starting line. Investors have been told to expect a public listing that could value the five-year-old company at as much as $2 trillion. But according to people familiar with the matter, the filing is expected to name a risk that cuts against the very premise of the AI boom: mounting public backlash against artificial intelligence and the data centers that power it.
That disclosure lands at an awkward moment. On September 12, co-founder and CEO Dario Amodei published an essay calling on the AI industry to "slow the pace" of model development — a stance that reads as principled safety advocacy to supporters and as a ladder-pulling exercise to skeptics. The tension between the two positions is the story: Anthropic is asking public-market investors to pay a record price for a company whose own filing warns that voters, communities, and politicians may try to slow it down.
The Numbers Behind the Warning
The scale of the offering is what makes the risk factor matter. Anthropic's annualized revenue run rate topped $65 billion by the end of July, more than seven times its pace a year earlier, according to people familiar with the company's investor updates. In May, the company reported a $47 billion run rate; it has projected 2028 revenue of roughly $190 billion to $200 billion, a forecast that underpins the $2 trillion valuation some investors are discussing.
For context, SpaceX — which went public in June in what remains the largest IPO on record, raising $85.7 billion including the underwriter option — is now valued at about $2 trillion. Anthropic's backers are betting the AI lab can match or beat that debut.
But the expected disclosure points to a vulnerability that no amount of revenue growth can easily fix. Companies are required to outline material risks in their prospectuses both to inform investors and to shield themselves from liability. SpaceX's filing carried the usual macroeconomic and geopolitical warnings. Anthropic's, by contrast, is expected to flag something more existential: that Americans increasingly oppose AI itself, and are taking that opposition to city halls, state legislatures, and the ballot box.
The evidence is not hard to find. A Gallup poll conducted in March 2026 with 1,000 U.S. adults found that seven in 10 Americans oppose constructing AI data centers in their local area, including 48% who are strongly opposed; barely a quarter favor such projects, and only 7% are strongly in favor. Ballotpedia tracked at least 16 data-center-related ballot measures in 2026, and of the four decided by late August, each produced more limits on development, not fewer. Opposition cuts across party lines and centers on power consumption, water use, and local control — tangible local costs rather than abstract fears about superintelligence.
The legal language is already being written elsewhere in the AI complex. SB Energy, a data-center developer backed by SoftBank, OpenAI, and Nvidia, stated in its September 1 registration statement that "growing public skepticism and resistance to AI" — including concerns about employment, privacy, safety, and "broader societal implications" — "could reduce consumer and enterprise willingness to adopt AI technologies and therefore reduce demand for the data center infrastructure that supports them." The filing also warned of "community opposition, local moratoria, and hyper-local dissent." Anthropic's own disclosure is expected to run along similar lines, according to the people familiar with the matter.
The mechanism is direct. Compute capacity is directly correlated to revenue for AI labs like Anthropic: more data centers mean more model training and more inference served to customers. Any delay in construction limits available computing resources, slows product delivery, or raises operating costs. A risk factor that threatens the buildout therefore threatens the growth curve that justifies the valuation.
The Transmission Channel Is Real, Not Boilerplate
Most IPO risk factors are legal padding: generic warnings about competition, regulation, and macroeconomic conditions that investors skim and ignore. Anthropic's expected disclosure is different because it identifies a specific transmission channel from public opinion to revenue.
The chain runs: voters oppose data centers → local moratoria and ballot measures pass → permitting and construction slow → compute capacity fails to expand on schedule → model improvement and inference volume stall → revenue growth decelerates. Each link is observable. The Gallup opposition number is not a sentiment abstraction; it translates into city-council votes. The ballot-measure trend is not hypothetical; four measures have already produced restrictions. And the capital-intensity link is mechanical: AI labs cannot serve inference they do not have the hardware to run.
The infrastructure pipeline data reinforces the point. One industry tally counted plans for 3,969 new U.S. data centers on top of 5,427 existing facilities reported at the end of last year — yet only 802 of those planned facilities are currently under construction. Goldman Sachs has estimated that only about half of the AI compute capacity scheduled to activate between now and 2028 will come online by its target date. U.S. data center construction spending reached $68.3 billion in June, up 7% from the previous month and 46% from a year earlier — growing, but not fast enough to match the announced pipeline.
This is where the disclosure bites. A company valued at $2 trillion on a projection of $190 billion to $200 billion in 2028 revenue is underwriting uninterrupted exponential growth. The risk factor is an admission that the physical and political infrastructure required for that growth is becoming harder to secure. Planned compute is not the same as available compute.
Cyclical Backlash or Structural Constraint?
The central question for investors is whether the backlash is cyclical — a temporary wave of anxiety that will recede as the technology proves its usefulness — or structural, a durable feature of the political landscape that will not self-correct.
The evidence points to structural. Three observations support that call. First, opposition is geographically broad and politically cross-cutting; it is not concentrated in a few activist enclaves that can be isolated or outspent. Second, the opposition is about tangible local costs — electricity, water, land use — rather than transient sentiment, and tangible costs produce durable political coalitions. Third, the backlash is institutionalizing: ballot measures, state moratoria, and prospectus disclosures themselves create a feedback loop in which every new filing normalizes the risk and arms the next local campaign with language from the securities document.
A cyclical read would require evidence that opposition is softening as communities observe benefits, or that state preemption laws are shielding developers. The opposite is happening. The feedback loop runs the other way.
That said, the cyclical leg exists and should not be conflated with the structural one. In the short run, the intensity of backlash is tied to the pace of the buildout itself — a slower construction cycle would reduce visible friction. And history offers some comfort: previous technology backlashes, from genetically modified crops to 5G networks, proved containable, and the AI industry has far more political capital than those predecessors. But those analogies cut both ways: 5G opposition slowed deployment for years without ever eliminating the technology, which is precisely the kind of drag that matters for a company underwriting uninterrupted growth.
Broader sentiment data points in the same direction. A Pew Research Center survey conducted in June 2026 found that 52% of Americans say they are more concerned than excited about the increased use of AI in daily life, up from 37% in 2021; only 9% are more excited than concerned. For the first time, a majority of adults under 30 — 55% — say they are more concerned than excited. The constituency that will inherit the technology is also the constituency most skeptical of its pace.
The Second-Order Problem: A Slowdown Plea From a Company Racing to Sell
Amodei's September 12 essay adds a second-order complication that no risk factor can fully capture. The CEO is publicly arguing that the industry should slow the pace of AI development at the very moment his company is preparing to monetize acceleration. The market reaction among observers has been mixed.
Some investors frame the positioning as rational risk management.
I don't know that investors are necessarily going to see it as a negative. Unless the companies are genuine and say, 'OK, we're not going to IPO, we're not going to use any more compute, we're not going to train any more models.' That's not what they're saying.
That was Gil Luria, an equity analyst at D.A. Davidson. Others see a competitive moat being built out of regulation. "That could actually favor Anthropic and OpenAI if smaller competitors cannot afford the rigorous safety, evaluation and security investments required for frontier-level models," said Arun Chandrasekaran, an analyst at Gartner. Luria put the point more sharply, saying Anthropic and OpenAI appear to be engaging in "monopolistic behavior" and that the slowdown push "feels more and more like a ladder pull."
The second-order effect is this: if the slowdown framing succeeds, it does not reduce Anthropic's risk — it socializes it. Stricter standards raise the fixed cost of competing at the frontier, which protects incumbents' margins while doing nothing to guarantee that the underlying demand for compute will be permitted. The risk factor and the essay point in the same direction: growth will be harder, slower, and more politically contested than the valuation assumes.
The Concentration Risk Behind the Risk Factor
The backlash disclosure is not the only structural tension in the filing. Anthropic's business sits inside a three-way dependency that the prospectus will also have to confront: Amazon, Google, and Microsoft are simultaneously the company's largest equity investors, its primary cloud infrastructure providers, and its main distribution channels through marketplaces such as AWS Bedrock and Google Cloud Vertex AI.
That entanglement is a strength until it is not. The hyperscalers give Anthropic capital, compute, and customers at a scale no standalone lab could match. But they also operate rival model franchises, and each has an incentive to prioritize its own offerings when pricing, placement, or capacity allocation comes under pressure. If public opposition raises the value of already-permitted capacity, the infrastructure owners capture more of the surplus — and the model lab's margin absorbs the squeeze.
The SB Energy precedent shows how explicitly this risk is now being drafted. The data-center developer's filing does not merely gesture at "regulatory uncertainty"; it names AI skepticism as a demand-side threat. For Anthropic, the two risks compound: the same public opposition that slows data-center construction also raises the bargaining power of the few partners that control permitted sites and grid connections.
The Strongest Counter-Thesis — and Why It Only Partly Holds
The bullish case is not weak, and it deserves its due. The market knows how to price risk, and investors have shown an enormous appetite for AI exposure. Brad Gerstner, chief executive of Altimeter Capital and an investor in both Anthropic and OpenAI, wrote on X that "the market knows how to price risk - see SpaceX. There is huge appetite to invest in the AI leaders." Gene Munster of Deepwater Asset Management predicted that "nothing will change and the AI leapfrog game will continue," adding that "AI's long-term opportunity is too big for them to slow down" and that the safety comments were "motivated to reduce the regulatory pressure."
The counter-thesis has real force: demand for AI is genuine and growing, enterprise adoption of tools like Claude Code is expanding, and capital is abundant. A $2 trillion valuation is not absurd if Anthropic captures even a fraction of the productivity gains it is selling.
But the counter-thesis answers the wrong question. Whether AI demand is real is not in dispute. What the risk factor highlights is not demand but supply — the physical and political capacity to build the infrastructure that demand requires. Gerstner's SpaceX comparison is instructive and not entirely reassuring: SpaceX went public at a valuation that priced in execution risk, and its shares have traded on the company's ability to actually launch rockets. Anthropic's risk factor is an acknowledgment that some of the launchpads may not get built.
The bet here is on the long term — now with tempering thoughts about the control of the technology. The dramatic growth and possibilities of these companies, now more publicly coupled with the potential very serious concerns and risks, will likely persist whether the IPO happens in Q4 or next year or whenever.
That was Lise Buyer, a partner at IPO advisory firm Class V Group, capturing the balance between the growth story and the newly acknowledged risk.
What Would Prove the Bearish Read Wrong
The falsifying signal is specific and observable. If, over the next two election cycles, more than half of the remaining 2026 data-center ballot measures are defeated or produce permissive language, and if state preemption laws shield developers in the states where Anthropic's partners are building, then the backlash is cyclical noise rather than a structural constraint — and the $2 trillion underwriting holds. The second test is operational: if Anthropic's disclosed compute-capacity additions track its revenue guidance through 2027, with at least half of the capacity scheduled through 2028 coming online on time, the supply-side argument collapses.
Conversely, if restrictive ballot measures continue to pass at the rate seen so far in 2026 and the company's compute additions fall behind guidance, the risk factor will have been the understated sentence in the filing.
Who Benefits, Who Is Exposed, and What to Watch
The immediate beneficiaries of this dynamic are the companies one layer up the stack from the model labs: the hyperscalers and power developers that control the physical infrastructure. Amazon, Google, and Microsoft are simultaneously Anthropic's largest investors, its cloud providers, and its distribution channels — a concentration that the prospectus will also have to address. If permitting friction raises the value of already-permitted capacity, those partners capture more of the surplus.
The exposed party is the equity holder in the model lab itself. A valuation built on uninterrupted exponential growth is fragile to anything that inserts friction into the buildout, and the risk factor is a formal acknowledgment that friction exists. Debt holders are less exposed than equity holders, which helps explain reports that Anthropic's bankers are also pushing for top-tier credit ratings after a listing — a structure that would let early investors monetize while the company retains flexibility.
Split by time horizon:
- Short term (sentiment and timing): The IPO itself may be delayed or repriced. People familiar with the matter said in early September that Anthropic pushed its marketing push to mid-October, with the public prospectus expected in late September — a slip that investors read as either prudent preparation or a sign that the numbers need more polishing.
- Medium term (fundamentals): Revenue will track compute additions, not model quality. Watch the company's disclosed capacity buildout against its guidance; any divergence is the first hard signal that the risk factor is materializing.
- Long term (structural): If the backlash institutionalizes through state preemption battles and federal safety rules, the industry consolidates around the players that can afford compliance and permitted sites. Anthropic and OpenAI may emerge stronger relative to competitors, even as the sector's overall growth rate slows.
Three scenarios frame the path:
- Base case: The IPO prices near the lower end of the $1.5 trillion to $2 trillion range, the risk factor is absorbed as a known discount, and growth continues but at a slower, more contested pace.
- Upside case: Public opposition proves localized and surmountable; state preemption shields developers; compute additions meet guidance; the stock rerates toward the full $2 trillion thesis.
- Downside case: Ballot measures proliferate through the 2026 midterms, permitting delays compound, compute additions miss guidance, and the growth multiple compresses sharply.
The specific signals to watch are the 2026 midterm results in states with data-center ballot measures, the final text of Anthropic's public prospectus, and the company's first disclosed compute-capacity additions against its revenue guidance. If core PCE is the tell for inflation, permitted gigawatts are the tell for AI.
Anthropic's prospectus asks investors to underwrite exponential growth while warning them that the public may not permit the infrastructure exponential growth requires — and that contradiction, not the technology, is the mother of all risk factors.
Data current as of September 16, 2026. All figures sourced from company disclosures, SEC filings, government surveys, and market-data providers; the IPO prospectus referenced has not yet been made public.
Explore more exclusive insights at nextfin.ai.
