AI · Nov 5, 2026
Perplexity Teammates bets that always-on work agents are worth $200 a monthAnthropic’s $107 billion debut puts its safety brand and data-center fights on the balance sheet
· Updated · Signal Desk newsroom
Analysis of reporting by WIRED. How we work

Anthropic closed history’s largest IPO and jumped more than 30% on debut. WIRED’s reporting shows the risks that remain: a $900 billion vendor bill, local revolts over data centers, and a safety culture that can slow growth.
Key takeaways
- The IPO raised about $81 billion, more than $107 billion with private share sales, topping SpaceX’s $75 billion record; shares jumped over 30% on Nasdaq debut under ANT.
- Anthropic has about $200 billion in cash but more than $900 billion in vendor commitments over 19 years, mostly for data centers.
- Quarterly sales hit $17 billion, up more than 1,100%, while stock-compensation accounting drove over $104 billion in losses through September.
- Dario Amodei called safeguards the limiting factor in deploying models, ahead of electricity access.
- Community groups in West Virginia and Mississippi are fighting Anthropic data centers, and some labor groups urged pension funds to skip the IPO.
A record, with a long tail of obligations
Anthropic is now a public company with a war chest that WIRED says exceeds every tech giant’s except Alphabet’s: about $200 billion. The IPO raised about $81 billion, or more than $107 billion including private sales, beating SpaceX’s June record of $75 billion. Shares jumped more than 30% on their Nasdaq debut. The initial valuation of roughly $1.5 trillion fell short of SpaceX’s, and WIRED reports Anthropic sought a higher one and didn’t get it, a hint of Wall Street nerves.
Those nerves have a number attached. Anthropic has committed more than $900 billion to vendors over the next 19 years, mostly for data centers. Bulls point to $17 billion in quarterly sales, up more than 1,100% year over year, and a twelvefold jump in customers spending six figures annually. Our coverage of the IPO’s circular-financing test laid out why those commitments hinge on continued growth. Losses of over $104 billion through September are mostly a stock-compensation accounting effect, but they still show how much of the value is promised rather than earned.
Safety as both product and brake
WIRED’s framing is that Anthropic’s worst enemy is itself, and the strongest evidence is from Dario Amodei. At an investor meeting, he called safeguards the limiting factor in deploying models, above electricity, which most rivals cite first. As a public benefit corporation, Anthropic is allowed to weigh social impact against profit. Amodei says it held back a model this year at the cost of sales, and screens hires for public-benefit orientation, which he admits is painful in the short run.
We don’t see that as a weakness to apologize for. It’s the reason many enterprise customers trust the company, and it’s consistent with what the prospectus said about catastrophic risk. But public markets reward growth every quarter. The real test is the first time a safety hold collides with an earnings target. Concentrated control matters here: the cofounders hold a majority of voting power, and employees who donate shares to company-offered philanthropic funds hand their votes to Amodei. That protects the mission from short-term investors. It also means few checks on the people running it.
The ground game is going badly
The data-center opposition is concrete. In West Virginia, Mason County United, a 70-member group, is fighting an under-construction site intended for Anthropic. After extreme rain and grading at developer Nscale’s site, neighbors’ backyards flooded; Nscale says it fixed damage and bought out some homeowners above market rate. In Mississippi, a homeowner coalition member vows to fight a planned Anthropic facility across the street from her, set to open next year. WIRED notes US voters largely rejected data centers this month when given the chance, that Anthropic has made no renewable-power commitment, and that it has never published an environmental impact report.
That last point is an easy fix for a company that sells itself on responsibility. If the public benefit is real, it should start with the neighbors. Pointing to future philanthropy from newly rich employees, as the filing does, isn’t a substitute for transparency where the concrete is being poured. Theory Ventures’ Tomasz Tunguz warns permitting in the US will be extremely difficult and could push building abroad, adding risk and cost.
Who is exposed
- Public investors
- Strong growth and cash, set against $900 billion in commitments and a governance structure they can’t easily influence.
- Communities near data centers
- Flooding, traffic and power-bill concerns with little local benefit so far, and no published environmental reports.
- Enterprise customers
- A safety-first supplier with more capital, but possible deployment delays when safeguards lag.
- OpenAI
- Beaten to market; WIRED notes it was once expected to list first and is raising at a $1.25 trillion valuation.
What to watch
- Whether Anthropic publishes environmental impact reports or renewable-energy commitments.
- Permitting outcomes for the West Virginia and Mississippi sites.
- The first post-IPO earnings and whether any model release is held back for safety reasons.
- Pension funds’ response to labor groups urging them to avoid the stock.
- Any shift of data-center building outside the US.
FAQ
How much did Anthropic raise?
About $81 billion in the IPO, or more than $107 billion including private share sales, according to WIRED.
Is Anthropic profitable?
No. It reported over $104 billion in losses through September, largely due to stock-compensation accounting, though operating losses are declining.
Why do communities oppose its data centers?
Residents cite flooding, construction traffic, housing pressure, pollution, higher power bills and little local financial benefit.
Who controls Anthropic?
Its cofounders collectively hold a majority of shareholder voting power.
Companies: Anthropic
More on Anthropic
AI · Nov 21, 2026
Anthropic’s safety promise now answers to one real boss: its compute billAI · Nov 10, 2026
New York’s RAISE Act gives frontier labs a 72-hour clock for reporting dangerous AI incidentsAI · Nov 19, 2026
Apple wants to build the Mac agent it is warning users aboutAI · Nov 13, 2026
A week with Claude Duet: good at catching what falls between apps, shaky on tone and urgency