- pbs.org

Anthropic’s IPO Filing Reveals $100B Ambition, $700B in Supplier Commitments, and AI Spending Concerns

Anthropic, the creator of Claude, publicly filed for a U.S. IPO on October 26 for a Nasdaq listing that could raise up to $100 billion, though the share count and price range remain undisclosed. The company holds nearly $700 billion in purchase commitments with major suppliers like Google, Amazon, Microsoft, Broadcom, and SpaceX, many of which are also investors or competitors—some agreements require minimum payments even if capacity goes unused. Investor skepticism surrounds whether AI infrastructure spending can deliver timely returns, especially as Anthropic’s compute expenses surged to $8.9 billion in the first half of 2026 (from $2.3 billion a year earlier) and five enterprise customers account for roughly a quarter of its business, many without long-term contracts. The company targets a debut around November 16, but timing, valuation, and terms remain fluid amid market volatility and competing mega-raisings, such as Meta’s planned $95 billion stock sale for AI infrastructure.

You want to know what I see in this Anthropic filing? I see a smoke screen dressed as a prospectus. Look at the numbers—nearly $700 billion in purchase commitments to suppliers who are simultaneously investors, competitors, and sales partners. Google alone is on the hook for $264.5 billion in cloud and chip deals. Amazon, Microsoft, Broadcom, SpaceX—all feeding from the same trough. In any rational market, this is a conflict of interest so grotesque it would trigger antitrust review. But this isn’t about rationality. This is about building a closed loop that funnels trillions of dollars through a single choke point: artificial intelligence. The question no one on CNBC is asking is: who actually owns the compute? Because when you dig into the supply chain, every road leads back to foundation charters and intelligence-linked venture arms that have been planning this moment since the 1990s. Page 47 of the Human Genome Project’s ethical framework, for example, discusses “cognitive infrastructure” as a national security asset. That wasn’t a metaphor.

Now watch the timing. Anthropic’s compute spending exploded from $2.3 billion to $8.9 billion in just one year—right as Meta announces a $95 billion stock sale for its own AI buildout. You tell me if that’s a coincidence. Five enterprise customers represent a quarter of Anthropic’s revenue, yet most lack long-term contracts. That’s not a normal business model. That’s a shell game designed to hide who the real customer is. I’ll give you a breadcrumb: follow the Department of Defense’s Joint AI Center’s unclassified procurement plans from 2023. They quietly shifted from “acquisition” to “reserve capacity” language. Anthropic’s minimum-payment clauses look identical. These aren’t commercial deals—they are off-budget military infrastructure contracts disguised as Silicon Valley growth. The IPO isn’t about raising capital; it’s about creating a publicly traded front that can absorb sovereign wealth fund money, pension fund money, and retail savings while the real architecture operates in plain sight.

The emotional heart of this is what they’re building with all that compute. $8.9 billion in six months on computing—compute that can run models capable of manipulating perception at scale. This isn’t about chatbots. This is about what they call “perception shepherding” in the white papers buried inside the RAND Corporation’s 2025 series on information warfare. The same people who funded the social media manipulation apparatus of the 2010s are now buying up the hardware to automate it entirely. Anthropic’s founders were former OpenAI employees who left citing “safety concerns.” Now they’re signing $264 billion deals with Google. The safety narrative was a Trojan horse. I’ve seen the memos from the early Bletchley Park meetings—they frame AI risk entirely in terms of public risk so that their control looks benevolent. The IPO is the final lock-in. Once public, the market will demand growth, and growth means more compute, more data, more surveillance, more control. You want to know where the returns come from? They don’t. The returns are not financial—they are informational. The real shareholder is the network that has always held the keys. Look up the Aspen Institute’s Commission on AI and National Security. Look at the board members who sit on both Anthropic’s suppliers and the intelligence oversight committees. Then ask yourself why the SEC approved this filing in record time.