Anthropic filed updated IPO paperwork with the SEC this morning. In the second quarter, revenue grew 144 percent over the prior quarter. In the third quarter, that growth rate fell to 47 percent. The growth rate was cut by two-thirds in three months. At the same time, the company has signed up for about $895 billion in compute and data center spending, with roughly $70 billion due in the next 12 months. In the most recent quarter, Anthropic generated $17 billion in revenue while spending $17.5 billion to get there.
These three figures—how fast growth is slowing, the size of the spending commitments, and the amount burned last quarter to produce $17 billion—matter more than anything else in the filing. Other items such as lockups, voting rights, and the Japan sale sit on top of them. If the three core numbers hold up, the rest is noise. If they do not, the rest becomes a problem. The business is real, and no one is alleging fraud. The question is the price: is a company with slowing growth and accelerating bills worth a $1.5 trillion valuation?
Revenue Growth Trajectory Turns Lower
Steve Eisman noted that the most striking items in Anthropic’s S-1 were its long list of future capital commitments, summing to $800 to $900 billion, with the numbers running only through the second quarter. He flagged that the reported third-quarter figure shows sequential revenue growth falling from 147 percent to 45 percent.
The path has been uneven. Growth sat in the single digits, then rose to 45 percent, jumped to 147 percent and 135 percent, and has now dropped to 43 percent. The more recent trajectory points toward continued 43 percent growth rather than a return to the earlier triple-digit rates. Reaching $17 billion in quarterly revenue this quickly is notable, yet the company remains committed to spending even more. The growth in what it plans to spend is rising, not shrinking.
Anthropic reported $17 billion in Q3 revenue against $895 billion in purchase commitments. For the nine months ended, the company generated $33 billion in revenue, up roughly 1,200 percent from the same period last year. The initial prospectus included financials only through the end of the prior year; the updated filing added the more recent figures.
The revenue-to-compute ratio also declined, falling from 2.08 to 1.7 after a period of token maxing. That earlier phase of heavy token usage appears to have tapered.
Adjusted Metrics Versus Reported Losses
Anthropic reported an $80 billion net loss for the first half of 2026, yet its non-standard adjusted metrics showed $2.7 billion of profit. Another reference put the net loss at $87 billion with an adjusted figure of $2 billion. The adjusted numbers exclude $72 billion of liability evaluation losses, $9 billion in losses on other financial instruments, and $2 billion in stock-based compensation. The $72 billion charge largely reflects the increased accounting value of convertible notes as Anthropic’s valuation soared.
This approach is comparable to a restaurant claiming strong profits while ignoring the cost of food. Meanwhile, Amazon booked $63 billion in equity investments, primarily related to Anthropic, accounting for most of its pre-tax profit. Alphabet recorded a $136 billion investment gain. In the second quarter of 2026 alone, gains from Amazon and Alphabet represented 15 percent of S&P 500 earnings. Wall Street’s treatment has been inconsistent: Anthropic’s paper losses are excluded from its adjusted earnings while the corresponding paper gains at Amazon and Alphabet are counted. The result is that the AI boom may be far less profitable than headline growth suggests, with net losses far outweighing revenue.
IPO Filing Details and Structural Features
Growth came in higher than many expected. People anticipated closer to $11.6 billion; the company reported nearer $17 billion. Customer and channel concentration is high. Two customers accounted for a combined 20 percent of revenue in the first half, while 48 percent of revenue was generated through cloud and marketplaces including Google, Amazon, and Microsoft. Losing any of those relationships would matter, particularly since Google and Microsoft are competitors and Amazon has stepped back.
Governance is unusual. As a public benefit corporation, Anthropic must balance its mission with profits. An independent trust elects most of the board, while a single co-founder held shares controlling 50.01 percent of the vote, leaving IPO buyers with little say.
Compute commitments are massive. Anthropic is committed to spend more than $600 billion on computing infrastructure through 2045, requiring continued rapid growth to justify the outlays. A staggered lockup release schedule, similar to approaches used by Cerebras and SpaceX, includes 22 distinct lockup releases. These could create selling pressure in the months after the IPO, especially if the stock enters major index funds relatively quickly.
Anthropic states that its product impact exceeds $40 trillion in economic value. That figure is well overstated. Nothing has ever come close to $40 trillion in economic value, and the claim assumes Anthropic would be the only AI maker. It is a completely silly and made-up number.
The company has also encouraged employees to donate to charity as it prepares for the offering, with the message that they are about to become rich and should give to charities. Anthropic has moved toward what could be the largest IPO ever, filing publicly for a planned Nasdaq listing. The broader market reaction was short-lived; Nasdaq saw selling pressure following poor results from other companies.
Data Center Financing and Operational Headwinds
A state grid audit, not investor demand for AI, is now the binding constraint on refinancing Anthropic’s Texas compute campus. Morgan Stanley-led banks shelved a $5 billion to $7 billion junk bond sale tied to an Anthropic-leased Texas data center. The pause followed Governor Greg Abbott’s freeze on approving new grid connections for data centers until Texas regulators finish an audit. That leaves seven banks holding a $14 billion one-year loan they have already dispersed, each with roughly $2 billion at stake.
A major Anthropic data center has been delayed, leaving banks holding the bag. In its S-1, Anthropic warns that it will begin using NeoCloud on-demand services that cost three to five times as much and will significantly hurt operating margins.
Competitive Pressure and the Path Ahead
The growth rates of 140 percent and 137 percent are over. Open-weight models are starting to eat into Anthropic’s position. Token maxing has tapered off. At Startuphakk, teams have begun moving away from Claude Code toward internally run open-weight models. A few projects remain on Claude Code because of their size, but the overall direction is toward owned open-weight models such as Qwen Flash Next (also referenced as Quen 3.8 Flash Next). These can be run using the open-sourced Open Monoagent.ai framework on hardware as accessible as an RTX 3090 paired with 64 GB of RAM—even older DDR3 systems.
The expectation is that open-weight models will continue to take share, that token maxing will keep declining, and that efficiency gains will spread. Anthropic and OpenAI are unlikely to remain the only meaningful players.
Is the Valuation Justified?
The filing puts three numbers front and center: sequential growth that has fallen from the mid-140 percent range to the mid-40s, commitments that total roughly $895 billion (with more than $600 billion extending through 2045), and a recent quarter in which $17.5 billion was spent to produce $17 billion in revenue. Adjusted profit figures exclude large charges tied to valuation increases and other items, while reported losses remain substantial. Customer concentration, unusual governance, staggered lockups, and delayed data-center financing add further pressure. Anthropic’s claim of more than $40 trillion in economic impact sits far outside any historical precedent.
The question remains whether 47 percent growth supports a $1.5 trillion price tag while bills accelerate. One view is that Anthropic is unlikely to IPO, generate large gains, and sustain the earlier growth trajectory. That is not financial advice; it is an assessment of the numbers now on paper.
