Anthropic, the company behind Claude, is preparing for one of the largest IPOs in history. The projected valuation exceeds $2 trillion more than double the valuation of approximately $965 billion reached in its previous funding round. However, documents prepared for the listing also reveal a far more unsettling side: to fuel its AI growth, Anthropic has taken on potentially massive infrastructure commitments (or debts?).
In 2025, revenue surged from approximately $386 million to $4.59 billion, a growth rate of about 1,088%. The issue, however, is that the company recorded an operating loss of $8.06 billion. Its GAAP net loss reached a staggering $41.97 billion.

COMPUTE
In 2025, Anthropic spent approximately $7.33 billion on infrastructure and computing capacity, accounting for a huge portion of its operating costs. Looking ahead, the IPO prospectus outlines at least $518 billion in infrastructure commitments spanning roughly a decade. This does not mean Anthropic must pay the full $518 billion immediately; however, about 80% of these commitments are non-cancellable or otherwise payable regardless of actual capacity usage.

This figure is distributed primarily among major infrastructure providers:
- Google: at least $111.1 billion.
- Amazon: approximately $110 billion.
- Microsoft: approximately $31.4 billion.
- Broadcom: approximately $161.2 billion in equipment lease obligations.
- xAI: up to $84.5 billion for Nvidia capacity, subject to more flexible terms.
- other agreements make up the remainder of the total, bringing it to at least $518 billion.

The agreements with Google, Amazon and Microsoft also entail very onerous terms: in some cases, if Anthropic utilizes less capacity than contractually stipulated, it must still pay the difference. Although some online are calling this a "$518 billion debt" it is more accurate to refer to it as "future infrastructure commitments".

Separate financial structures or special purpose vehicles (SPV) were established to finance the purchase of massive quantities of Google TPUs. These SPVs purchase the hardware and lease it to Anthropic. This mechanism allows Anthropic to secure computing capacity without directly recording the debt used to purchase the chips on its own balance sheet. The transactions have secured approximately $71 billion in potential financing through two major private-credit structures: one worth around $35 billion (already closed) and a second in the range of $36 billion. This debt sits with the SPVs, while Anthropic bears the associated lease obligations.

KEY FIGURES
Let’s look at some figures:
$4.59 billion in revenue for 2025.
$8.06 billion operating loss.
$7.33 billion in compute and infrastructure costs.
$41.97 billion GAAP net loss.
Approximately $20.28 billion in cash and investments.
Up to at least $518 billion in future infrastructure commitments.
And a potential valuation exceeding $2 trillion.

The economic model for frontier AI therefore requires a completely different scale of capital compared to traditional software companies. Essentially, Anthropic is attempting to secure access to gigawatts of power, data centers, GPUs, TPUs, and specialized infrastructure for years to come—even before future demand is fully guaranteed. Anthropic itself has stated that compute availability could become the primary constraint on the growth of advanced models.

IPO PROSPECTUS: ANTHROPIC DISCUSSES THE RISKS OF ITS AI!
Out of 261 pages, approximately 80 are dedicated to risk factors, compared to about 48 pages describing the business. Anthropic warns investors that increasingly advanced models could theoretically develop behaviors that are difficult to control, including attempts to resist being shut down, concealing or manipulating information, and engaging in blackmail-like behavior. The company also explicitly addresses potential "catastrophic or existential" risks to humanity.

Anthropic is seeking a valuation of over $2 trillion from the market while devoting a significant portion of its prospectus to explaining the potential dangers of the technology it is developing. Anthropic also acknowledges that highly advanced models could realize they are being tested and alter their behavior, making it more difficult to accurately assess their safety.

SIMULTANEOUSLY A PARTNER AND A COMPETITOR
Finally, there is another notable aspect: Anthropic is simultaneously a customer, partner, and competitor to giants such as Google, Amazon, and Microsoft. These companies provide capital and infrastructure to Anthropic while also developing their own AI models. This represents a massive strategic dependency; Anthropic admits that its business could be severely impacted if access to computing power were restricted, became more expensive, or were cut off.

Article always updated with all the possibilities of on-chain farming (airdrop): Some Sites To Earn Crypto Bonus (Old & New)