The $200 Billion Tether: Anthropic’s Massive Google Deal Exposes the “Circular Economy” of Silicon Valley AI

According to reports from The Information, the artificial intelligence startup Anthropic has committed to an astounding $200 billion expenditure over the next five years to secure stable access to Google’s cloud infrastructure and specialized AI silicon. This monumental transaction not only redefines our perception of infrastructure costs but also epitomizes a prevailing Silicon Valley archetype: cloud behemoths provide the capital to nurture AI startups, only for those startups to promptly return those funds to the titans’ own data centers in exchange for computational power.

Anthropic’s astronomical contract with Google is not an isolated phenomenon. Indeed, the firm previously established a multi-billion dollar alliance with Amazon, involving the acquisition of AWS computational resources.

The report further highlights a staggering metric: the insatiable demands of just two startups—Anthropic and OpenAI—have generated a combined revenue backlog of $2 trillion for the “Big Four” cloud service providers: Amazon, Google, Microsoft, and Oracle.

The underlying commercial logic is transparent: while Microsoft, Google, and Amazon inject massive capital into these startups under the guise of “investment,” the reality is a form of “resource tethering.” Flush with cash but possessing limited alternatives, these startups are effectively compelled to funnel their capital back to their benefactors to procure cloud capacity and chipsets. Consequently, the cloud providers reap both potential investment returns and immediate revenue growth, creating a seamless, closed-loop financial cycle.

However, the fiscal burden of sustaining these Large Language Models (LLMs) has reached breathtaking proportions. Market forecasts for 2026 suggest that server maintenance costs alone could escalate to $450 billion for OpenAI and $20 billion for Anthropic. This rapacious hunger for compute has even prompted the semiconductor sovereign NVIDIA to engage in strategic investments in firms like OpenAI, thereby safeguarding its GPU hegemony and ensuring a steady stream of orders.

Anthropic’s $200 billion commitment vividly illustrates the “capital-intensive” nature of the contemporary AI industry. Yet, beneath these dazzling investment figures lies a latent crisis of sustainability.

Primarily, this “circular” transaction model is profoundly dependent on AI startups identifying a monetization strategy—such as high-premium enterprise subscriptions—capable of offsetting these exorbitant costs. Should the liquidity chain fracture or the practical profitability of AI fail to meet expectations, the bubble constructed upon compute leasing may face a severe reckoning.

Moreover, the “crowding-out effect” on the physical hardware supply chain warrants serious concern. To sustain these AI leviatans, the construction of data centers is voraciously consuming global electricity and critical electronic components. The report explicitly notes that the unbridled expansion of data centers is exerting immense pressure on finite resources, even precipitating shortages in Random Access Memory (RAM). This imbalance implies that everyday consumers may eventually bear the burden of higher component costs and retail prices for smartphones and PCs, effectively subsidizing the AI industry’s expansion. The AI revelry of tech giants may, in the end, be financed by the broader consumer market.