Anthropic's IPO is expected to begin as early as Monday, with the AI giant set to go public. Previously, the market's highest valuation expectation for the company reached $2 trillion, but some investors have now lowered their expectations to $1.5 trillion. This $500 billion difference is not just a change in numbers, but also illustrates a shift in the valuation logic of AI companies: from "who grows the fastest" to "whether growth is sustainable and whether cash flow can be realized." Let's look at revenue first. Anthropic's revenue reached $16.3 billion in the first half of this year, compared to only $1.1 billion in the same period last year, a remarkable growth rate. Third-quarter revenue further increased to $16.8 billion, a sequential increase of approximately 45%. However, the growth rate has clearly slowed. Revenue growth exceeded 140% in the previous quarter, but has now dropped to 45%. This means that while Anthropic is still growing rapidly, the growth rate is declining. In the early stages of financing, the market is willing to pay for growth rate and market share; once it enters the IPO stage, investors are no longer focused on "how fast it can grow," but rather "what price will this growth cost." Now let's look at profits. The company reported a staggering $80.7 billion net loss in the first half of the year. However, approximately $72 billion of this came from valuation adjustments for financing instruments such as convertible bonds, which doesn't necessarily mean the company actually burned through an equivalent amount of cash. Anthropic also stated that it achieved profitability in the third quarter on an adjusted basis. However, it's crucial to note that adjusted profitability does not equate to an end of cash flow pressure. The company's future commitments to infrastructure such as cloud computing, chips, and data centers exceed $600 billion, representing a more significant long-term cost. AI models can be monetized quickly, but the computing power, energy, and data center investments required to train and run these models will also expand rapidly. Whether revenue growth can cover these capital expenditures will directly determine the quality of Anthropic's profitability. Furthermore, there are two easily overlooked risks. First, its revenue streams are highly concentrated. By 2025, approximately 47% of its revenue will be sold through Amazon and Google's cloud platforms. These two companies are not only investors but also computing power providers and key business channels. Such collaborations can accelerate growth, but they also mean Anthropic is heavily reliant on a few giants, and its bargaining power may not be entirely in its own hands. Second, who will truly control the company after the IPO? The seven co-founders plan to control 50.1% of the voting rights through a special shareholding structure; meanwhile, a long-term charitable trust composed of non-shareholders will still have the power to appoint and remove a majority of board members. This means that ordinary investors buying shares may not necessarily acquire corresponding control of the company. For an AI company still in a high-investment, high-uncertainty phase, governance structure itself is part of the valuation. Therefore, the real focus of Anthropic's IPO is not whether the final valuation reaches a certain integer threshold, but rather three things: Whether revenue growth can be sustained, how much cash will be burned through infrastructure investment, and how much say ordinary shareholders will actually have after the IPO. The $2 trillion story is certainly tempting, but what the capital market ultimately recognizes is sustainable growth, real cash flow, and future returns. The AI narrative can inflate valuations, but it cannot forever justify financial statements.