Anthropic估值传闻破2万亿美元? AI投资是否陷入泡沫再度引发热议

The US stock market is about to witness another historic moment.

According to media reports and secondary market speculation, AI company Anthropic could officially go public as early as October, with market expectations valuing the company at over $2 trillion (this figure has not yet been officially confirmed by the company; please refer to the formal prospectus for specifics). If this number materializes, it would set a record for the largest global IPO—surpassing even SpaceX's previous listing record.

A company valued at only around $965 billion during its fundraising in May 2026 has been pushed to its current heights by secondary market trading prices, driven by the rapid growth of the Claude model series. And before this IPO has even landed, an old question has been reignited: Has the AI investment boom already entered bubble territory?

1. Bulls Do Hold a Strong Hand

First, let's look at where the supporters' confidence comes from—Anthropic's just-disclosed Q2 operating data was impressively strong, exceeding expectations.

According to operating data circulating in the market (not yet officially disclosed by the company; please refer to the prospectus or official financial reports for specifics), Anthropic generated quarterly revenue of $11.5 billion, compared to just $787 million in the same period last year—a year-on-year increase of 1,361%. In Q1 of this year, revenue was $4.7 billion, meaning it doubled in just three months. The company has also crossed a profitability watershed: it has begun generating profits, with gross margins in its inference business jumping from 38% a year ago to over 70%.

The criticism that "AI companies only burn cash without producing output" appears weak in the face of these numbers. The bulls' narrative thus follows logically: the high valuation is paying for the speed at which AI commercialization is being realized, and for the long, snowballing runway of large model adoption in the enterprise market.

2. But the Quality of Those Gross Margins Deserves Scrutiny

Why have profit margins improved so quickly? Breaking down the cost structure, the answer lies not with Anthropic itself, but within its list of chip suppliers: Google TPUs and Broadcom's custom chips.

The current compute landscape is this—in the training phase, Nvidia's dominance remains unshaken; however, spending on the inference phase is increasingly shifting toward custom chips, and the larger the business scale, the faster this shift occurs. Anthropic's gross margin leap largely benefits from declining inference costs rather than enhanced pricing power.

This directly impacts an upcoming observation window: Nvidia's earnings report on August 26. The revenue growth rate of its inference business will serve as the most direct yardstick for measuring the progress of this "inference migration."

3. The Bears' Hand: Overvalued Pricing and Liquidity Drain

Now let's consider what the skeptics are worried about.

First is the valuation method. The $2 trillion valuation corresponds to revenue expectations for 2028—today's price tag is attached to a report card two years out. This kind of pricing demands nearly flawless growth: if revenue growth slows from tenfold year-over-year to double or single digits, the current valuation framework would lose its support.

Second is liquidity. The largest IPO in history means siphoning over a hundred billion dollars from the secondary market at once; worse, OpenAI is queuing up behind them. By the time Anthropic rings the bell, there may simply not be enough money to go around in the market.

So there are two things worth watching: the growth rate of inference revenue in Nvidia's earnings report; and whether the revenue figures disclosed in Anthropic's official prospectus align with the data circulating in the secondary market. A mismatch in the numbers is often a signal that the story is starting to sour.

4. Bulls and Bears at a Standoff? BofA Offers a "Hedge Both Sides" Approach

Regarding whether AI has entered a bubble zone, bulls and bears each hold their ground. According to publicly reported market information, Michael Hartnett, Chief Investment Strategist at BofA Securities, proposed in his latest research report that investors need not choose sides between bulls and bears but may consider allocating to two types of assets simultaneously (for specific strategy details, please refer to BofA's official research report; this summary may involve information loss).

His suggestion is to go long simultaneously on AI technology leaders and long-neglected, out-of-favor assets, capturing the gains from the final blow-off phase of the nominal GDP bubble from both directions, while shorting AI-related bonds.

This judgment is supported by a set of data. BofA's Bull & Bear Indicator has fallen from 9.7 to 9.3, still remaining in extreme bullish territory; the "sell" signal has not been lifted—yet since this signal flashed in May, global equities have continued to climb higher. Capital flows are also diverging: gold and commodities continue to see structural inflows, tech stocks just experienced their largest weekly outflow in seven weeks, and private clients' equity holdings have risen to record highs.

Hartnett cites historical bubble patterns, pointing out that as the dominant bubble nears its peak, emerging markets or oversold cyclical assets often benefit from the spillover effect. This time, the consumer sector is the most likely candidate to take over. The reasoning for shorting AI bonds is equally clear: with over $1 trillion in AI capital expenditures paired with negative cash flows, a flood of related corporate bond issuance is imminent, and with supply pressure mounting, prices are unlikely to improve.

The core of this strategy lies in this idea: the most manic final stretch of a bubble often yields the richest returns—but beware, your chips cannot all be placed on the same side.

Important Notice: BIT Securities operations rely on its group company BIT (formerly Matrixport)'s subsidiary, Matrix Gelephu Pte. Ltd., and are regulated by the Gelephu Financial Services Office (GFSO) in Bhutan, providing eligible clients with access channels to US stock trading services, with relevant trades executed and cleared through US registered broker-dealers. This article was written and provided by an external contributing author. The Anthropic valuation, financial data, and listing timeline mentioned herein are compiled from public reports and market speculation and have not been officially confirmed; please refer to the company's formal disclosures for specifics. The views, analyses, and judgments expressed herein represent the author's personal opinions and do not represent the official stance of BIT or BIT Research. BIT also makes no guarantees regarding the accuracy, completeness, or timeliness of related content. The views of third-party institutions cited herein (including those of the BofA Securities strategist) represent only their own or the analyst's personal opinions, do not represent BIT's position, and their accuracy is not guaranteed. This article does not constitute investment advice, an offer, or a solicitation of an offer. Valuations of unlisted companies and AI-related assets are highly volatile, and historical growth rates do not guarantee future performance. Investors should make independent judgments and consult professional advice.