Anthropic Races Toward IPO: Why Did the A-Share AI Supply Chain Rise First and Then Fall?

Anthropic is officially making a push for an IPO, rekindling enthusiasm in the long-dormant AI sector.

After the star AI company Anthropic officially submitted its S-1 prospectus to the U.S. Securities and Exchange Commission (SEC) on local time October 26, the Philadelphia Semiconductor Index surged 2.82% overnight, and on October 27, Asia-Pacific tech stocks rebounded collectively. South Korea's KOSPI index rose more than 3% during the session, SK Hynix closed up 5.18%, and Samsung Electronics closed up 4.23%.

In the A-share market, on October 27, AI chain sectors such as computing power, optical communications, memory semiconductors, and PCB were active at the open, but the index opened higher and then fell back, with funds rotating toward robotics, banks, and other directions. Some market participants said this looks more like structural catch-up gains and high-low switching under a repair of global AI trading sentiment rather than a trend reversal, and the key going forward is whether cloud giants' earnings reports and domestic model iterations can carry the momentum.

Rising first, then falling

On October 27, the AI computing power industry chain was once the most active main line in early trading, but by the close it turned into the leading decliner.

In early trading, computing power leasing, optical communications, memory semiconductors, and PCB took turns being active, but funds did not spread across the board, and structural divergence was obvious.

As of the close, the robotics (motors, reducers) direction bucked the market trend and set off a wave of limit-up gains. The AI computing power that was active in early trading faced concentrated profit-taking in the afternoon, with the IDC (computing power leasing) index down 3.78%, the optical communications index down 3.22%, the optical module (CPO) index down 4.13%, and the STAR 50 plunging 4.81% to 1252.28 points.

Jinxin Fund believes that the tech sector ultimately closing lower and adjusting today is high-level chip digestion and style rebalancing in a low-volume environment, not a turn in the industrial logic of tech growth. From the perspective of industrial fundamentals, in the first nine months, profits of high-tech manufacturing above designated size grew 52% year on year, and profits in the electronics industry doubled year on year, among which profits in the manufacturing of electronic special materials and the manufacturing of semiconductor discrete devices grew 2.2 times and 51.7% respectively, while profits in the manufacturing of optical fibers and optical cables grew 5.3 times. Demand realization driven by AI computing power and advanced storage is still being continuously verified.

Guotai Fund analyzed that as third-quarter reports enter a concentrated disclosure phase, uncertainty over earnings realization is putting pressure on directions that had high expectations earlier. Theme stocks previously driven by industrial expectations face earnings tests, and for AI hardware segments such as optical modules and storage, the market is divided on whether their high prosperity can continue to translate into reported profits.

"Overall, since the tech stock correction in July, their total market capitalization has fallen significantly, and their marginal drag on the index has also diminished," Guotai Fund said, noting that judging from downside room, the degree of chip clearing, and the panic index, this round of adjustment is nearing its end.

CSC further analyzed that with overseas core models making concentrated breakthroughs from September to October, domestic mainstream models will usher in intensive product and algorithm iterations in the fourth quarter, and the next generation of core technology versions is expected to emerge one after another. The narrowing technology gap and expectations for application scenario implementation are injecting fundamental support into related tech targets.

Whether the market can upgrade from sentiment repair to fundamental revaluation will ultimately be tested by investment returns.

The prospectus shows that in the first half of 2026, the company's revenue reached $16.286 billion, with Anthropic's revenue growing explosively. Its revenue in 2023, 2024, and the first half of 2025 was $41 million, $386 million, and $4.585 billion, respectively. Anthropic said revenue growth and improvements in computing resources and operational efficiency drove the improvement in the company's operating performance.

But under GAAP (generally accepted accounting principles), Anthropic's net loss in the first half of the year reached $80.7 billion. The company explained that this huge loss mainly came from "fair value changes in convertible notes, derivative liabilities, and SAFEs," which amounted to $72.388 billion, with the remaining loss mainly from computing power and infrastructure spending.

If non-cash items such as equity incentives are excluded, under the company's self-defined non-GAAP basis, adjusted operating profit in the first half of 2026 was a positive $1.618 billion, with an adjusted operating margin of about 10%.

The prospectus first responds to two questions that had previously suppressed AI hardware valuations: whether AI companies actually make money, and whether computing power demand has long-term constraints. The company's computing power spending as a share of revenue has fallen from 72% in the first quarter to 48% by the end of the second quarter, and economies of scale are beginning to show.

Global trading sentiment was quickly ignited, but that does not equal a turn. For sentiment repair to upgrade into fundamental revaluation, earnings reports still need to carry it forward.

Guosen Securities analyzed that about 80% of Anthropic's revenue comes from token-consumption-based billing, its net revenue retention rate is close to 900%, no single customer contributes more than 10%, and the quality and stickiness of its revenue are rare among large-model companies, but nearly half of its revenue relying on distribution through three major cloud channels is the main hidden concern.

On the industry side, many popular AI industry chain stocks have halved even with strong fundamentals. Recently, earnings reports from Meta and Tesla hit tech stocks hard, and their share prices continued to fall after they announced third-quarter results. Meta's results showed revenue hitting a record high, but its free cash flow turned negative for the first time since 2012, highlighting the pressure that large-scale AI investment is putting on the company's ability to generate cash. At the same time, the company's AI infrastructure spending rose 80% year on year, further raising investor concerns about its massive capital expenditures.

On the financing side, PIMCO analysis pointed out that as of September 23, AI-related issuers already accounted for nearly a quarter of U.S. dollar non-financial corporate bond issuance, up from less than 13% in 2025 and about 4% in 2024; against the backdrop of rising U.S. Treasury yields and higher financing costs, AI companies are still crowding into debt issuance, and the concentration of AI-related debt in the credit bond market continues to rise. JPMorgan Asset Management estimates that by 2030, complete data center construction may require about $5 trillion in funding, of which about $2 trillion may be financed through the investment-grade credit bond market.

Some industry insiders further analyzed that with Anthropic's listing imminent, the global AI narrative remains, and the more prominent current feature is structural catch-up gains and sector rotation. A one-day pullback should not be inferred as a pessimistic expectation; the next step is to see whether cloud giants' earnings reports and domestic model iterations can push sentiment repair into fundamental revaluation.