Everything About AI
NextFin News - Asian semiconductor stocks surged on Monday as Anthropic PBC's initial public offering filing reignited investor appetite for the artificial-intelligence trade, lifting South Korea's benchmark index by as much as 4.4% and pulling memory-chip makers Samsung Electronics and SK Hynix up more than 5% each in early trading. Japan's Nikkei 225 rose as much as 2.2%, with chip-equipment names Advantest and Sumco climbing more than 6% each. The move is a sharp reversal for a sector that had just endured one of its worst stretches of 2026, and it raises a question the market has been avoiding: is the AI spending boom being repriced on fundamentals, or merely on relief that the biggest private AI lab has finally opened its books?
Anthropic, the maker of the Claude chatbot, filed a Form S-1 registration statement with the U.S. Securities and Exchange Commission on October 26, applying to list its Class A shares on the Nasdaq Global Select Market under the ticker "ANT," with Goldman Sachs and Morgan Stanley leading a 22-bank syndicate. The filing did not disclose a share count or price range, but it laid bare the economics of the AI race: revenue of $16.29 billion in the first half of 2026, up from $1.16 billion a year earlier, and a net loss of $80.7 billion over the same six months. The combination — explosive growth paired with staggering costs — is what will set the valuation benchmark for every AI company that follows, and it is why chip stocks from Seoul to Tokyo rallied on the news. A company that spends hundreds of billions is a chipmaker's best customer, even when that company is losing money on paper.
The Market Reaction: Relief Rally, Short-Covering, and the Asymmetry of the Bounce
The tape told a story of short-covering as much as conviction. The biggest beneficiaries of Monday's move were the memory makers whose shares had been punished hardest, not the foundry whose order book is already locked in for years. That asymmetry is the first clue that this was a relief rally rather than a fundamental re-rating.
The relief had something to work with. In the days before the filing, SK Hynix had fallen roughly 26% across five trading sessions on fears that AI infrastructure spending was peaking, with Samsung down about 22% and Micron down about 11% over the same span. A 4.4% jump in the KOSPI on a single filing is not a re-rating of earnings; it is the market exhaling on the belief that the AI trade is not yet over. The decline itself had not been accompanied by any deterioration in the underlying demand for high-bandwidth memory used in AI servers. It was a positioning unwind, and positioning unwinds reverse quickly when a catalyst gives nervous holders an excuse to stop selling.
That pattern has repeated through 2026. Broadcom's guidance earlier in the summer triggered a correction. Reports of China's rapidly improving large-language models triggered another. Each found buyers. The October selloff is the third sharp downdraft in a single year, and like the others it reversed on a sentiment catalyst rather than a fundamentals one. This is the cyclical layer of the AI trade: the stocks swing violently on positioning and headlines while the underlying capital commitment grinds upward.
"The AI supercycle is unlikely to be any different. Unless upcoming earnings reveal evidence that enterprise AI demand is slowing materially, the selloff appears more consistent with a healthy valuation reset than the end of the AI investment theme."
That assessment, from a market strategist writing ahead of the rally, captures the fragile logic now at work: investors are treating the selloff as a buying opportunity only so long as the next wave of earnings confirms that enterprise demand is still accelerating. Anthropic's filing gives them a reason to believe it is — but the confirmation has to come from the hyperscalers' own numbers, not from a private company's prospectus.
What the S-1 Actually Says: Growth That Defies History, Losses That Defy Intuition
The numbers in Anthropic's prospectus are difficult to overstate. Second-quarter revenue alone was $11.56 billion — more than the company earned in its entire history before 2026, which totaled roughly $5.0 billion across 2023, 2024 and 2025. Annualized revenue has surpassed $30 billion, putting it ahead of rival OpenAI's roughly $24 billion to $25 billion run-rate, according to figures the company has disclosed. Revenue growth in the first half ran at roughly 1,300% year over year, a pace almost no company of any size has ever sustained. Full-year revenue climbed to $4.59 billion in 2025 from $386 million in 2024 and $41 million in 2023 — a growth curve that looks less like a company and more like a phenomenon.
But the loss side is where the filing becomes a Rorschach test for the AI thesis. The $80.7 billion net loss sounds catastrophic until it is decomposed: approximately $72.4 billion of it — about 90% — came from the non-cash revaluation of convertible notes, derivative liabilities and similar instruments that rise in value as the company's own valuation climbs. In other words, the accounting loss is largely a mirror of how much richer Anthropic's backers have become on paper. The actual operating loss from running the business was about $693 million in the first half, and core cash and operational losses were in the single-digit billions.
This distinction is not cosmetic. It is the difference between a company burning through cash and a company whose paper losses are the price of financing a buildout that its lenders have already agreed to fund. Yet the dilution risk is real: those convertible instruments can become shares, and when they do, existing holders will own a smaller slice of whatever Anthropic becomes. An $80 billion loss that does not touch the bank account is still a loss that lands on the cap table.
The scale of the buildout is what links Anthropic directly to the chip rally. The company has committed more than $500 billion in spending with six partners over the coming decade, including at least $110 billion with Amazon, about $114.5 billion with Google, and $44.9 billion with Microsoft for cloud computing over seven to ten years. Many portions of those agreements are non-cancelable or require payment regardless of usage. That is a revenue floor for the hyperscalers — and, one layer deeper, for the chipmakers and equipment suppliers they buy from. As long as Anthropic keeps spending, the AI supply chain keeps getting paid. The rally is not betting that Anthropic will be profitable; it is betting that Anthropic will keep spending.
The Second-Order Question: Who Is Really Funding the AI Boom, and Can It Compound?
Here the filing exposes something the market has preferred not to examine too closely. Amazon and Google are not merely Anthropic's largest customers; they are also among its top shareholders. This circular structure — where the companies buying the compute also own a stake in the company spending the money — has dominated capital flows in the AI industry, and it is the mechanism through which a private valuation boom gets converted into public-market revenue.
The transmission runs like this: hyperscalers raise capital or commit balance-sheet spending to build data centers; they sign long-term, often non-cancelable contracts with AI labs; the labs book that capacity as the infrastructure needed to serve enterprise customers; and the labs' growth then justifies the hyperscalers' own capital expenditure to investors. Anthropic's filing makes the loop visible. Data from corporate-spending tracker Ramp showed Anthropic accounted for 60.1% of AI token spending in the week ended October 11, a concentration that is both a sign of dominance and a vulnerability. A single customer base carrying that share of industry spend means the entire AI revenue story is sensitive to the health of a narrow set of enterprise buyers.
The second-order implication is uncomfortable for the chip bulls. The rally assumes Anthropic's spending will continue, but that spending is itself a bet on enterprise adoption that has not yet been proven at public-company scale. If enterprise customers slow their token consumption, the non-cancelable contracts keep the hyperscalers paying in the near term — but they will not sign new ones, and the growth rate that justifies today's valuations compresses. The chip stocks are not pricing the contracts Anthropic has signed; they are pricing the contracts it has not. That is the gap between a relief rally and a sustained uptrend, and it will be closed by the hyperscalers' earnings, not by Anthropic's filing.
The Counter-Thesis: A Valuation Test the Market Has Not Yet Taken
The strongest argument against reading too much into Monday's move is that the hard part has not happened. Anthropic has filed paperwork; it has not priced an offering. And the valuation being discussed in private markets — as much as $2 trillion — has already drawn some of the sharpest skepticism of the AI era.
Independent research firm New Constructs, in a report released this week, called a potential Anthropic offering the "most ridiculous IPO of 2026" and valued the company at $150 billion, dramatically below the $2 trillion figure circulating in investor discussions. The firm had not seen the actual prospectus, but it worked from reported figures, including revenue run-rate estimates. Michael Burry, the investor known for betting against the housing bubble before 2008, added his own comparison: at the discussed valuation, Anthropic would be worth as much as 78 S&P 500 companies combined, including top household brands and established industrial firms.
The math behind the skepticism is specific. Dr. Chan Ahn, a former Goldman Sachs and JPMorgan executive, estimated that a $2 trillion valuation would require Anthropic to generate roughly $725 billion in annual revenue by 2036, assuming a 10% cost of equity, a 25% free-cash-flow margin and a 25x terminal multiple. At a 13% discount rate, the required revenue rises to about $950 billion. For context, Anthropic's first-half 2026 revenue of $16.3 billion annualizes to roughly $33 billion. Bridging from there to $725 billion in a decade requires compounding at a rate that has almost no precedent for a company of that size.
This is the real test the IPO will impose, and it is why the filing matters more than the rally. SpaceX, which completed its own landmark IPO in June at a $1.77 trillion valuation after raising about $75 billion, showed that public markets will absorb AI-priced assets — but SpaceX came with Starlink, a profitable satellite-internet business, as a floor. Anthropic comes with a consumption-revenue model and an $80 billion accounting loss. The two are not the same asset class, and the market will price them accordingly. The IPO roadshow, when it begins, will be the first time the AI valuation thesis faces public-market discipline rather than private-round enthusiasm.
Cyclical Bounce, Structural Uptrend: Separating the Two
The correct read of Monday is not "the AI boom is back" or "the AI boom is a bubble." It is both, at different time horizons. The stock prices are cyclical; the capital commitment is structural. Confusing the two is how investors buy at the top of a sentiment wave and sell at the bottom of a positioning wash.
The cyclical case rests on the price action itself. Three sharp corrections in 2026 — the summer Broadcom-guidance episode, the China large-language-model reports, and the October oversupply scare — each reversed without any fundamental change in AI demand. Mean reversion in sentiment is the pattern, and Monday's 4.4% KOSPI jump is the mirror image of the prior week's decline: same mechanism, opposite sign. When valuations price in near-perfect execution, modest changes in sentiment trigger disproportionately sharp moves in both directions. The memory names that rallied hardest are the ones most exposed to the next reversal.
The structural case rests on harder ground. Anthropic's $500 billion-plus of committed compute spending is not contingent on next quarter's sentiment; much of it is contractually locked. Its revenue grew roughly 1,300% year over year on enterprise demand rather than consumer hype, and its focus on coding and workflow automation ties its product to measurable productivity gains. The transition of general-purpose computing workloads onto accelerated infrastructure is a decade-long reallocation, not a sentiment cycle. The picks and shovels get paid as long as the digging continues, and Anthropic's filing confirms the digging is far from over.
The falsifying signal is specific and observable. If hyperscaler cloud-revenue growth falls below the mid-teens on an annual basis over the next two earnings seasons while AI-specific revenue line items flatten, the structural thesis is damaged and the rally becomes a bear-market bounce. A second signal sits in the IPO itself: if Anthropic prices materially below the $965 billion post-money valuation set by its May 2026 Series H round — the $65 billion raise led by Altimeter, Dragoneer, Greenoaks and Sequoia — the private-market benchmark that has anchored the entire AI complex will have cracked. Until one of those prints, the burden of proof sits with the bears.
What Comes Next: The Roadshow, the Earnings, and the Watch List
The filing is the opening move, not the event. Anthropic must still have its registration statement declared effective, begin a roadshow, and price an offering. Reports have suggested a debut timed after the U.S. midterm elections in November, though the company has announced no date. It will also face competition for AI-focused capital: OpenAI has filed its own confidential S-1, and the market will be comparing the two lab valuations line by line.
For investors, the watch list is concrete. First, the final price range and share count, which will reveal the implied valuation and how much dilution the convertible instruments represent. Second, the upcoming technology earnings season, which will show whether enterprise AI demand is still accelerating or merely holding — this is the confirmation the relief rally needs. Third, memory pricing and high-bandwidth-memory supply data from Samsung, SK Hynix and Micron, which will confirm whether the physical bottleneck that has powered the rally is still binding.
The scenarios split cleanly by horizon. In the short term, sentiment can swing either way on headlines, and the chip names that rallied hardest are the ones most exposed to a reversal. Over the medium term, the earnings season will determine whether this was a genuine demand signal or a reflexive bounce. Over the long term, the structural shift in computing infrastructure is intact regardless of which company wins the model race — the supply chain gets paid as long as the spending continues.
Who benefits and who is exposed is not symmetric. The beneficiaries are the memory makers and equipment suppliers whose revenue is tied to committed capex; the exposed are the stocks that rallied on sentiment alone, with no contracted demand to fall back on. And the ultimate arbiter is not Anthropic's prospectus but the hyperscalers' next earnings call, because that is where the circular financing loop either tightens or breaks.
The irony of the moment is worth sitting with: a filing that shows $80 billion of losses triggered a rally in the very stocks that profit from those losses. That is not a contradiction. It is the market's verdict that in the AI boom, spending is the signal, and Anthropic has just signaled that it intends to spend hundreds of billions more. The rally is rational — it is also incomplete, because spending without compounding revenue is a bridge to nowhere, and that bridge has not yet been tested in public.
Explore more exclusive insights at nextfin.ai.
