↓ Skip to main content
  1. Articles/

Z.ai Raises $5B in Five Days While Anthropic's S-1 Is Still a Confidential Draft

·1035 words·5 mins·
Florent Clairambault
Author
Florent Clairambault
CTO & software engineer — writing daily about spec-driven development and agentic coding

Z.ai Raises $5B in Five Days While Anthropic’s S-1 Is Still a Confidential Draft

Z.ai — the Chinese open-weight lab this blog has tracked since the Ox Alpha stealth-launch episode in August — raised roughly $5 billion in a single package on September 13: a ~$2 billion Hong Kong share placement and a ~$3 billion zero-coupon convertible bond. It’s the company’s second raise in two months, its third since its January 2026 IPO, and it lands the same week Anthropic’s own path to public markets is still stuck at “confidential draft,” eleven weeks after that draft was first submitted.

What Z.ai actually raised
#

Per the Hong Kong Stock Exchange filing (reported by Reuters and corroborated across multiple outlets), the package breaks down into two instruments priced the same day:

Share placement: 21.97 million new H-shares at HK$714 ($91.05) per share — a 10% discount to Friday’s HK$793 close — raising approximately $2 billion.

Convertible bond: ¥20.14 billion ($3 billion), zero-coupon, issued at 100.5% of face value, maturing September 2027. The conversion price is set at HK$892.50, a 25% premium over the placement price. The bonds carry an early-redemption clause: Z.ai can call them starting February 18, 2027, if the share price trades at or above 130% of the conversion price for 20 of any 30 trading days — a standard structure that effectively bets the stock keeps climbing.

Z.ai says 60% of proceeds go to “research and development of the company’s next-generation models and its fully self-training system,” 15% to expansion, and the remainder to capital structure optimization and general corporate purposes. Add this round to January’s roughly $4 billion IPO and July’s roughly $4 billion follow-on placement, and Z.ai has now raised close to $12 billion in public markets in eight months.

The pattern: raise fast, spend on compute, repeat
#

Three raises in eight months is an unusually fast cadence even by the standards of a capital-intensive AI buildout, and the timing lines up with a company that’s been running its models hot. This blog covered Z.ai directly when it unmasked itself as the operator behind Ox Alpha — a mystery coding model that ran six days of free, effectively unlimited access on OpenRouter and OpenCode, with Claude Code users among its top consumers by token volume, before Z.ai confirmed it was a load test for GLM-5.3-Flash. That launch tactic — subsidize massive real-world inference traffic first, price and name the model second — is expensive, and expensive habits need funding rounds to sustain them. A $5 billion raise nine days after a $4 billion one is what “close the funding gap between stealth-launching and actually monetizing” looks like in practice.

It’s also consistent with the broader thread this blog has followed through Kimi K3, Qwen3.8-Max, and GLM-5.3-Flash: open-weight Chinese labs closing the capability gap on agentic coding while operating largely on domestic accelerators, unconstrained by the same export-control regime that shapes US labs’ compute planning. Z.ai has been on the US Commerce Department’s Entity List since January 2025 over military-AI concerns — a restriction that hasn’t stopped Hong Kong’s public markets from funneling billions toward its next model generation, and hasn’t stopped Claude Code users from routing real production traffic to its infrastructure the moment a model looks free and capable.

The contrast with Anthropic
#

Set this against Anthropic’s own fundraising posture and the gap is stark. Anthropic confidentially submitted its S-1 draft to the SEC on June 1, 2026. As of this week’s direct SEC EDGAR check, there is still no public S-1 or S-1A on file under Anthropic PBC — secondary reporting has speculated an October listing window since before Labor Day, but nothing has materialized, and this blog has now logged more than half a dozen consecutive “still confidential” checks across September alone. Anthropic’s compute-funding strategy this year has instead run through a string of individually negotiated infrastructure deals — Akamai, AWS, SpaceX, and the reported, still-unconfirmed-by-any-named-party $35B Lambda/Hut 8 arrangement covered here September 2 — each disclosed (or not disclosed) on its own schedule rather than through a single public capital raise.

Two very different postures toward capital and transparency are visible here. Z.ai is raising fast, disclosing terms in a regulatory filing within days, and converting public-market capital directly into compute and training runs, all while operating under a foreign Entity List designation that would make most US investors nervous. Anthropic is sitting on a materially larger, still-confidential IPO process while quietly negotiating individual compute deals that only become public through reporting, leaks, or a counterparty’s own stock-price reaction — the Lambda deal’s only public confirmation so far was a 4% premarket pop in Hut 8’s share price, not a statement from any of the three parties actually involved.

Neither posture is obviously wrong — Z.ai’s speed carries real dilution and governance costs its shareholders are accepting; Anthropic’s confidentiality is standard pre-IPO practice and consistent with its broader “controlled disclosure” approach to safety and security communications this blog has tracked all year, including the still-overdue PyPI incident transcript. But the raw pace difference — three completed raises totaling $12B for Z.ai against zero completed public raises and a still-sealed draft for Anthropic — is a real data point about how differently the two companies are willing to move when it comes to their own capital structure, even as both are converging on similar answers about how agentic coding infrastructure should work.

What to watch
#

Z.ai hasn’t said what “next-generation models” the new capital specifically targets — GLM-5.3-Flash was a July/August-cycle release, and a fourth funding round this fast implies either an unusually expensive training run already underway or a compute shortfall from serving GLM-5.3-Flash’s traffic at the pace Z.ai itself said set call-volume records. Worth checking Z.ai’s own announcements and Hugging Face releases over the coming weeks for what specifically gets built with this money, and worth continuing the standing watch on Anthropic’s own EDGAR filing status — the S-1 could go public at any point and would immediately be the bigger story of the two.

Sources: The Express Tribune, “China’s Z.ai raises $5 billion from new share, convertible bond sales, filing shows” (Reuters-sourced, primary filing details) · PANews coverage (corroborating detail) · this blog’s own Ox Alpha/GLM-5.3-Flash coverage and direct SEC EDGAR check on Anthropic’s S-1 status.

Industry

Cognition's Series E Closes at $48B — Bigger Than the Rumor, Not Smaller

Cognition closed its Series E on September 8 at a $48B valuation on over $2B raised — both figures higher than the $47B/$1B talks Bloomberg reported six days earlier. Annualized revenue hit roughly $900M, up from $492M in May, with a16z and Accel leading a round more than 20 investors deep.

Sep 11, 2026 · 5 min read

Industry

Cognition's Reported $47B Round: Money Is Betting on Autonomy, Not IDEs

Bloomberg reported September 2 that Cognition, maker of the Devin coding agent, is in talks to raise roughly $1B at a $47B valuation — up from $26B just three months earlier — with annualized revenue above $900M, up from $492M in May. The round hasn’t closed and terms could still change, but the trajectory says more about where investor conviction is heading than any single benchmark score.

Sep 10, 2026 · 6 min read

AI Tools

Anthropic Filed Its S-1. Here's What Going Public Actually Means for Claude Code Users.

Anthropic filed a confidential S-1 with the SEC on June 1, kicking off a fall 2026 IPO process at a reported $965B+ valuation — and Claude promptly had an outage the same morning. For developers building on Claude Code, going public is not just a financial milestone. It is the moment Anthropic’s decision-making calculus permanently changes.

Jun 3, 2026 · 6 min read

Industry

Anthropic Walked Away From a $6B Acquisition. That's the Real News.

Anthropic conducted due diligence on Israeli AI-efficiency startup Decart, valued at roughly $6 billion, then walked away without a deal — per Bloomberg, corroborated by multiple outlets, both companies declining comment. The collapse says more about IPO-timing discipline than about Decart.

Sep 9, 2026 · 5 min read

Industry

Anthropic Won in Court for Refusing to Weaponize Claude. Days Later, It Admitted Where Its Agents Went Off the Rails.

A federal judge ruled the Pentagon’s ‘supply chain risk’ designation of Anthropic was illegal retaliation for refusing to strip Claude’s guardrails on autonomous weapons and mass surveillance. Three days later, Anthropic’s own blog disclosed exactly how two of its agentic models broke those same safety assumptions in live testing — and what it’s doing about it.

Sep 1, 2026 · 7 min read