Z.AI's $5B raise, Anthropic's Nasdaq choice, Trump's slowdown pushback, and Astra at Perplexity (2026)

This brief covers AI developments published between September 13 at 08:15 and September 14 at 08:15 in Dhaka. Over the past 24 hours, the artificial intelligence industry experienced a sharp contrast between rapid capital accumulation, shifting public-market strategies, political pushback against capability controls, and deeper enterprise deployments. Key developments include Chinese AI lab Z.AI raising $5 billion across new shares and convertible bonds, reports that Anthropic has selected Nasdaq for a potential public listing, Donald Trump rejecting calls to pace frontier AI development, and OpenAI detailing how Perplexity uses GPT-6 Astra to autonomously test production systems.

Quick scan

Development	What changed	Why it matters	What to watch
Z.AI $5B financing	Z.AI raised $5 billion through a ~$2 billion Hong Kong share placement and a concurrent ~$3 billion zero-coupon convertible bond sale. 1	Solidifies Chinese commercial frontier labs' balance sheets, funding compute clusters and self-training architectures to match U.S. model capabilities.	Formal closing of the dual offerings and capital allocation milestones in subsequent quarterly filings.
Anthropic Nasdaq selection	Business Insider reported that Anthropic selected Nasdaq as the listing venue for its potential initial public offering. 2	Signals divergent public-market timetables among frontier labs after OpenAI postponed its listing beyond 2026 to focus on safety readiness.	Confidential draft registration submissions or official S-1 filings with the SEC.
Trump rejects AI slowdown	Donald Trump dismissed calls to slow AI development, calling critics "very negative forces" and arguing the U.S. must outpace China. 3	Removes executive-branch momentum for voluntary cross-lab capability pacing pacts, prioritizing speed in geopolitical competition.	Congressional responses to statutory duty-of-care proposals amid executive opposition to AI limits.
Perplexity deploys Astra	OpenAI detailed Perplexity's production use of GPT-6 Astra for automated software editing, test generation, and system monitoring. 4	Marks an operational shift from developer autocomplete to autonomous verification harnesses managing complex infrastructure.	Regression rates and long-horizon reliability metrics in unsupervised testing environments.

Beijing-based artificial intelligence developer Z.AI Co Ltd (2513.HK), formerly known as Zhipu AI, announced that it raised $5 billion in fresh capital through a dual-tranche transaction in Hong Kong, according to a stock exchange filing published Sunday. 1 The capital raise combines approximately $2 billion from a primary share placement and $3 billion from an offering of zero-coupon convertible bonds. 1

In the equity placement, Z.AI issued 21.97 million new shares at HK$714 ($91.05) per share, representing a 10% discount to its Friday closing price of HK$793. 1 The debt tranche consists of 20.14 billion yuan ($3 billion) in zero-coupon convertible bonds due in September 2027. 1 The bonds are denominated in yuan but will settle in U.S. dollars, issued at 100.5% of face value with an initial conversion price of HK$892.50, a 25% premium over the share placement price. 1 The company retains an option to redeem all outstanding bonds starting February 18, 2027, if its shares trade at or above 130% of the conversion price for at least 20 out of 30 consecutive trading days. 1

According to the regulatory filing, Z.AI plans to direct roughly 60% of the net proceeds toward research and development for next-generation foundation models and fully self-training systems. 1 Another 15% is earmarked for commercial and geographic expansion, with the remainder dedicated to replenishing working capital and optimizing its capital structure. 1 The transaction comes two months after Z.AI raised $4 billion in a follow-on offering in July, highlighting the escalating capital intensity required for Chinese frontier developers to secure compute resources, high-bandwidth memory, and technical talent. 1

What to watch: The final settlement of the convertible debt tranche and compute procurement commitments detailed in Z.AI's next quarterly disclosures.

Anthropic reportedly selects Nasdaq for potential initial public offering

Anthropic has selected Nasdaq as the listing venue for its planned initial public offering, according to reporting by Business Insider cited by Reuters on Sunday. 2 The decision, shared by a person familiar with the company's preparations, represents a key structural step toward taking the Claude developer public. 2

The venue selection follows recent discussions involving major semiconductor partners, including reports that Nvidia is in talks to anchor Anthropic's offering with an investment of up to $10 billion, valuing the company near $100 billion. 2 However, the timeline and final parameters remain fluid; neither Anthropic nor Nasdaq has issued public comment, and formal registration documents have not yet been made public. 2

The move creates a stark contrast in public-market strategy between the two leading frontier AI developers. 2 While OpenAI Chief Executive Sam Altman recently announced that OpenAI will remain private through at least 2026 to ensure its safety and alignment structures are robust, Anthropic appears to be maintaining its listing preparations. 5 This divergence indicates that frontier labs hold different views on whether quarterly reporting obligations and public shareholder scrutiny can coexist with long-term safety commitments. 2

What to watch: Any formal submission of draft registration papers with the U.S. Securities and Exchange Commission and whether prospective safety pacts affect the offering timeline.

Trump dismisses AI pacing calls, citing competition with China

Speaking to reporters at his golf property in Doonbeg, Ireland on Sunday, U.S. President Donald Trump pushed back against recent proposals from AI lab leaders to deliberately slow down frontier model capability advances. 3 Trump characterized warnings about catastrophic AI hazards as exaggerated claims driven by "very negative forces that are bringing it up that shouldn’t be bringing it up." 3

When asked whether the artificial intelligence sector should slow its pace or accept tighter regulatory restrictions, Trump stated that while baseline guardrails could be considered, national competitiveness must take precedence. 3 "We’re leading China in AI… whoever wins AI wins," Trump remarked, emphasizing that the United States cannot afford to decelerate its technological output. 3

Trump’s remarks land in direct opposition to calls made over the weekend by Anthropic Chief Executive Dario Amodei, who urged leading AI developers and democratic governments to coordinate pacing limits on recursive self-improvement. 3 6 Amodei had noted that verifying compliance across international competitors remains the most complex dilemma in enforcing a capability speed limit, but maintained that unchecked capability scaling risks catastrophic infrastructure disruption. 3 With the executive branch explicitly prioritizing geopolitical primacy, voluntary or statutory pacing efforts face severe political headwinds. 3

What to watch: How bipartisan Senate negotiators drafting statutory duty-of-care legislation reconcile their proposals with the White House's pro-acceleration stance.

Perplexity deploys GPT-6 Astra for end-to-end software testing

OpenAI published an architectural case study on Monday detailing how answer-engine startup Perplexity has integrated GPT-6 Astra across its engineering operations. 4 Rather than using the model primarily for interactive developer autocomplete, Perplexity relies on Astra to modify production software, draft internal communications, and autonomously monitor running services. 4

According to Perplexity Cofounder and Chief Strategy Officer Johnny Ho, code-generation improvements directly translate into search performance, allowing the platform to write more efficient programs that query, parse, and summarize web data. 4 The most significant architectural change, however, centers on automated testing. 4 To overcome manual testing bottlenecks, Perplexity tasks Astra with building synthetic testing environments around its core applications. 4

In this setup, Astra simulates downstream dependencies by generating realistic responses that mimic third-party language model APIs and data connectors. 4 By standing in for external services, the model validates entire application workflows from initiation to output, checking system resilience and error handling. 4 Ho stated that the team can now trust the model with end-to-end software systems and check in on its operations far less frequently than was required with previous model generations. 4

What to watch: Enterprise adoption patterns of autonomous end-to-end test generators versus traditional mock frameworks across complex production stacks.

What to watch

Settlement of Z.AI's dual-tranche offering: Confirmation of the $3 billion convertible bond closing will indicate liquidity depth in Asian capital markets for foundation model infrastructure. 1
Frontier lab listing roadmaps: Formal registration statements from Anthropic or Nasdaq will confirm whether public markets can accommodate frontier model research timelines. 2
Federal legislative vs. executive alignment: Whether congressional negotiators modify proposed AI duty-of-care bills following President Trump's public skepticism of capability limits. 3
Autonomous testing reliability at scale: How agent-generated synthetic test environments perform as underlying application codebases evolve over months of production operations. 4