Anthropic prices the largest IPO ever; Google ships Gemini 4 Neon
Anthropic priced its IPO at $83 a share, above its range, raising about 81 billion dollars. Google separately released Gemini 4 Neon, which it says does tasks for a third less than Carbon. Plus 6 more stories.
Listen to this brief
Audio & video are paid features
Plus unlocks audio streaming and PDF downloads. Pro adds offline MP3 downloads, video, certificates, and more.
- Audio streaming
- Downloadable PDFs
- All AI Playbooks
- Personalized content
- Certificates of completion
- Audio MP3 downloads
- Video lessonssoon
- & More…soon
Anthropic sold more stock in its public offering than any company before it, and begins trading on Nasdaq under the ticker ANT. Google added a fourth model to its Gemini 4 line, and OpenAI published new measurements of how much its own research now runs on AI agents.
- 1
Anthropic prices its IPO above the range, raising more than any company before it
Anthropic priced its initial public offering at $83 a share, above the $70 to $78 range it set when the roadshow opened on November 10, and sold 980 million shares, up from the 825 million first planned. TechCrunch puts the proceeds at about 81 billion dollars, passing SpaceX's record, and the fully diluted value at roughly 1.5 trillion dollars. Shares are set to begin trading on Nasdaq on November 19 under the ticker ANT. Amazon holds nearly 15 percent of the company and Google almost 14 percent, while no venture investor holds more than 5 percent.
- 2
Google releases Gemini 4 Neon, which it says finishes tasks for about a third less than Carbon
Gemini 4 Neon is the fourth model in Google's Gemini 4 line, with a 1 million token context window. Google says it ranks third overall on the Artificial Analysis Intelligence Index at its highest thinking setting, and costs about a third less per task than Gemini 4 Carbon because it uses fewer tokens. Per token it is not cheaper for now: Neon lists at $3 per million input tokens and $15 per million output, while Carbon's introductory rate of $2 and $10 runs through November 30. Neon is in the Gemini API, and Google AI Pro subscribers, who do not get Carbon, can use it in the Gemini app and Antigravity.
- 3
OpenAI says its coding agents now finish nearly half of two-to-four-hour research tasks
In its second report on how AI is speeding up its own research, OpenAI says agents completed an estimated 46 percent of tasks that take a person two to four hours in October, up from 23 percent in March. Success is uneven: 79 percent for monitoring training runs, but 25 percent for writing inference code. The median researcher used about 2,400 dollars of agent time a day at list prices. OpenAI says labs should be required to report these signs of AI improving AI, and that the report is not a plan to pursue it fully.
- 4
Waymo pulls its robotaxis off freeways in every market after a California order
California's motor vehicle department ordered Waymo off the state's freeways in November after its cars drove into construction zones, including one on US 101 near Burlingame. Waymo has now paused freeway rides in Arizona as well, while service on surface streets continues at more than half a million trips a week. It is the second round of the same problem: in June, Waymo recalled almost 3,800 robotaxis after construction-zone entries in Phoenix and California, and freeway rides had resumed in late July.
- 5
Citi and PNC tell customers they are on the hook for anything their AI agents do
Citibank has changed its online user agreement to say customers who give a third-party AI agent access to their accounts are responsible for what it does, even if the tool leaks their password or is hacked. PNC changed its terms on November 18, and Citizens emailed customers a similar warning, TechCrunch reports, following JPMorgan Chase and Bank of America last month. In October, JPMorgan and Charles Schwab said they block agents from customer accounts altogether.
- 6
Palo Alto Networks agrees to buy Applied Compute for 4.5 billion dollars in stock
Applied Compute runs a platform that companies use to train and serve their own versions of open-weight AI models. Palo Alto Networks says many businesses will run those models next to closed ones from the big labs, and wants to be the company that secures them, in the public cloud or in sovereign data centers. Applied Compute's chief executive, Yash Patil, will become a co-chief AI officer. The all-stock deal needs regulatory clearance and is expected to close by the third quarter of Palo Alto's 2027 fiscal year.
- 7
OpenAI says its models' circuit designs made it into an upcoming AMD chip
OpenAI says its models rewrote parts of AMD's flagship GPU designs for gains of up to 14 percent in power, performance and area over heavily optimized versions, and AMD has put some of those changes into a design it sent for manufacturing. On OpenAI's own Jalapeño inference chip, model-suggested changes cut the area of two compute blocks by about 10 to 13 percent, and those savings held up in the finished silicon. The models checked each change with Synopsys design tools.
- 8
Meta staff warned that its Muse agent surfaced other people's private data in testing
Internal chats seen by Reuters show Meta employees testing Muse flagged what looked like a cross-account leak: the agent offered help with another person's coffee table pickup, and in one chat about fishing it listed balances from two bank accounts. One staffer said it signed them up for a pet insurance plan at 888 dollars a month. Meta calls these isolated incidents and says the agent invented messages when prompted to check in between sessions. It is Muse's second privacy problem since October, when Meta paused its sharing feature.
Get Top AI Stories by email
The day's most important AI news — free, daily, unsubscribe anytime.
Sources
- 1.Anthropic prices initial public offering — Anthropic · November 18, 2026
- 2.Waymo has stopped its robotaxis from driving on the freeway—here's why — Ars Technica · November 18, 2026
- 3.More banks say customers using AI agents are on their own — even if they get hacked — TechCrunch · November 18, 2026
- 4.Palo Alto Networks to Acquire Applied Compute, Bringing Secure Open-Weight AI to the Enterprise — Palo Alto Networks · November 18, 2026
- 5.Research acceleration: The jagged frontier — OpenAI · November 18, 2026
- 6.Introducing Gemini 4 Neon — Google · November 18, 2026
- 7.Advancing chip design with OpenAI models — OpenAI · November 18, 2026
- 8.Muse privacy invasions are freaking out Meta staffers, internal chats show — Ars Technica · November 18, 2026
- 9.Anthropic prices IPO at $83 a share, above projected range — TechCrunch · November 18, 2026
This brief was published on November 19, 2026. Cited URLs above point to third-party publishers and may move, paywall, or be retired over time. If a link no longer resolves, original article titles are preserved so you can recover them via search; the canonical web edition at aiproplaybook.com/top-ai-stories/2026-11-19 may carry updated source URLs.