Skip to content
ai0.news
Go back

AI News — November 19, 2026: Anthropic Prices at $83 Above Range, Google Neon Drops Same Day

Listen to this briefing

Chapters (9)

Good morning. Anthropic priced the largest IPO in history overnight at a $1.5 trillion valuation, Google answered with Gemini 4 Neon aimed straight at Opus 5.5, and Microsoft used Ignite to argue it can deliver 96% of frontier performance at 5% of the price. If you had “trillion-dollar AI lab goes public the same week three competitors try to undercut it” on your bingo card, today’s for you.

Anthropic goes public at $83. Anthropic priced its IPO at $83 per share, above the $70-78 range, raising $81.3 billion and surpassing SpaceX as the largest IPO ever by proceeds. The fully diluted valuation lands around $1.5 trillion — below the reported $2 trillion target but well above the $965 billion private mark — with Amazon holding ~15% and Google ~14%. TechCrunch notes the first-day trade will set the tone for OpenAI and the rest of the private AI book. HN was less reverent: one commenter pointed out Apple didn’t hit $1.5T until 2020, and several flagged that employees being allowed to sell immediately is not typically a confidence signal.

Google ships Gemini 4 Neon. DeepMind released Gemini 4 Neon at $3/$15 per million tokens with a 1M context window, ranking #3 on the Artificial Analysis Intelligence Index and #2 on the new Agent Arena Index for steerability. It’s pitched directly at Opus 5.5 territory — slightly cheaper, and notably faster on OpenRouter at ~127 tokens/sec versus Carbon’s 46. A third-party analysis projects Neon winning ~48.3% of head-to-head matchups against Opus 5.5, though it leans heavily on extrapolation. HN reactions split between “first Google model that actually feels good for coding” and skepticism that Gemini has historically looked strong on benchmarks and faded in long-horizon agentic work. Dropping this the day Anthropic prices feels pointed, even though Google owns 13% of the company going public.

Microsoft’s seven-model Ignite dump. Microsoft used its Ignite keynote to announce seven new MAI licensed-weight models, headlined by a claim that MAI Code 2 hits 96% of Opus 5.5’s coding performance at 5% of the cost. Early testers on HN say Code 2 is genuinely good, but the broader reception was skeptical about Microsoft’s position — comparisons to Oracle and Windows Phone came up more than once, and several commenters said GitHub Copilot’s pricing changes have damaged developer trust in ways a cheap model can’t fix.

Nvidia stops shipping H200s to China. Interim CEO Colette Kress told analysts on the earnings call that licensed Hopper data center products are no longer being shipped to China — a reversal of the White House-approved deal Jensen Huang secured last December before resigning. Context helps: roughly 600,000 H200s are reportedly stuck in Hong Kong awaiting Beijing’s import approval, and vetting of Chinese buyers had already become a mess. The China revenue line is now effectively zero.

Genie 4 opens to Ultra subscribers. Google launched Genie 4, its interactive world-generation model, with real-time text-prompted environment editing, native audio, and a community gallery. Access is gated to Google AI Ultra, which produced the usual HN frustration about Workspace accounts being locked out and no public demo. One commenter noted the demo video carefully avoids panning away from objects and back, suggesting object permanence is still shaky.

OpenAI gets into chip design. OpenAI published details on using its models for RISC-V work and an IP partnership with AMD on GPU design, pitched as accelerating EDA workflows. HN readers asked the obvious question — are the RISC-V designs actually being shared? — and noted the pattern of AI labs moving into every one of their customers’ industries while keeping the enabling models private. One commenter put it dryly: this is how Nvidia fuels its own demise.

AI agents outperform KPMG juniors. European Business Magazine reports that KPMG’s “Project Exponential” deployed over 100 AI agents on a FTSE 250 audit, with the firm’s audit head acknowledging the AI sometimes asked better questions than human staff. HN was unimpressed: no quantitative metrics, KPMG grading its own homework, and — as one commenter noted — experienced auditors’ value is knowing what to dig into, which is exactly the judgment call the headline glosses over. Another observed Big Four audits are typically staffed by overworked juniors, so the bar isn’t where the press release implies.

That’s a lot of pricing pressure landing in one 24-hour window for a company that just asked public markets to value it at $1.5 trillion. Tomorrow we find out what the ANT ticker does when it opens.

Get this in your inbox

One post every morning. Unsubscribe anytime.


Share this post on:

Previous Post
AI News — November 20, 2026: Anthropic Closes at $97 on Nasdaq, Gemini 4 Neon Tops Coding Index
Next Post
AI News — November 18, 2026: GPT-6.2 Cuts Task Costs 5x, Nvidia Board Locks Out Jensen Unanimously