• The Deep View
  • Posts
  • Anthropic's IPO is a trillion-dollar bet on trust

Anthropic's IPO is a trillion-dollar bet on trust

Welcome back. CES 2027 in January will put AI's shift to the center of the tech ecosystem on full display, from smart glasses to coffee makers and far more. But a larger AI footprint also raises a practical question: How will AI actually make our lives better? OpenAI’s latest research offers a different reality check. Agents are accelerating AI development, but original ideas and complex engineering still depend heavily on people. The path to self-improving AI remains uneven, and its potential benefits come with serious questions about risk. Meanwhile, we examine what the Anthropic IPO could mean for the company, its ambitions, and the growing expectations around the business of AI. —Jason Hiner

IN TODAY’S NEWSLETTER

1. Anthropic's IPO is a trillion-dollar bet on trust

2. Why self-improving AI still depends on humans

3. CES shows how AI became tech's default

MARKETS

Anthropic’s IPO is a bet on trust, not technology

Anthropic officially landed on the Nasdaq on Thursday, marking a watershed moment for the AI industry—even as many, many questions remain. 

The maker of the popular Claude chatbot, Claude Code, and enterprise-friendly AI models such as Mythos, Fable, and Opus, set a base price of $83 per share, closed up 17% at $97.17 on its first day of trading, and raised $81 billion to fund future growth.

The company's valuation is a $1.27 trillion market cap, which immediately vaults it into the top 10 most valuable companies in the US stock market, coming in the 10th spot right behind Tesla. (On a fully diluted basis, the company is worth $1.7 trillion, which would rank it 7th). 

It's going to need that $81 billion to pay for the data center commitments the company has made to fund the growing number of users of its products in the years ahead. Still, Anthropic is not the money-burning furnace that other AI companies are. It reportedly turned its first operating profit in June, according to The Wall Street Journal, and has projected that it will have positive EBITDA in 2028.

The fact that Anthropic went public before its archrival, OpenAI, which has been around longer, is a testament to the fact that Anthropic has been almost solely enterprise-focused since Claude launched in 2023. And the enterprise is where the lion's share of revenue is being generated in AI today. 

Despite Anthropic's successful debut on the Nasdaq, there are major questions swirling around the AI trade in the market and the course of the AI industry in general:

  • Earlier this year, Anthropic's IPO was expected to reach $2 trillion at its debut. Is the discounted price a reflection of investors' wavering confidence in the future of AI?

  • When OpenAI's IPO arrives in 2027, where will it land in comparison to Anthropic and is it reasonable for the two of them to be valued at the same level as tech companies that took decades to create trillions of dollars in value?

  • Where will the broader AI trade go for the rest of 2026 and into 2027 now that Nvidia is in turmoil from the resignation of CEO Jensen Huang, the AI boom's No. 1 evangelist?

  • The recent mid-term elections in the US were overwhelmingly won by anti-data center candidates. Will that put constraints around AI expansion across the country in the critical years to come?

  • As I wrote in October at the Gartner Symposium, 81% of CIOs expect to increase AI spending in 2027 while only 11% of CFOs report that they are seeing clear returns from AI investments. Will enterprises close this gap and keep investing more resources in AI to the benefit of Anthropic and other enterprise AI vendors?

Anthropic's market debut comes at a time when AI models are commoditizing and enterprises are turning toward efficiency, with a focus on model routing and using less expensive models, small language models, and domain-specific models. Even Anthropic itself tacitly acknowledged the trend with its recent release of Fable 5.2, which builds in its own routing features that can delegate tasks to faster, lower-priced models like Sonnet and Haiku. As a result, when we think about the future of Anthropic, it's likely to be less about selling lots of metered access to enterprises for the most advanced and most expensive AI models that money can buy. Instead, think about Anthropic as the company that has created a brand around being the safest, most trustworthy, and most business-friendly AI company in the world. In the long run, that's likely to be enormously more valuable than creating the most advanced models, especially in a world that's still so distrustful of the future of AI.

Jason Hiner, Editor-in-Chief

TOGETHER WITH CHARGEBEE

Field notes on the new, new CFO org

Finance teams are moving from static spreadsheets to real-time dashboards—vibe coding deal desks, wrangling MCPs and databases for custom automations, deploying agent swarms to triage invoices, and reimagining every repetitive workflow with AI. 

At Chargebee’s Traceable summit, CFOs, controllers, and finance engineers from Vercel, PostHog, Public.com, Celigo, and Campfire shared hands-on field notes on shipping AI-native finance workflows. Every panel and practitioner playbook is now streaming on-demand.

RESEARCH

Why self-improving AI still depends on humans

Though frontier AI is becoming increasingly capable as it inches closer to being able to improve itself, not all capabilities are advancing equally. 

On Thursday, OpenAI released a report examining the state of research acceleration within the lab, tracking how coding agents have changed the way that AI is developed as the question of recursive self-improvement or the ability for AI to develop and improve itself, still looms. The findings show that, while AI is speeding up certain parts of model development, these agents still struggle with complex engineering tasks and coming up with original research ideas. 

The lab noted that transparency about "indicators" of RSI is "essential to inform a public discussion of the capabilities, risks, and required safeguards for increasingly powerful AI; however, our reporting should not be read as an intention to pursue full RSI."

However, it's undeniable that agents are getting better at doing their research: 

  • According to OpenAI, internally, agents now successfully complete 46% of "two-to-four-hour" research tasks independently as of October, up from 23% in March. Meanwhile, tasks that take four to eight hours of human work now see a success rate of 35%, up from 13% in March. 

  • However, that success varied depending on the task. Agents were successful 79% of the time at monitoring training runs, 35% of the time at launching training runs, and only 25% of the time at writing code for inference and serving stacks. "Overall, uplift is jagged, and humans remain a core part of the AI research loop," OpenAI said in its report.  

  • Humans, meanwhile, are also still responsible for "research taste," OpenAI said, as these models are currently not good at coming up with scientific ideas on their own, something that could create a "bottleneck for recursive self-improvement," the company mentioned. 

While OpenAI's research points to an uneven course of reaching self-improving AI, a recent essay from the DeepMind Institute may see it as the key to advancing even further. On Thursday, Google's research arm published an essay exploring the paths to artificial general intelligence and, beyond this, artificial super intelligence. The essay notes that recursive self-improvement could dramatically accelerate the path to these breakthroughs through things like code and model architecture design, data generation and agentic collaboration. 

However, DeepMind notes that RSI itself may be jagged, as the feedback loops that allow these models to self-improve can easily "stall or degenerate" if the feedback itself is poor. "RSI should advance fastest in domains that allow for automated verification without humans in the loop, such as mathematical proofs, while other domains might lag behind or taper off sooner," the essay reads.

Recursive self-improvement has become a hot-button topic in the AI industry, as many buy into the promise that creating a way for these machines to rapidly improve themselves could open the door for technological and scientific breakthroughs beyond human imagination. However, with that kind of power comes an incredible amount of risk. It's why California governor Gavin Newsom outlined rules around RSI in his regulatory framework published earlier this week, and why many of the industry's safety advocates have been calling out the issue of containment and potential loss of control of these systems. The labs themselves understand this too, hence why OpenAI explicitly noted in its essay that its research does not mark an intention to pursue RSI, and why Anthropic called out the dangers of RSI in research that it published over summer. Though this research shows that AI has a long way to go to reach this level of capability, with so many calling out the dangers, the question that we should be asking shouldn't be If it's possible, but rather if the benefits of it outweigh the risks.

Nat Rubio-Licht

TOGETHER WITH SIMA.AI

Build Physical AI Fast: Our Engineers Show How

SiMa.ai’s Palette 3.0 lets developers run any Physical AI model on a single Modalix MLSoC, combining vision models, VLMs and LLMs in one application within strict power limits.

On Dec. 2 at 8 AM PT (5 PM CET), the engineers behind the release go under the hood in a free live webinar. Learn how to tune new cameras without recompiling code, run multiple models side by side and get up to 3.5x more tokens/sec with speculative decoding. See how Mixture-of-Experts support brings GPT-OSS-20B to a 16GB module, and how optimized BEV kernels speed up camera–LiDAR fusion.

Bring your questions for the live Q&A. Can’t make it live? Register anyway and we’ll send you the recording.

CONSUMER

CES shows how AI became tech's default

If you need another reminder that AI has taken over the tech world, look no further than CES.

This January marks the 60th anniversary of CES, the world's largest tech show. The annual trade-only event, held in Las Vegas, has become a staple of the tech community, as well as a bellwether for the next big wave of tech hardware innovation. This year, unsurprisingly, the star of the show is AI.

CES has long been known as a showcase for the latest gadgets, with department store-sized booths from every major electronics maker. In recent years, though, AI has steadily moved to center stage. 

At a CES media luncheon in New York, Brian Comiskey, senior director of innovation and trends at the Consumer Technology Association, the trade group that produces the conference, offered a look at what attendees can expect: 

  • Vertical AI: This is AI tailored to a specific industry or product, offering deeper expertise than a general-purpose chat assistant

  • Agentic AI: This is the next step past call-and-response AI assistants, where it can actually take action for users, though this use case is still nascent. However, CTA data shows how much AI has already influenced the enterprise: In the US and Korea, 70% of the population has used AI at work. They are followed closely by the Netherlands (69%), Hong Kong (68%), France (68%) and Italy (66%). 

  • Semiconductor ecosystem: These are the companies responsible for providing the GPUs, CPUs, memory, and storage that power the AI ecosystem. Comiskey noted that 80% of the largest semiconductor companies in the world take part in the show, including AMD. 

  • Physical AI: This category includes humanoids, dexterity, and autonomous vehicles, including demos from Zoox. 

"The age of intelligence—that's what 2027 will probably be remembered as—is where AI really hits capacity," added Comiskey.

Beyond the broader trends, he offered a preview of AI hardware expected at CES, including smart glasses and AI glasses. He highlighted that both Halliday and Snap's AR brand, Specs, would showcase their latest glasses technology on the show floor. AI will also make its way to everyday products in simpler ways, able to optimize everything from TVs to coffee makers. 

"The home is serving as a microcosm of all the different technology trends that we see on a day-to-day basis, whether that's security, energy, AI, and it's coming increasingly into our control to be able to pack our day-to-day lives," said Comiskey.

There seem to be countless tech conferences each year, as you’ve probably noticed from The Deep View’s coverage and how many our team attends. But CES is different: it’s the industry’s flagship event, encompassing both consumer and enterprise innovations. Nearly every major company, Apple being the notable exception, shows up to unveil what’s coming in the year ahead. The show is so central to the industry that simply walking the floor gives you a real pulse on where tech is headed. When I first attended four years ago, AI hardware and AI-focused news were mostly confined to corners like Eureka Park, the show’s startup showcase. As the category has grown, so has its footprint.

LINKS

  • Live Dictation: OpenAI launches it in ChatGPT with no time limits

  • Genie 4: Available in Project Genie, bringing native audio, real-time editing, and a new agent to create interactive, virtual worlds

  • Slack: Unveiled many updates, "Workflows, Slackbot, and AI agents working as one"

  • Perplexity Computer: Brain automatically creates self-improving agent skills for users

  • Gemini: Google introduces computer Use in the Gemini App for Mac

GAMES

Which image is real?

Login or Subscribe to participate in polls.

POLL RESULTS

Have you tested out using persistent agents in your day-to-day work?

Yes (15%)
Not yet, but I’m interested (44%)
No, and I don’t plan to (34%)
Other (7%)

The Deep View is written by Nat Rubio-Licht, Sabrina Ortiz, Jason Hiner, Faris Kojok and The Deep View crew. Please reply with any feedback.

Thanks for reading today’s edition of The Deep View! We’ll see you in the next one.

“[This image] has better reflection.”


“Reflections and shadows on the water were more realistic in [this image].”


“[This image] needed to be cropped, scenery around the boat in the water distracts from the action: The boat floating in water.”


“The small dock in the background looks realistic.”

“[This image] looked so much better, so I guessed it was enhanced. Crazy logic...”


“AI use very high contrast color to make it look real.”


“Not enough variation of trees and water in [this image]”


“Waves in the water are too perfect.”

If you want to get in front of an audience of 750,000+ developers, business leaders and tech enthusiasts, get in touch with us here.