What insiders really think about the AI race
Buy Anthropic. Short Perplexity. Notes from this year’s Cerebral Valley Summit. Also: Robinhood's Vlad Tenev this week on ACCESS.
I spent yesterday at
’s Cerebral Valley Summit in San Francisco. I’ve attended this event for three years in a row because Eric does a great job curating the speakers and the audience, and the conversations are more substantive than a typical industry event.This year was no exception; however, I found the most interesting part of the day to be when the results of an anonymous audience survey were shared onstage. The more than 300 attendees who participated in the survey primarily consisted of AI company founders, followed by investors, other industry professionals (including product leaders and engineers), and members of the media.
Here are the results of the survey in order of how they were shared onstage:
What will be OpenAI’s annualized revenue be at the end of 2026?
Median answer: $30 billion.
What will Nvidia be worth at the end of 2026?
Median answer: $6 trillion.
What year will an independent committee of experts, as dictated by the Microsoft-OpenAI agreement, declare that we have reached AGI?
Top answer: 2030
Which venture capital firm’s AI portfolio are you the most jealous of?
The top three most voted for, from first to last: Andreessen Horowitz, Khosla Ventures, and Sequoia.
If you could put money in any private technology companies today, what would they be?
Top five companies in order from first to last: Anthropic, OpenAI, Cursor, Anduril, SpaceX, and OpenEvidence.
What global company’s model will top the LMArena web development leaderboard at the end of 2026?
In order from first to last: OpenAI, Anthropic, Gemini, Grok, Qwen.
If you could short a $1 billion-plus valuation startup, which would it be?
First place was Perplexity. Second place went to OpenAI. Other names shown onstage: Cursor, Figure, Harvey, Mercor, Mistral, and Thinking Machines.
What stood out to me from these results (Newcomer has also published the full slides for his paying subscribers):
A softening on OpenAI: Given that Sam Altman has said OpenAI plans to end this year with $20 billion of annualized revenue, this group of AI insiders doesn’t expect next year to be as exponential for the business as the leap from 2024 to 2025. The prediction that AGI won’t be declared until 2030 suggests a lack of faith in model progress meaningfully improving in the near term, although that answer could also be clouded by the complexity of how OpenAI and Microsoft must settle on how it’s decided. (I’m still waiting for either company to share information on who its “independent committee of experts” will be and how they’ll decide.) It was also notable that more attendees wanted to buy Anthropic stock than OpenAI’s, despite the consensus being that OpenAI would lead LMArena next year.
Meta wasn’t in the conversation. It wasn’t named on the list of models likely to lead LMArena next year. The presence of a Chinese model (Alibaba’s Qwen) in the top five signals a shift that’s already underway, as many companies fine-tune open-source Chinese models rather than Llama. Meta has a lot to prove if it wants to re-enter the model race.
Perplexity is controversial. But everyone working in AI already knows that.
Other takeaways from Cerebral Valley:
What’s driving reverse acquihires? I attended a breakout session about AI acquihires, such as Meta’s deal with ScaleAI to hire Alexandr Wang and Google’s deals with Character and Windsurf. I’ve closely covered many of these deals over the past couple of years, but it was interesting to hear the group’s perspective on what drives them. Antitrust scrutiny of Big Tech certainly plays a factor, but some who have been involved in these kinds of transactions also made the point that bigger companies are racing each other to shore up talent and move faster than their competition. They have seemingly “infinite money,” as one member of the group put it, and see it as a game of placing bets on a very finite pool of talent. One AI founder in the group, who fielded multiple offers of this kind, recalled a member of a Big Tech company’s corporate development team asking him how much he wanted his startup to be valued for a deal.
No one cares about AGI anymore. At the first Cerebral Valley conference, the topic of AGI was a major throughline. A startup founder onstage said that “we’re going to be dead” by the time OpenAI releases GPT-10. This year, multiple onstage conversations noted how AGI barely registered as a discussion topic. Instead, most of the interviews focused on the business applications of AI. Multiple companies represented onstage at the first Cerebral Valley event didn’t exist and are now worth billions of dollars. There was a strain of AI bubble fear throughout the day, but mostly, everyone seemed dialed in on how they could win market share and provide products that people want to pay for.
Standout quotes from onstage interviews:
Replit CEO Amjad Masad: “If you are competing on price, then maybe you don’t have a business.”
Elad Gill: “Most companies should sell at some point. There’s often a market-maximizing moment where you’re going to get the best deal you can. A very small number of companies should never ever sell.”
S.F. Mayor Daniel Lurie: “People are starting to complain about traffic. Thank goodness. I want those complaints. We still have a lot of empty office space.”
Anthropic CPO Mike Krieger: “Time spent is, I can tell you, not on any of the dashboards that I look at. It’s just not a main consideration.”
xAI co-founder Jimmy Ba: “Knowledge is just crystalized computation from the past.”
A MESSAGE FROM OUR SPONSOR
Finance built for speed and control.
When it comes to managing company spend, founders and finance leaders often have to choose between financial responsibility and operational speed. Brex breaks that tradeoff. Brex is the intelligent finance platform that helps companies spend smarter and move faster. With Brex, you get corporate cards with built-in expense management plus a team of AI agents that handle manual finance tasks for you. That means more time back for you to focus on the business.
Sources audience survey
I’m thrilled to be a launch partner for Alltogether, a new initiative to partner brands with leading voices in independent journalism. Subscriptions will always be the backbone of my business, but advertising can also play a critical role in helping me grow. I’m a fan of Alltogether’s approach because it respects not only my readers but also my commitment to editorial independence.
Part of this effort involves getting to know my subscribers better. Please take a few minutes to complete this audience survey. (I emailed a broken link in my last issue, apologies!) Results are anonymous and not tied to your email. Thank you so much for your support!
This week on ACCESS: Robinhood CEO Vlad Tenev
This week on the show, Ellis Hamburger and I discuss our first live podcast recording in San Francisco, grandma-centric design, and how I discover new sources. Then, we’re joined by Robinhood CEO Vlad Tenev to discuss the rise of prediction markets, tokenization, juggling priorities as a public CEO, and much more.
Like and subscribe on YouTube, Spotify, and Apple Podcasts.
breaks down the last issue of Sources
Shoutout to
and for dissecting yesterday’s scoop about OpenAI. You can watch the full clip above. (I never reveal my sources!)More links
- and ’s excellent interview with Satya Nadella that I linked yesterday is rightfully making the rounds They got the Microsoft CEO to open up about the terms of his new deal with OpenAI (it seems like no love is lost). This is by far the most remarkable quote to come from OpenAI’s largest shareholder with deep IP rights to its tech: “I can make the argument that if you’re a model company, you may have a winner’s curse. You may have done all the hard work, done unbelievable innovation, except it’s one copy away from that being commoditized.” Their full conversation is worth watching, but don’t miss Patel’s complimentary deep dive on Microsoft’s AI strategy and infrastructure buildout.
Cursor 12x’d its valuation in less than a year and is now making. A notable addition its its cap table is Google, which already reverse acquihired Windsurf. Another remarkable figure: $1 billion in annualized revenue with just over 300 employees.
Google has shown off SIMA 2 (“scalable instructable multiworld agent” just rolls off the tongue), a video-game-playing agent. It was trained on people playing No Man’s Sky and Goat Simulator 3. As I’ve covered already, expect to hear a lot more about world models and these kinds of agents in the coming months.
China state-sponsor hackers used Anthropic’s Claude to target dozens of entities with phishing techniques and malware, “literally with the click of a button, and then with minimal human interaction.” Anthropic’s full report on the investigation is worth checking out.
Polymarket has apparently started beta testing its US exchange. The rumors that most of its trading volume comes from bots are about to be put to the test!
