The Top Takes from Newcomer’s Cerebral Valley AI Summit
- 01Theme 1: Compute cost is the central leverage point, driving multi-model routing and open-weight adoption
- 02Theme 2: Fight against an "intelligence duopoly" and rise of owned/post-trained models
- 03Theme 3: Nvidia leadership change (Jensen Huang's departure) reshapes the chip landscape
- 04Theme 4: AI agents are creating new security risks and new defensive requirements
- 05Theme 5: AI's public perception problem and the bubble debate
1. Key Themes
Theme 1: Compute cost is the central leverage point, driving multi-model routing and open-weight adoption
Insight: Cost of compute is the defining operating constraint for 2026
"Contemplating the year ahead, founders and investors were laser-focused on what's emerged as a key leverage point in the frantic AI buildout: the cost of compute. We heard from CEOs with various approaches to bringing down costs, ranging from purpose-built chips and application-specific inference to open-weight models and aggressive real-time price arbitrage based on how much intelligence is needed for certain tasks."
Insight: Enterprises route queries by difficulty, and the harness, not the model, is the sticky layer
"Databricks routes more than four quadrillion tokens across OpenAI, Anthropic, Gemini, and open-source models, and its new smart router sends simpler queries to cheaper models. In his view, models aren't sticky; the harnesses developers work in, like Claude Code and Codex, are."
Insight: Open-weight usage is exploding at the application layer
"At the start of 2026, open models accounted for about a basis point of the tokens passing through Factory... By March it was roughly ten basis points, by May a single-digit percentage, and it is now 'way into double-digit percent.'" Grinberg added: "In many of our customers, more than half of their tokens are to open models."
Insight: "Token-maxxing" is over; mixed model portfolios are in
"After the era of 'token-maxxing' ran up many budgets, Vercel CEO Guillermo Rauch says companies are now choosing a mix of small, large, fine-tuned, and open-weight models to save on costs."
Theme 2: Fight against an "intelligence duopoly" and rise of owned/post-trained models
Insight: Founders are building to prevent concentration of intelligence in a few labs
"'I just simply cannot see a future where we'll just have one giant AI company and we all have to ask for permission if we want to use intelligence,' he said." (Igor Babuschkin, xAI co-founder, on River AI)
Insight: Post-training your own model pencils out at scale
"Mercor is now pitching enterprises on post-training their own models, which he said generally makes sense once a company spends more than $100 million a year on inference. Customers including Cursor, Cognition, Harvey, and Sierra have already made that math work."
Insight: Big-lab concentration is already visible in the data
"A lot of this value is concentrated in the big foundation model labs, especially OpenAI and Anthropic, which accounted for 55% of the valuation total."
Theme 3: Nvidia leadership change (Jensen Huang's departure) reshapes the chip landscape
Insight: Huang acted as a financial backstop for the AI ecosystem, and that confidence is gone
"'Whenever there was hesitation and things on the bubble, he was there to say, don't worry about it, I got their back,' Ghodsi said. 'That's no longer there. So I think on the margin, it's going to start hurting a lot of players.'"
Insight: Near-term damage limited; opportunity for challengers; diversify the compute stack
"The near-term damage is limited, he said, because the chips pipeline is three years out. 'We're buying chips today that they were working on a few years ago.'" And: "a weakened Nvidia could be an opportunity for more chip startups to flourish. Founders should become more flexible about their compute stack, he said, to avoid over-dependence on a single chip provider."
Insight: Open-weight advocates face uncertainty
"Jensen Huang's resignation from Nvidia creates 'a lot of uncertainty' for open-weight advocates, Babuschkin said, calling Huang the 'leader of the pack' when it came to open-weight models."
Theme 4: AI agents are creating new security risks and new defensive requirements
Insight: Attack speeds have collapsed and no single model finds everything
"Attack times have already shortened to 18 minutes, Arora said, since agents can overwhelm some security systems. But no single provider has the model to rule them all: each cutting-edge model from OpenAI, Anthropic, or Google can successfully find about 30% of vulnerabilities, while open-source models can make up the final 10%."
Insight: Compute itself is now an attack target ("LLM jacking")
"She described it as 'LLM jacking': an organized attack, possibly by people with a crypto-jacking background, aimed at monetizing compute and tokens. 'Apparently there is a black market for that,' she said."
Insight: Agent identity and permissions are the missing layer
"Businesses won't have the luxury of keeping proprietary data out of the hands of AI agents, he said. But identity protocols will have to be added on top, so agents will be able to work on behalf of users with secure permissions." Rauch also noted: "In the past two weeks... Vercel has talked with five CISOs whose companies were breached, with attackers using swarms of agents to move through their systems."
Theme 5: AI's public perception problem and the bubble debate
Insight: Value concentration threatens the industry's social license
"'Capitalism is a privilege,' he noted, one the industry could lose if it didn't pay more attention. 'What the world is seeing is this concentration of value in a handful of companies — it's not about Silicon Valley versus the rest of the world. This is about a handful of companies versus everyone.'"
Insight: Audiences believe it's a bubble yet remain bullish on Anthropic's IPO
"79% of respondents agreed that there was an AI bubble and 67% of them thought it was likely to pop in 2027... Yet the room also thought that Anthropic's market cap after the first day of trading would sit between $1 and $2 trillion."
Insight: The capex math requires labor displacement
"About $8 trillion of capex for the AI buildout is planned for the next five years, but for that to pay off, businesses would need close to $6 trillion of annual revenue to justify those expenses, per Bessemer's data. Some of that will have to come from AI replacing parts of the labor force, Goldberg noted."
2. Contrarian Perspectives
Contrarian 1: Perplexity is not dead, and enterprise plus harness beats chatbot
"Now he says its naysayers got it all wrong, and will again. 'I would bet the same this year. If you guys short us, you're gonna lose more money next year,' he said."
Supporting reasoning: Srinivas rebuilt around infrastructure rather than a consumer chatbot, serves enterprises that can pay for token costs, and argues usage is power-law concentrated: "eighty percent of the revenue is probably coming from twenty percent of the users… It might even be ninety-ten." Perplexity also post-trains its own models rather than relying on inference providers, and says it "could IPO any time from now to 2028."
Contrarian 2: Agents will make chip design less valuable; the physical stack is the moat
"He predicted that the value of chip design would decline as agents took over, but physical-world problems like vertically integrating chips and interconnects into data center racks would remain difficult to solve." (Robert Wachen, Etched)
Supporting evidence: Etched already runs with agent leverage ("We have more agents than people in the company working on every problem 24/7"), and Wachen says "We're still in the biggest compute shortage we'll ever be in."
Contrarian 3: Skepticism of humanoids, scale-only robotics, and SF echo-chamber thinking
Younis said "the commercial model hasn't been figured out yet" for humanoid robots and concurred that "Scale is not all you need in robotics."
Supporting evidence: Applied Intuition's real demand came from unexpected places: "The surprise, Younis said, was where demand turned up: in areas like military drones rather than automobiles." He also argued SF VCs can distort outcomes: "they help, but they can also get, like, really weird outcomes."
3. Companies Identified
- Description: Data/AI platform that sits on top of hyperscalers.
- Why mentioned: Case study in multi-model routing and cost control; CEO says 2026 is a bad IPO year.
- Quotes: "Databricks routes more than four quadrillion tokens across OpenAI, Anthropic, Gemini, and open-source models, and its new smart router sends simpler queries to cheaper models."
- Description: Cybersecurity company led by Nikesh Arora.
- Why mentioned: Frames AI-agent threats; assisted Fireworks on forensics.
- Quotes: "Attack times have already shortened to 18 minutes." Qiao: "Fireworks worked with Palo Alto Networks on forensics."
General Catalyst
- Description: Megafirm VC led by Hemant Taneja.
- Why mentioned: Pursuing AI roll-ups and integrating tech into the service economy; backer of River AI.
- Quotes: "It's a strategy General Catalyst has already leaned into, with the embrace of AI roll-ups and the purchase of a hospital under his leadership."
- Description: Dominant AI chip provider; Jensen Huang's departure was a major topic.
- Why mentioned: Ecosystem backstop; investor in Flapping Airplanes and River AI; leadership uncertainty.
- Quotes: "Nvidia is also an investor in the firm." And "River AI has raised $1 billion from General Catalyst and AMP PBC, as well as Nvidia and AMD."
Etched
- Description: AI chip-to-rack startup.
- Why mentioned: Beneficiary of a more competitive chip market; agent-heavy operations.
- Quotes: "...where more than 100 former Nvidia employees work." "We have more agents than people in the company working on every problem 24/7."
Applied Intuition
- Description: Physical AI software for autonomous systems; #13 on the AI 100.
- Why mentioned: Case study on where hardware/autonomy demand shows up.
- Quotes: "sells 'any tool that you would need to design, test, and deploy autonomous systems' to 'anyone that makes a physical machine.'"
Perplexity
- Description: AI search/research company; now primarily enterprise.
- Why mentioned: Comeback narrative; strategy pivot to harness and infrastructure.
- Quotes: "Perplexity now primarily serves enterprise customers, Srinivas said, a recognition of token costs and companies' willingness and ability to pay for them."
Vercel
- Description: Developer platform led by Guillermo Rauch; runs an AI Gateway.
- Why mentioned: Signal on model demand and agent-based breaches.
- Quotes: "He said they were the most-requested models on Vercel's AI Gateway."
Crosby
- Description: AI-native law firm selling legal AI to non-lawyers; raised $85M (Lux, Index, Sequoia).
- Why mentioned: Novel vertical AI model that eliminates the billable hour.
- Quotes: "we are the first company that we know of that is selling legal AI to non-lawyers to interact with."
Harvey, Legora, Manifest
- Description: Other legal AI competitors.
- Why mentioned: Validate legal as the biggest AI market after coding.
- Quotes: "His intuition that legal services would be AI's biggest market after coding has also been borne out with a host of other companies like Harvey, Legora, and Manifest playing in the same market."
Flapping Airplanes
- Description: Research lab "devoted to making AI efficient"; Nvidia-backed.
- Why mentioned: Competing with big labs via different incentives.
- Quotes: "I think the advantage we have is that we don't have the innovator's dilemma."
Mercor
- Description: Expert-labor marketplace for training/evaluating AI models.
- Why mentioned: Case study in who trains models now.
- Quotes: "It now has experts in every occupation the Labor Department tracks across more than 100 countries, and pays them over $7 million a day."
River AI
- Description: Igor Babuschkin's startup enabling companies to post-train and serve models on their own hardware.
- Why mentioned: Open/owned-model thesis; $1B seed.
- Quotes: "Babuschkin joked that the 10-figure seed round was 'table stakes' for AI these days."
Factory
- Description: Coding-agent startup (Khosla, Sequoia backed); agent called Droid.
- Why mentioned: Real-world data on open-weight adoption.
- Quotes: "'For so many of these tasks, it is not worth six dollars per million tokens to do it,' he said."
Fireworks
- Description: Inference provider, valued at $17.5B in July; presenting sponsor.
- Why mentioned: Open-model hosting; disclosed a security incident.
- Quotes: "Qiao, whose company was valued at $17.5 billion in July, also addressed a major security incident that Fireworks, an inference provider, disclosed last week."
Bessemer Venture Partners
- Description: VC firm co-producing the AI 100 with Newcomer.
- Why mentioned: Provided market-wide valuation and capex data.
- Quotes: "The total valuation for the inaugural Cloud 100 list in 2016 was $100 billion; for the inaugural AI 100, it's 36 times higher, at $3.6 trillion."
OpenAI and Anthropic
- Description: Leading foundation model labs.
- Why mentioned: 55% of AI 100 valuation; Anthropic IPO anticipation.
- Quotes: "there was also near-unanimous optimism about the upcoming IPO of Anthropic."
CrowdStrike and Wiz
- Description: Security vendors.
- Why mentioned: Detection tools used by Fireworks.
- Quotes: "used CrowdStrike and Wiz sensors for detection."
Oracle, IREN (sponsors)
- Description: AI cloud/infrastructure providers.
- Why mentioned: Presenting/core sponsors.
- Quotes: "[Oracle] has become one of the largest AI cloud providers, handling many of the industry's most important compute deals."
4. People Identified
Hemant Taneja
- Description: CEO of General Catalyst.
- Why mentioned: Opening speaker on AI's perception problem and VC evolution.
- Quotes: "Leaders should talk less about future job losses and more about 'abundance,' such as more accessible healthcare and education, and 'agency'..."
Jensen Huang
- Description: Departing Nvidia leader.
- Why mentioned: His exit was the main subtext across talks.
- Quotes: "Jensen Huang's departure from Nvidia, it leaves a 'huge concern' about continuity in AI industry leadership."
Ali Ghodsi
- Description: CEO of Databricks.
- Why mentioned: Cost/choice insights; Genie agent usage.
- Quotes: "I'm like 10x better CEO now than a year ago."
Nikesh Arora
- Description: CEO of Palo Alto Networks.
- Why mentioned: Cybersecurity risk from agents; floated as potential Nvidia CEO.
- Quotes: "Although Shaun Maguire and other tech leaders were floating his name as a potential Nvidia CEO, Arora deflected on the question."
Shaun Maguire
- Description: Tech investor/leader.
- Why mentioned: Floated Arora as potential Nvidia CEO.
- Quotes: See above.
Robert Wachen
- Description: Etched co-founder/exec.
- Why mentioned: Chip competition, agent-heavy company.
- Quotes: "We're still in the biggest compute shortage we'll ever be in."
Qasar Younis
- Description: Founder/CEO of Applied Intuition; former Y Combinator COO, ex-GM.
- Why mentioned: Physical AI and anti-echo-chamber perspective.
- Quotes: "People are unhappy with the status quo, and I think they just want somebody to yell at."
Aravind Srinivas
- Description: CEO of Perplexity.
- Why mentioned: Contrarian comeback, enterprise pivot.
- Quotes: "There are only a few people who really know how to take the most out of these things and automate a lot of these things without being wasteful."
Guillermo Rauch
- Description: CEO of Vercel.
- Why mentioned: Cost-driven model mixes and agent security.
- Quotes: "spreading AI across many companies, rather than concentrating it in a few, is what's best for the world and for America."
Ryan Daniels
- Description: Founder of Crosby; a lawyer.
- Why mentioned: Reimagining legal services with agents.
- Quotes: "Every single person has to have a pliable mind and be comfortable with the fact that every three months, what they thought only they could do before, an agent can do or a computer could do."
Asher Spector
- Description: Co-founder of Flapping Airplanes.
- Why mentioned: Competing outside big labs.
- Quotes: "the thing we're actually the most focused on is training... it doesn't just make chips more efficient, it actually makes them more expressive."
Brendan Foody
- Description: Mercor founder/exec.
- Why mentioned: Expert-driven model training market.
- Quotes: "there can't be a duopoly on intelligence."
Igor Babuschkin
- Description: xAI co-founder; River AI founder.
- Why mentioned: Owned/open models and safety through distribution.
- Quotes: "very likely by next year, we'll have these super intelligent systems that are better than any human at any economically valuable task."
Matan Grinberg
- Description: Factory co-founder/CEO.
- Why mentioned: Open-model adoption data.
- Quotes: "it is now 'way into double-digit percent.'"
Lin Qiao
- Description: Fireworks CEO.
- Why mentioned: LLM jacking incident; agent-era forecasts.
- Quotes: "If an agent burns $50 of tokens for a $1 job, it's not that useful."
Byron Deeter and Talia Goldberg
- Description: Bessemer partners.
- Why mentioned: Presented AI 100 data.
- Quotes: "Cybersecurity and biotech were promising categories that were underrepresented, said Goldberg."
5. Operating Insights
Insight 1: Build routing and eval infrastructure, not model loyalty
Route by task difficulty and use cheaper or open models where quality allows.
"Factory's agent, Droid, separates tasks that need frontier closed models from those that can go to cheaper, faster ones." And: "In his view, models aren't sticky; the harnesses developers work in, like Claude Code and Codex, are."
Also, Qiao's warning on agent ROI: "If an agent burns $50 of tokens for a $1 job, it's not that useful." Pair this with Taneja's advice to be "more flexible about their compute stack."
Insight 2: Rebuild the business model around where money actually concentrates, and use agents inside your own company
Perplexity found revenue is power-law concentrated and pivoted to enterprise. Etched and Databricks use agents internally.
"'eighty percent of the revenue is probably coming from twenty percent of the users… It might even be ninety-ten.'" And: "We have more agents than people in the company working on every problem 24/7."
For AI-native service companies, hire for adaptability: Crosby hires "the ones who genuinely want to offload their minds to agents," and Daniels notes the need to "get rid of the billable hour."
Insight 3: Treat security as an agent-era priority; instrument for detection and add identity layers
"every company needs such tools 'because you may not even know you're being attacked.'" And: "The best way to stem attacks will be to train smart cyber systems on a specific industry's typical data flow, and then repeat the process 'one at a time' with each enterprise customer."
6. Overlooked Insights
Overlooked 1: Hiring data signals no AI-driven hiring boom, even among believers
"64% of respondents were also hiring fewer people, and not a single person said that they were hiring significantly more because of AI."
This sits in tension with the "abundance" narrative and with Daniels' relaxed view on job disruption, and it's a leading indicator of the labor-displacement the capex math depends on.
Overlooked 2: Startup-age compression and the $1B-acquisition problem
"20% of the companies on the AI 100 didn't exist two years ago." And: "with incumbents so large that a $1 billion company may no longer be an interesting acquisition target on its own, the real advantage can come from taking that technology and integrating it more directly into the broader service economy."
Together these suggest faster company formation, shifting exit dynamics, and a structural rationale for AI-enabled roll-ups over traditional acquisition exits.