SpaceX’s expected record-shattering June IPO will test the public markets’ faith in Elon Musk — and AI. Y Combinator and Ron Conway’s SV Angel show how volume can be a winning strategy in startup investing. Amazon backs Anthropic with another $5 billion and loads of compute. Chinese model-maker DeepSeek is raising outside capital. AngelList offers retail investors a chance to back private startups with debatable terms. Bob Iger returns to investing at Thrive Capital. Meta & Microsoft announce AI-efficiency layoffs and buyouts. Bill Gurley reminds everyone the AI boom is still heavily subsidized by VC cash.

The Main Item

SpaceX Financials Show How Faith in Elon Powers Musk Inc.

The SpaceX IPO, now expected in June, has become the black hole of technology and finance.

It’s sucking up the attention and the dreams — and soon the dollars — of everyone in its path. Elon Musk’s agreement to buy coding star Cursor for $60 billion sometime down the line, or pay $10 billion, marked only the latest unusual twist.

Data from SpaceX’s confidential S-1 filing has been dribbling out (itself an unusual phenomenon) and it’s all very… unprecedented!

The $75 billion SpaceX aims to raise is an order of magnitude more than what even today would be considered a massive IPO.

The company is a newly constituted combination of the old SpaceX, the old Twitter, and the new xAI, and its financials don’t remotely support the $1.5 trillion to $2 trillion valuation it’s expected to seek. Yet the Elon Musk reality distortion field has been a feature more than a bug for quite some years now (see Tesla). He believes his investors will follow him anywhere, and who’s to say he’s wrong?

Risk factors in an S-1 are doom-and-gloom by design, but SpaceX’s warnings are especially stark. “Yeah, we don’t really know if any of this will work,” would be a fair summary.

It’s widely assumed that Musk will be looking to combine a newly public SpaceX with Tesla. In the meantime, he’s trying to get SpaceX included in major stock market indexes right away. The prospect of the IPO being followed by a steep price decline as private investors cash in their winnings looms over the process.

No one has seen anything like this before.

It’s definitely a banker bonanza, with Morgan Stanley, Goldman ⁠Sachs, Bank of America, Citigroup, and JPMorgan Chase all serving as ​bookrunners and 16 other banks involved in smaller roles.

Here are some key factoids that have emerged from the S-1, mostly via Reuters and The Information:

• Musk holds a bit over 40% of the equity ahead of the IPO. Google could see its $900 million SpaceX investment from 2015 turn into more than $100 billion. Founders Fund (Luke Nosek, since departed), Valor Equity Partners (Antonio Gracias), and Fidelity all stand to win big, along with Sequoia Capital (Shaun Maguire), DFJ (Steve Jurvetson, since departed), a16z, Nvidia, and the Qatar Investment Authority.

• Merging SpaceX with xAI swung the company to a loss. The combined entity lost $4.94 billion last year on $18.67 billion in revenue, Reuters reported, after heavy investments in xAI infrastructure. In 2024, SpaceX had $791 million in profit on $14.02 billion in revenue. Capex exceeded $20 billion last year, with more than half going to AI development.

• Starlink is carrying SpaceX’s revenue. The Information reported that the satellite internet service brought in $11.4 billion in 2025, about 60% of the company’s total, with EBITDA margins of 60%. Rocket launches generated $4.1 billion in revenue, and AI accounted for $3.2 billion.

• Insiders are set to retain control. SpaceX will use a dual class structure that gives Class B shareholders — including Musk and a handful of other insiders — ten votes each, while the Class A shares will be worth one vote each. SpaceX will retain its “controlled company” status after going public, which means that it will not require a majority independent board.

• Retail investors may not have much board representation, but they will get a seat at the cap table. Musk is allocating 30% of IPO shares to retail, or around three times the norm, with banks assigned to specific pools or “lanes”: Morgan Stanley is handling smaller retail investors, Bank of America is serving family offices and high-net-worth individuals in the US, and UBS will take on these investors abroad, Reuters reported. Around 1,500 retail investors were invited on a roadshow tour of SpaceX’s Starbase facilities in Boca Chica, Texas, once it kicks off in earnest on June 8.

All of this is unfolding as Musk juggles the capital demands of his broader empire. Tesla’s Q1 earnings surprised analysts Wednesday with a projected $25 billion in capex — $5 billion more than predicted, and twice its previous peak.

The question hanging over all of it is whether the market can absorb Musk’s grand ambitions.

Bankers are thrilled that SpaceX could reignite the IPO market more broadly, but they’re pushing to get their own IPOs out ahead of SpaceX — or give it some breathing room and wait until after the summer.

Since buying SpaceX is really buying Musk Inc., there could be a rotation by Elon-stans out of Tesla and into SpaceX — not what Musk is looking for. The Nasdaq is eager to get SpaceX on board, but that could lead to concentration risk.

All of this will no doubt set precedents and influence the timing of the next turn of the wheel: the expected mega IPOs of OpenAI and Anthropic. It will certainly put the durability of the AI boom — or bubble — to the test.

Newcomer Podcast

Amanda Askell on AI Consciousness

Anthropic AI safety researcher Amanda Askell came on the Newcomer podcast to break down one of the most existential conversations in AI to date: can AI systems like Claude become conscious, and if they do, what do we owe them?

Listen to the Podcast

Cerebral Valley Show

Have the Vibes Turned Against Anthropic?

On the Cerebral Valley Show this week, Eric, Max, and James debate whether the “vibes” are actually turning against Anthropic or if this is just noise from power users, and later unpack a rough stretch for Sam Altman.

One Big Chart

Accelerators & SV Angel Top List of Seed Investors that Backed $100 Million Companies

Volume can be a very good strategy in the seed investing business.

Dealroom published a list of funds that have written the most seed-stage checks to startups that eventually reached $100 million in revenue, and it’s little surprise that the storied startup accelerator Y Combinator came out on top with 94 companies so far. The accelerator backs around 500 to 600 startups a year nowadays.

Ron Conway’s SV Angel has taken a similar small-check-wide-net approach, with around 50 to 100 new investments a year.

Sequoia Capital stands out as the most successful non-exclusive seed or accelerator fund.

Five Notable Deals

Anthropic, VAST Data, Blue Energy, Reliable Robotics, Omni

Amazon doubles down on its investment into Anthropic with more funding and compute capacity in a week otherwise filled with infrastructure and hardware deals.

• Amazon agreed to invest an additional $5 billion into Anthropic, with an option to put up $20 billion more as part of an agreement to build AI infrastructure.

• Data infrastructure startup VAST Data raised around $1 billion in Series F funding co-led by Drive Capital and Access Industries. Fidelity, NEA, and Nvidia also participated, among others.

• Nuclear power plant startup Blue Energy raised $380 million in combined equity and debt funding led by VXI Capital. Other investors include Engine Ventures, At One Ventures, and Tamarack Global.

• Reliable Robotics, which builds autonomous systems for aviation, raised $160 million in a new funding round led by Nimble Partners. Other investors include Eclipse, Lightspeed Venture Partners, Coatue, Pathbreaker Ventures, Island Green Capital, Socium Ventures, AE Ventures, RTX Ventures, Presidio Ventures, UP Partners, KAS Venture Partners, What If Ventures, Calm Ventures, Gaingels, and Mana Ventures.

• Omni, an AI analytics platform, raised $120 million in Series C funding led by Iconiq. Other investors include Theory Ventures, First Round Capital, Redpoint Ventures, and GV.

Stories We’re Reading

DeepSeek Seeks Funding, AngelList Pushes Private Investor Limits, Iger Returns to Thrive, Meta & Microsoft Slash Jobs, Anthropic Burns Figma, Gurley Reframes Compute Shortage

Read more

Tuesday afternoon, Cursor CEO Michael Truell convened an all-hands meeting to make an announcement about its new partnership with SpaceX, according to a person familiar with the matter.

Shortly afterwards, SpaceX followed with a carefully worded post on X, which announced that Cursor had “given SpaceX the right to acquire Cursor later this year for $60 billion or pay $10 billion for our work together.”

The announcement caused a whiplash that rippled through the tech world — and was also slightly vague. But there are a few things to keep in mind, both on the specifics of this announcement and how we got here.

Read more

Big news. We’re heading back across the pond.

Last year, we brought the Cerebral Valley AI Summit to Europe for the first time.

London was too good to do just once. We’re returning with another world-class lineup of AI leaders, founders, and investors for what’s shaping up to be one of our best summits.

Apply to Attend

Over the years, CVAI has earned a reputation as the summit where real things happen — new partnerships, fresh funding rounds, and deals you’ll read about later.

Databricks’ acquisition of MosaicML for $1.3 billion came after Ali Ghodsi met Naveen Rao at our first Cerebral Valley AI Summit. Suno’s $250M Series C began at our summit the following year.

We’ll be hosting the Cerebral Valley AI Summit in the heart of London on Wednesday, June 24.

We’re thrilled to announce our first batch of premier speakers, including:

• Alex Mashrabov, founder & CEO, Higgsfield

• Grant Lee, founder & CEO, Gamma

• R. Martin Chavez, board member, Alphabet

• Jesse Zhang, founder & CEO, Decagon

• Luciana Lixandru, partner, Sequoia Capital

• Aidan Gomez, founder & CEO, Cohere

• Oege de Moor, founder & CEO, XBOW

We’ll announce more speakers in the coming weeks. You won’t want to miss them.

The event is hosted by me, Eric Newcomer, along with Weekend co-founders Max Child and James Wilsterman.

Special thanks to our first set of sponsors — Felicis, Baseten, and Nebius — for helping us bring together this leading global AI summit.

Apply to Attend

• To see Newcomer’s full event calendar, click here.

• To reach out with questions or to sponsor the event, email riley@newcomer.co.

• To learn more about our past Cerebral Valley AI Summits, go to cerebralvalley.com. We will be back in San Francisco with our voice summit in May and our anchor event in November.

The Cerebral Valley AI Summit is an invite-only event. We welcome founders, investors, AI executives, and members of the media to apply to attend. The summit will bring together a couple hundred top AI leaders from around the globe for an intimate, one-day event.

Next up, the Cerebral Valley Voice Summit in San Francisco on Wednesday, May 6. Reach out if you’re interested in participating.

Our look at 13 polls shows that Americans are scared of AI taking their jobs, and don’t trust our own government to regulate it. Kara Swisher enters the Newcomer Podcast studio to skewer Silicon Valley. Electric truck maker Slate Auto leads an otherwise slow week for funding rounds. Jensen Huang spars with Dwarkesh over Nvidia’s chip sales to China. Accel raises a fresh $5 billion in funds, while Sequoia rakes in $7 billion for late-stage bets. Google is back in talks with the Pentagon to license its AI models. Anthropic’s Mythos could spur governments to regulate AI.

Plus, we get ready for the Cerebral Valley Voice Summit on May 6.

The Main Item

Recent Polls Show Job Fears & Lack of Oversight Driving Negative Views of AI

It’s no secret that AI is not as popular with the American public as it is in Silicon Valley. Facing a political backlash, industry leaders are now fretting about whether they’ve painted the wrong picture. In light of that, we thought it worth a close look at what public opinion polling says — and doesn’t say — about attitudes towards AI.

We reviewed 13 polls published since September of last year. All but two are from well-known mainstream polling organizations, and all but one have sample sizes of 1,000 or more.

Subscribe now

Our four biggest takeaways from the data:

• The biggest fear about AI, hands down, is that it will take people’s jobs. Americans are using AI more in the workplace, but both blue- and white-collar workers are wary of its impact on their livelihoods and careers.

• There is a strong appetite for AI regulation. The demand for oversight is broad and bipartisan, although Democrats are increasingly concerned about AI while Republicans are trending slightly towards less concern, per Pew Research polling from 2020 to 2025.

• Most Americans don’t have strong opinions on the data center buildout in general, even as local opposition mounts to the data center next door. Democrats are more hostile to data center construction in their communities than Republicans, according to Politico data.

• “AI experts” (defined as AI conference presenters or authors with technical or applied AI expertise) are neutral to positive about AI tools and much less concerned about job destruction compared to the average American. Only 39% of AI experts think AI adoption will bring about fewer jobs, compared to 64% of Americans.

A complete list of the polls we used, with links to their data, is at the bottom of this post. We found the polling from Quinnipiac, Pew Research Center, YouGov, Politico, and Stanford/ Ipsos to be the most interesting and have a further breakdown below.

Quinnipiac, March 2026: White-Collar and Blue-Collar Workers Alike Think AI Will Decrease Available Jobs. Gen Z Is the Most Pessimistic

Across employed Americans, 71% of white-collar workers and 73% of blue-collar workers think AI advancement will lead to fewer job opportunities, according to the poll, which surveyed 1,397 US adults. Despite their pessimism, Gen Zers were using AI the most: YouGov found that Gen Z had the greatest familiarity with AI tools in their workplace, with 51% of Gen Zers saying they used AI weekly at their jobs.

Pew Research Center, September 2025: Partisan Opinions Have Converged Towards the Center

At the moment, the American public’s AI skepticism is solidly bipartisan. Republicans had been historically more wary of AI, while Democrats viewed the technology more favorably, but now both have effectively met in the middle, according to polling from the Pew Research Center published at the end of last year.

Outside of political affiliation, the progressive think tank Data for Progress found that support for AI is more split across age, race, and gender groups. Women view AI unfavorably by a 10 point margin, while men view it favorably by 16 points. White voters are slightly unfavorable towards AI at -3 points, while Black and Latino voters are favorable of the technology by 29 points and 10 points, respectively.

Gallup, September 2025: AI Safety Is Seen as More Important Than Rapid Progress

An overwhelming majority of Americans would prefer many more guardrails around the development of AI technology, even if that means AI progress would slow.

Some of this could be due to fears around a dangerous all-powerful AI that CEOs have warned about: the December YouGov survey found 77% of Americans are concerned that AI could be a “risk to humanity,” with 39% saying they were “very concerned.”

Politico, February 2026: US Adults Are Still Making Up their Minds on Data Centers

Polling around data centers is fuzzier, but Politico’s February survey found that a majority of American voters are undecided or are slightly positive about data centers coming to their communities.

Trump 2024 voters are slightly more likely to support a data center buildout near them, while Kamala Harris supporters are more likely to oppose such construction, but both sets of voters are largely undecided.

Politico didn’t survey why people were opposed to data centers, but in Quinnipiac’s similar research, 72% of respondents who oppose data center construction said electricity costs were their biggest reason, while 64% said water use and 41% said noise. Research has shown that water usage concerns are largely overblown, but the narrative remains quite sticky.

Per Pew’s relatively recent polling on data centers, “Americans tend to think that they’re bad for the environment, home energy costs, and quality of life,” said Pew senior researcher Colleen McClain, but they do see positives in data centers’ impact on local jobs and tax revenue.

Stanford AI Index, April 2026: US Sentiment on AI Is Much More Negative Than in Most Other Countries

Compared to the rest of the world, US adults lean more apprehensive rather than excited when it comes to AI development. Per Ipsos’ 2025 survey, the US is on the low end of getting “excited” about new AI products at 38%, and 64% of Americans are made nervous by the increasing use of AI. In China, by contrast, 84% of those surveyed were excited about AI.

“Not only are they willing to use the technology and learn about it, but they’re more willing to welcome it into their national economies and trust that their politicians are going to do the job of protecting them,” Stanford AI Index lead Sha Sajadieh told us.

The US also showed the lowest trust in its own government’s ability to regulate AI, with only 31% of respondents feeling confident about it.

What’s the Bottom Line?

Taken together, the data paints a fairly consistent picture that economic disruption is the clear standout in what makes Americans the most nervous about AI. Notably, this polling did not differentiate between Americans’ feelings on the tech industry at large relative to AI.

An NBC News poll on Americans’ attitudes towards political figures and topics showed AI polling lower than ICE, but it did not break out tech beyond artificial intelligence. Gallup polling from the end of 2025 showed that 59% of Americans sampled view “the computer industry” positively or somewhat positively, which was a drop from a high of 75% in 2017 but still overall above water.

Politically, distrust of AI is strikingly bipartisan: both sides are looking for clearer guardrails and regulation around the technology. The Trump Administration’s deregulatory philosophy, supported by much of the industry, looks distinctly out of step with public sentiment in this regard. Americans also don’t trust their own leaders to work on the problem, though.

One very telling but very simple finding from all the polls: experts and the public are living in entirely different realities about what AI means for society. Until that gap closes, the wariness isn’t going anywhere.

For our analysis, we looked at the Pew Research Center’s survey of 5,023 US adults from September 2025, Gallup’s poll of 3,128 US adults from September 2025, Future of Life Institute’s poll of 2,000 US adults from October 2025, Navigator’s poll of 1,000 US adults from December 2025, YouGov’s survey of 1,287 US adults from December 2025, Gallup’s poll on US views of internet companies from the end of 2025, the Tech Oversight Project/Morning Consult’s poll of 1,800 US voters from January 2026, Data for Progress’ poll of 1,228 likely US voters from February 2026, Politico’s poll of 2,093 US adults from February 2026, NBC News’ poll of 1,000 US voters from March 2026, Quinnipiac’s poll of 1,397 US adults from March 2026, Gallup’s poll of 23,717 US adults from April 2026, and the Stanford AI Index’s findings from a 2025 Ipsos poll of 23,216 adults from 21 countries, including additional data from China.

Cerebral Valley

We’re Less Than Three Weeks Away from the Cerebral Valley Voice Summit on May 6. Apply ASAP to Secure Your Spot.

Time is running out to land a spot in the room at the first ever Cerebral Valley Voice Summit.

Apply Here

We’re bringing together the top founders building AI voice applications and infrastructure, along with leading venture capitalists, members of the media, and other AI notables.

We’ve already secured a stacked line-up of speakers and discussion leaders, like Sierra’s Bret Taylor, Runway’s Anastasis Germanidis, AssemblyAI’s Dylan Fox, Abridge’s Shiv Rao, a16z’s Olivia Moore, Cartesia’s Karan Goel, and Emergence Capital’s Jake Saper.

You won’t want to miss this one-day summit covering AI’s next big paradigm shift.

Apply Here

The Cerebral Valley Voice Summit is co-hosted by Eric Newcomer, founder of Newcomer and Max Child and James Wilsterman, the co-founders of Weekend.

Newcomer Podcast

Kara Swisher Unfiltered on Elon, Trump & Silicon Valley’s Biggest Lie

This week on the podcast, Eric sat down with legendary tech journalist Kara Swisher for a heated conversation on the state of Silicon Valley.

They discuss a whole host of topics, including her early prediction of tech’s rightward shift, why she thinks Silicon Valley has become addicted to playing the victim, her takes on Elon Musk, Mark Andreessen, and Peter Thiel, why she would rather have Ted Cruz making decisions about AI than the tech industry itself, and what she thinks is coming next.

It’s one of our spiciest episodes to date. Give it a listen.

Listen To The Podcast

Five Notable Deals

Slate Auto, Beeline Medicines, Sygaldry, Terremoto Biosciences, Slash

A slow week for big AI rounds pushed biotech and quantum startups up the rankings for funding hauls.

Read more

After a long day working on his prized new product, an autonomous fighter jet, Armor Harris was ready for a beer. As we sat at a cafe near the company’s Dallas headquarters in early March, the 33-year-old product guru at defense tech firm Shield AI gazed at the storm clouds gathering over the flat Texas farmlands and told me why he makes weapons.

“According to the Bible, you realize we’re in the last couple of years here in the grand scheme of things,” he said.

Harris has seen the signs — the creation of modern Israel, nuclear weapons, artificial intelligence — and believes there’s a “reasonable” chance that Jesus returns in our lifetime, and will be supporting nations that are on the “good side” when the end of the world comes.

Subscribe now

“Ultimately, it’s one of the reasons — maybe the main reason — why I’ve become passionate about working on defense for the last couple of years,” he said. Later, he added: “If you have pretty strong faith in the way that things are going to go, then would you not want to work on the things that will be most impactful to the course of human history?”

Harris, 33, is not a preacher. He’s an engineer — and an exceptionally talented one, according to colleagues who worked with him at SpaceX, where he spent a decade and won the confidence of Elon Musk. He joined Shield AI two years ago as its senior vice president of aircraft, charged with fixing and revitalizing the product line-up of an eleven-year-old AI drone startup that had spent most of its life in the shadow of better-funded, better-known competitors.

Read more

Anthropic dominates the conversation with its Mythos model, and spearheads a new cybersecurity consortium. Founders & investors at HumanX are wowed by frontier model progress, while Databricks and Snowflake spar onstage. Q1 data shows VCs getting much busier. Data center infrastructure and defense tech companies score growth rounds. Meta launches a new model & Amazon sees an AI payoff coming. OpenAI’s Stargate fades, but its infrastructure bets could yet pay off. John Carreyrou of Theranos fame takes on Satoshi Nakamoto. Private credit stares down the SaaS-pocalypse.

The Main Item

Anthropic’s Cyber Initiative Brings Welcome Collaboration on AI Risks

Anthropic’s limited release of its too-dangerous-to-touch Mythos model this week is a watershed in more ways than one.

It puts to rest any lingering worries about scaling walls: Industry executives and investors we spoke with this week were almost uniformly impressed with the progress of the frontier models in general and Mythos in particular — and suspect most people don’t realize what’s coming.

Mythos also solidifies Anthropic’s new-found status as the clear industry leader. Amazingly, the company is apparently set to leapfrog OpenAI in revenue run-rate, leaving many VCs kicking themselves.

Perhaps most importantly, Anthropic’s latest reveal is prompting some much-needed industry cooperation on AI risks. If generative AI is as powerful as promised — maybe it’s already AGI? — that comes not a moment too soon.

Anthropic apparently wasn’t ready to reveal Mythos at all until an accidental leak forced its hand. It then said that it had built a model powerful enough to systematically find and fix vulnerabilities in almost any piece of software — even those that had escaped notice for years — and was limiting access to trusted partners.

The company announced an industry consortium, dubbed Project Glasswing, to collectively address the cybersecurity threats inherent in state-of-the-art AI. If a model can write complex software without human help, after all, it can also spot even small problems in any stack of code. Google, Microsoft, and Amazon, among others, are all on board.

Steve Schmidt, chief security officer at Amazon, told us he was optimistic about the work ahead. “AI has helped attackers become more prolific,” he noted. “It’s up-leveled the skills of the bottom tier. But AI can really help the defenders.” He singled out Claude’s ability not just to spot weaknesses in software, but to reliably develop even complex fixes without human intervention.

Cybersecurity all but demands collective action by companies and the government, but it’s never been easy. Companies don’t like to talk about their cybersecurity problems, lest they alert attackers, competitors, customers, investors, or government regulators. It was only in 2023 that the SEC required companies even to disclose major breaches, let alone work together to prevent them. The cybersecurity industry, though steadily growing, often seems stuck in a never-ending game of whack-a-mole.

Government agencies have been unreliable partners too. Oftentimes, when cyber researchers at the NSA or the CIA discover a software vulnerability, they are inclined to keep it to themselves for their own hacking purposes, rather than inform the company so that it might be fixed. The decision on how to handle such situations has been governed by a so-called “equities” process that tries to balance risks and benefits, but it’s often contentious.

Though defending US companies, government agencies, and individuals against cyber attacks would seem to be the government’s job, at least in part, President Trump has a grudge against the main agency charged with that mission, the Cybersecurity and Infrastructure Security Agency (CISA). He blames it failing to endorse his false claims of election hacking in 2020 and now wants to slash its budget by $700 million, on top of earlier cuts (and a punitive investigation of its former director). CISA, unsurprisingly, is said to be in disarray.

Cybercrime, meanwhile, has grown into a massive worldwide industry. Ransomware alone yielded an estimated $57 billion for criminals last year and continues to disrupt a wide range of institutions. Foreign adversaries, including China, North Korea, Russia, and Iran, have widely penetrated commercial and government systems, with consequences that are still largely unknown. It’s easy enough to imagine how AI would amplify the threats.

Treasury Secretary Scott Bessent and Fed chair Jerome Powell summoned bankers to an urgent meeting this week on the potential dangers of Mythos, Bloomberg reported.

It’s more than a little ironic that it’s Anthropic leading a critical industry-wide security initiative even as the US government tries to designate it a “supply chain risk” and ban it from government systems. Anthropic had been negotiating to limit how its AI could be used for autonomous weapons and surveillance, but its bigger sin was likely opposing Trump’s anti-regulation agenda. (It’s now fighting the designation in court, with mixed results.)

On the one hand, the Trump Administration has insisted in the Anthropic case that private companies can have no role in deciding how the government uses their technology. On the other hand, it’s ideologically opposed to regulations and other measures that might help address AI safety risks, leaving private companies little choice but to take matters into their own hands.

Anthropic is doing Trump a huge favor by at least trying to get ahead of the cybersecurity issue and moving cautiously with its new model. Cooperation and competition don’t have to be a zero-sum game. The AI era will certainly require both.

HumanX Dispatch

Everyone Is Back on Team Foundation Model

Scaling laws are holding up. Reinforcement learning and other techniques are proving effective. LLMs are improving even more quickly than anticipated — especially at Anthropic.

That was the consensus among investors and executives at the big HumanX AI conference in San Francisco this week.

Anthropic’s industry-leading tech and impressive financial performance popped up constantly on the heels of the Mythos announcement and its revenue run-rate surpassing $30 billion. Investors were again wondering where AI applications would fit in: One told us that they were now doubting the prospects of highly valued legal AI companies in light of Claude’s capabilities.

Carta, the cap table software company, is literally building their product into Claude’s desktop application, so users can run queries and ask questions about fund management without switching applications, its CEO Henry Ward told Eric. Zoom CTO Xuedong Huang said an integration with Claude would let customers put meeting notes seamlessly into their Claude workspace.

Vinod Khosla (who is firmly Team OpenAI, of course) said he’s confident the big models will see massive growth, since it’s still very early days for large enterprise adoption.

A few more highlights:

• The Databricks and Snowflake rivalry was on full display, with Ali Ghodsi dismissing Snowflake as being “in the rearview mirror” during his panel. Sridhar Ramaswamy insinuated that Databricks’ financials were all over the place. Both companies are losing the war of attention to the foundation model labs.

• Multiple speakers said the public is massively underestimating AI’s future effects. Andrej Karpathy made a case on X on Thursday that it’s partly due to the paid capabilities of foundation models, particularly coding tools, far outstripping the free tier of Claude and ChatGPT that most people outside of tech are most familiar with.

• Despite all of the rampant AI optimism, many attendees were still questioning if SpaceX, OpenAI, and Anthropic will be able to pull off their IPOs as planned later this year or next. Many people we spoke to acknowledged the risk of an AI bubble.

The Newcomer team made a group showing at the conference. Eric, Tom, and Madeline caught up with many folks including Arvind Jain at Glean, Victor Riparbelli at Synthesia, Qasar Younis at Applied Intuition, Asheem Chandna at Greylock, Anna Barber at Everywhere Ventures, Rahul Vohra at Superhuman, Bhavin Shah at ServiceNow, and Josh Constine at Signalfire.

We were also very happy to host our Cerebral Valley After Dark party with Weekend at Saint Joseph’s Art Society. We set up 6 poker stations in a nod to the conference’s Vegas roots and to entice tech’s biggest poker fans.

Thank you to Weekend and to our sponsors — Airwallex, IBM, and IREN — for making the night possible.

A big congratulations to our winner Janice Chen from Mammoth Biosciences, who cleaned up.

Newcomer Podcast + Cerebral Valley Show

Jim Lanzone Talks Yahoo Resurgence, Cerebral Valley Show On OpenAI Buying TBPN

It’s a two-for-one podcast week at Newcomer! Jim Lanzone, longtime tech leader and current CEO of Yahoo, stopped by the podcast studio and highlighted his company’s comeback.

On the Cerebral Valley Show, Eric, Max, and James break down the TBPN acquisition, Anthropic’s likelihood of going public anytime soon, and the rise of Fruit Love Island and other AI slop content.

Listen To The Podcast

Two Big Charts

VC Activity Leaped in Q1 — and Not Just for Mega-Deals

This last quarter shattered venture funding records (depending on when you consider certain deals closed), though much of the dollar total is from just a few big deals.

Now new Crunchbase data shows that the uptick in activity was broad-based, with most firms doing far more deals than in the comparable quarter a year ago.

Accel stepped up its game, writing 16 checks in Q1, with Andreessen Horowitz and Lightspeed close behind at 15 and 14.

Read more

I’m in the air, bored, flying to San Francisco for HumanX, so here’s a rundown of everything that’s been happening, with links to key stories, excerpts, and our quick thoughts.

Read more

The massive SpaceX IPO, and expected offerings from OpenAI and Anthropic, will depend on retail enthusiasm as markets wobble. Two startup founders from the ET30 list come on the Newcomer Podcast. Snap gets an activist investor, a bit too late. Q1 VC fundraising smashes records, though with asterisks. Defense tech shipbuilder Saronic raises $1.75 billion. Mercor hit by damaging data breach as Anthropic loses some source code. OpenAI buys a tech-friendly talk show. Anthropic buys a biomedical startup. Doug Leone leans back in at Sequoia, taking a new “chairman” title.

AI Public Offerings Face a Down Market. But Fear & Greed May Yet Carry the Day.

A great paradox of today’s financial markets is that they have never been more heavily weighted towards the biggest players — and never been more dependent on the emotions and fortitude of retail traders.

The prospective IPOs of SpaceX, OpenAI, and Anthropic will put their faith to the test.

By historical market logic, the IPO window is currently closed. It’s always hard if not impossible to go public when stocks are in a down cycle. This year was supposed to be a big one for IPOs, but so far it’s been a bust. A lot of the companies that have managed to get out recently are trading below their offering prices, and the herd of private unicorns seeking an exit remains large.

The appetite for a piece of the top AI companies seems insatiable, even if it comes via an SPV at an inflated price. Private markets are the new public markets, and shrewd (or wealthy) individual investors are finding their way into the hottest trades.

OpenAI, though still private, is now letting them in directly: the extraordinary $122 billion in funding it announced this week included an allocation for wealthy clients of the big investment banks. Elon Musk is looking to cut out the middlemen too. SpaceX, now combined with xAI, confidentially filed its IPO prospectus this week and will be looking to raise $75 billion — the largest IPO ever by a wide margin — while setting aside a third of the shares for retail investors.

Normally, companies want institutions to own most of the shares post-IPO, because they’re less likely to flip for a quick profit. Musk is looking for retail to be that kind of anchor — and why not?

These are not your retail investors of yore. They have seen their irrational loyalty to Tesla — the ultimate meme stock — pay off in big way. They’ve envied those who got in early on Bitcoin. They’ve watched the big VCs tout the AI companies as trillion-dollar winners, and they’re not going to miss out.

This is an era where sports betting and prediction markets sit alongside equities and government bonds in the same apps. Crypto and meme stocks have taught a whole generation that it’s dumb to sell into a downturn and fundamentals are meaningless as long as there’s momentum. Every recent market swoon has been followed by a quick “buy-the-dip” rally.

Institutional investors have now heeded the lessons of Tesla and Bitcoin: don’t fight the crowd, however irrational it looks. They dare not sit out the SpaceX IPO, as The Information reported, and may face similar pressures with OpenAI and Anthropic.

The private capital markets are of course very eager to tap the new-found risk appetite of individual investors. They’ve already gotten the green light to market private equity and VC investments to 401K plans, arguing that everyone should be able to get in on private market gains.

Of course, the sky high IPO valuations mean a lot of the money will already have been made by the time regular retail investors get a crack. They won’t be buying low, but instead will be counting on eventually selling at valuations in the multiple trillions. Similarly, it’s hard to see 401K plans getting an early piece of the next big winners.

The risk, of course, is that the new-style retail investors also run in packs — meme stocks all but depend on it. They are hard to spook (see Tesla), but when sentiment shifts it can go hard in the other direction. Look no further than the current private credit meltdown: retail investors were lured into not-very-liquid products, and when risks emerged and they realized how non-liquid they were, a bank-run dynamic ensued.

Secondary market traders are reporting that OpenAI shares are suddenly wanting for buyers as enthusiasm shifts to Anthropic, another example of volatile sentiment.

Mom and pop investors — or maybe the better simile these days is Robinhood bros — are always at risk of being the greater fools, the ones left holding the bag when a financial mania runs its course. If they lose AI faith, they won’t be the only ones feeling the pain.

Newcomer Podcast

Two Stars of the ET30 Power Rankings

This week we published the ET30 list of the enterprise startups VCs are most excited about in collaboration with the venture capital firm Wing. Eric sat down with two of the top-ranking enterprise tech founders for the latest episode of the Newcomer Podcast.

First up is Han Wang, CEO of Mintlify, the knowledge infrastructure platform that quietly powers the docs for Anthropic, Lovable, and thousands of other companies. We found out their servers crashed overnight because of OpenClaw before Wang even knew what it was.

In the second half, Eric talks to Jesse Zhang, CEO of Decagon, ranked #4 on the late stage list in a category that includes some of the most well funded names in enterprise AI to talk all things voice AI.

It’s two of the most exciting bets in enterprise AI right now, in one episode.

Listen To The Podcast

Founder Hubris

Activist Foray at Snap Shows What Got Away

Well, it finally happened.

Snap, the fading star of the social era, has gotten a serious activist investor who wants to shake things up. And they’re taking direct aim at CEO Evan Spiegel and his stable of passion projects and unchecked control.

The activist, Irenic, disclosed this week it bought up 2.5% of the company’s class A shares and has put together an authoritative sounding plan (along with a surprisingly cheery website SaveSnapNow.com) to rescue the company. Irenic has some legitimate credentials in the hedge fund world among its founders, including one from feared activist Elliott Management.

The hedge fund’s core argument is that Snap, at nearly 1 billion monthly users, is one of the few scaled social media companies out there — behind only Meta and TikTok. But the company has profligate side projects and a bloated head count. Meanwhile it argues the app’s users are under-monetized.

The initial demands are for Spiegel to spin out or shut down its struggling AR glasses unit Spectacles. It also wants to cut a big chunk of its employee base, similar to Meta in its quest for efficiency. Down the list Irenic wants to improve corporate governance by giving class A shareholders 1 vote per share. Currently Snap’s Class A public shareholders have no voting rights at all.

Snap as a public stock has been, to put it kindly, a dog. It currently trades at $4.63 a share, well off its pandemic peak at $83 and its 2017 IPO price of $17. The company’s glory days, when its innovative product seemed a threat to Facebook and Instagram’s dominance, are a distant memory.

Spiegel’s vice-like grip over the company’s voting structure has left him completely unchecked and able to pursue idées fixes like Spectacles, which simply do not contribute in a meaningful way. Snap has long been a mirage to investors — Irenic among them — who see its large user base and dream of what could be.

But the deep irony of Irenic’s activist stake is that they’ve admitted their powerlessness themselves: “This is your company,” they write to Spiegel. “We can only attempt to persuade you.” Indeed, the only true option would be a board revolt, which feels unlikely given how personally close Spiegel is to people like board chairman Michael Lynton.

“Snap welcomes input from all shareholders and regularly engages with investors on strategy, capital allocation, and governance,” Lynton told CNBC in a statement.

The story of Snap is ultimately a cautionary tale of how innovation can grow stale in a hyper-competitive space and changing times can leave promising stars by the wayside — especially when their leaders have absolute power. Snapchat still matters insofar as people really do love it as a messaging product. But the company’s story is a sad one. And it feels already written.

One Big Chart

Q1 Was the Largest Fundraising Quarter of All Time — Depending on When You Consider a Round Closed

Read more

Anthropic has quietly purchased a Dimension-backed stealth startup called Coefficient Bio in a deal worth just over $400 million in stock, according to a letter that I’ve obtained.

The startup was working on AI models for biological research. It was pursuing nothing short of artificial superintelligence for science.

They are joining Anthropic’s Health Care Life Sciences team led by Eric Kauderer-Abrams.

The company, which had stayed under the radar and was formally founded only eight months ago, was half-owned by the venture capital firm Dimension.

Samuel Stanton, one of Coefficient’s co-founders, tweeted in January as part of a recruiting pitch, “We’re ushering biopharma into the Intelligence Age. It will change everything about how the industry learns and makes decisions.”

Dimension is boasting a 38,513% IRR on the investment. (I first broke the news in 2022 that two Lux Capital partners were leaving to join with a partner from Obvious Ventures to found Dimension.)

Dimension and Anthropic declined to comment for the story.

Anthropic’s previous acquisitions include Bun and Vercept.

It’s hard not to marvel at the contrast between Anthropic’s science-forward acquisition and OpenAI’s media-oriented purchase in the headlines today.

I’ve obtained Dimension’s letter to its limited partners. Newcomer subscribers can read it here:

Read more

Think Y Combinator is overrated and overvalued? We’ve heard it all before. Investors still won’t stop writing checks.

Last week’s YC demo day had VCs buzzing about several startups from the Winter 2026 batch.

This time around, there were fewer companies that could be referred to (in derogatory terms) as “ChatGPT wrappers” — especially compared with a couple of years ago. AI infrastructure for enterprise was a big theme this batch, as was AI tooling for specific industries, like law and finance. Startups building the infrastructure for robotics also made a good showing.

Read more

We’re a month away from the Cerebral Valley Voice Summit, and our agenda is filling up with voice AI leaders.

We’re bringing together the top founders building AI voice applications and infrastructure, along with leading venture capitalists, members of the media, and other AI notables.

Big things happen at our Cerebral Valley AI Summits: Databricks bought MosaicML for $1.3 billion after Ali Ghodsi met Naveen Rao at our first Cerebral Valley AI Summit. Suno’s $250M Series C began at our Cerebral Valley AI Summit.

Apply to attend now and see if you can get your startup’s name added to our highlight list.

Apply To Attend

Today we’re announcing 8 new speakers:

• Russ d’Sa, founder & CEO, LiveKit

• Scott Stephenson, founder & CEO, Deepgram

• Karan Goel, founder & CEO, Cartesia

• Olivia Moore, partner, Andreessen Horowitz

• Jeff Liu, founder & co-CEO, Assort Health

• Justin Uberti, head of realtime AI, OpenAI

• Grace Isford, partner, Lux Capital

• Tanay Kothari, founder & CEO, Wispr Flow

We’ve already announced a bunch of big-name speakers, including Sierra’s Bret Taylor, Runway’s Anastasis Germanidis, AssemblyAI’s Dylan Fox, Abridge’s Shiv Rao, and Emergence Capital’s Jake Saper.

Today, we’re also excited to announce our first batch of discussion group leaders. These breakout sessions have become a highlight of our summits where attendees participate in curated conversations.

The founders and investors leading these discussions include:

• Alex Levin, founder and CEO, Regal

• Connor Zwick, founder and CEO, Speak

• Lily Clifford, founder and CEO, Rime

• Jeremy Kaufmann, partner, Scale Venture Partners

• Morgan Blumberg, partner, M13

• Priya Vijayarajendran, CEO, ASAPP

Finally, we’re thrilled to have Baseten joining as a sponsor. They’re joining Felicis and a few sponsors still to be announced.

We’re hosting a dinner with CTOs of top voice AI companies ahead of the summit with Baseten, and a speakers’ dinner after the summit with Felicis.

We still have a few sponsorships available, so if you’re interested, reach out to riley@newcomer.co.

If you want to attend the summit on the next great interface shift, spots are filling fast. Apply now for a chance to secure a seat.

Apply To Attend

The two-man race is among the most dependable sources of drama in Silicon Valley. Uber v. Lyft, Rippling v. Deel, Brex v. Ramp — the path toward dominating a new field goes through a high-cost land grab with a directly competitive product.

The latest is a quietly fierce battle between legal tech startups Harvey and Legora. Harvey, co-founded by attorney Winston Weinberg, has jumped out to an early lead, scooping up the biggest investors in Silicon Valley and signing on more clients. But Legora is a fast-rising follower, started by Swedish engineer Max Junestrand (pictured), whose investors believe its deeper enterprise product will overtake Harvey in the near future

The big Silicon Valley VC firms are loaded up on one side or the other: Harvey has Sequoia, a16z, Kleiner Perkins, and Coatue; Legora has Benchmark, Bessemer, General Catalyst, Accel, and ICONIQ.

The two startups recently traded fundraising announcements: Harvey last week said it raised $200 million at an $11 billion valuation, following Legora’s late February raise of $550 million at $5.5 billion.

Harvey has more than $200 million in annualized revenue, which is around double Legora’s rate, according to people familiar with their finances. But Legora, which was started a year later, is growing faster, one of its investors tells me.

The dynamic has gotten chippy. Junestrand, a bright and canny 26-year-old, wouldn’t refer to Harvey by name in our conversation. He instead spoke vaguely of the competition while insisting he was paying it no heed.

“If we were in a professional swimming race, I would say that Legora is looking down and other people are looking to the side, losing speed,” Junestrand said.

The even-keeled Weinberg, 31, cast Legora as opportunistically trying to draft off Harvey’s success. I asked Weinberg if he often found clients weighing between the two offerings. “—And like 30 others,” he quickly responded.

“We’re pitted against each other because — second mover, that’s what you want to do. You want to make sure that you’re part of all the press releases of the bigger player because it kind of brings you up,” Weinberg said.

Entertaining as the two-man race may be, the real competition for both is coming from all sides. Harvey and Legora are built on top of foundation models from OpenAI, Anthropic, and Google; the specter of Claude and ChatGPT enterprise devouring both of their businesses looms over them as it does all companies that could be tagged with the AI wrapper label.

Outside of that, there are a series of more niche legal AI startups going after specific practices. Weinberg’s “30 others” was barely an exaggeration; other fast rising AI startups in the space include:

• EvenUp, focused on personal injury law

• Eve, for plaintiff law firms

• Spellbook, for contract drafting and review

• Darrow, for class action litigation

• Summize, for contract lifecycle management

Read more

What do a developer documentation startup, an IT agents platform, a voice model provider, and Anthropic all have in common?

They’ve got investors frothing at the mouth.

We partnered with the venture capital firm Wing for the second year in a row to survey top venture capitalists about what startups they think are the most promising right now. The ET30 list is broken up into early, mid, late, and giga stages. The name has become a misnomer — 60 companies made the list this year.

This year, Mintlify, Serval, ElevenLabs, and Anthropic took first in their respective categories.

Agent platform Dust, voice-powered note-taking application Granola, frontend platform and vibe coding tool Vercel, and infrastructure juggernaut Databricks took second place across early, mid, late, and giga.

With investors at most of the top venture firms participating, the survey provides a snapshot of VC sentiment, highlighting startups that are quietly fundraising, categories that are attracting investor dollars, and also indicating where enthusiasm has cooled.

You can see fun rivalries play out on the list:

• In the battle of customer support startups, the less highly valued Decagon beat out rival Sierra by two ranks.

• In legal, Harvey and Legora were neck-and-neck ranking 14 and 15 respectively on the late stage list.

• Anthropic beat out OpenAI this year, ranking first on the giga list compared to OpenAI’s fourth place.

• AI infrastructure provider Baseten ranked 10 on the late stage list with rival Fireworks AI coming in at 12.

• Neither Deel nor Rippling made the list this year. Corporate credit card startup Ramp ranked seventh on the giga list. Software companies need to figure out an AI story or lose the investor zeitgeist.

• OpenEvidence ranked third in the late stage list, but Abridge did not secure a ranking.

• Even with the massive outcome for Scale AI, data supply companies are getting largely snubbed. Mercor, which ranked fourth on the mid stage list last year, did not make the cut this year, nor did Surge AI. Only David AI, which ranked 11 on the mid stage list, got recognized for its work in voice data.

Wing’s Peter Wagner pointed out that the list features “a large number of companies that are building products that assist the development of agents — a group that didn’t exist a couple years ago.”

AI infrastructure is booming even as application companies are still getting their footing. In many cases, investors are more bullish on the infrastructure providers than the applications building on top of them.

On the application front, investors have suddenly opened their hearts to vertical applications. “It wasn’t long ago that venture capital would have thought these were killing fields,” Wagner observed. The wisdom used to be: “vertical applications — you can’t make money doing that.” Basis is building in accounting. Avoca answers calls, texts, and emails in home services. Pace provides AI operations for insurance carriers, brokers, and wholesalers.

Taking a step further, venture capitalists are even starting to make bets on full-blown services businesses that are selling outcomes instead of seats. That trend, which investors have been watching closely, is best represented by Crosby, number six on the early list, which is building its own law firm.

The startups on the list are grouped by how much money they have raised: early stage is up to $35 million, mid stage is $35 million to $150 million, late stage is $150 million to $1 billion, and giga stage is over $1 billion.

Voice companies did extremely well. ElevenLabs topped the late stage list and Granola ranked second on the mid stage list. (Newcomer is hosting the Cerebral Valley Voice Summit in San Francisco in May.)

You can read Wing’s full report on the ET30 here.

Read the report

In a presentation to prospective investors in January, Coatue offered a rare look at how it justifies Anthropic’s heady valuations.

An anonymous source sent me the slides, which give a unique look at how a top investment firm values Anthropic’s business. This is a moment in time when some of America’s most valuable, growing technology companies remain private.

In the January presentation, Coatue estimated that Anthropic would lose $14 billion in EBITDA on $18 billion in revenue in 2026, ending the year with $30 billion in annualized run-rate revenue.

Most recently, it’s been reported that Anthropic is at $19 billion in ARR, which could suggest Anthropic is already running ahead of Coatue’s expectations.

By 2031, Coatue projected that Anthropic would be generating $48 billion in EBITDA profits on $200 billion in revenue, with $224 billion of ARR. That would give Anthropic a 41x forward EBITDA multiple at a projected $1.995 trillion valuation in 2030.

Coatue told the investors it was pitching that Anthropic could be worth $2.413 trillion in 2031.

In February, Coatue co-led a $30 billion Series G investment in Anthropic that valued it at $380 billion. Coatue declined to comment on this story.

I first shared this slide and talked about the presentation on my new podcast, the Cerebral Valley Show, which I host with my CVAI co-hosts Max Child and James Wilsterman.

I’ve got a few other slides from the Coatue presentation for Newcomer subscribers.

Read more

The AI race with China had even Jamie Dimon endorsing industrial policy at Tuesday’s Hill & Valley Forum. OpenAI deems Sora, and Disney, a ‘side quest.’ San Jose Mayor Matt Mahan says Dems need to focus on affordability without attacking tech. Defense tech heavyweight Shield AI raises $2 billion, while AI applications earn some investor love. Kleiner Perkins lands $3.5 billion in early and growth stage funds as comeback builds under Mamoon Hamid. Sen. Bernie Sanders files his moonshot AI data center moratorium bill. Two juries find Meta responsible for social media harms.

The Main Item

Industrial Policy Gets New Life as Tech Eyes Government Incentives

The remarkable transformation of erstwhile free-market capitalists into industrial policy boosters is all but complete, if Tuesday’s Hill & Valley Forum in Washington was any indication.

For at least half a decade, some of the Founders Fund and Andreessen Horowitz investors who’ve embraced the Hill & Valley ethos have been arguing that America needs to shore up its ability to build things.

But they were in a minority. Most business leaders were entrenched in the belief that the government shouldn’t be “picking winners and losers” by directing taxpayer dollars or other incentives to specific companies or technologies.

Now, no less a personage than Jamie Dimon, the honorary chairman of corporate America (pictured), has found himself on board with President Trump’s enthusiasm for deploying government money and power on behalf of the tech industry. The competition with China demands it.

“I think you have to use industrial policy now,” said Dimon, while hedging that he came to that view “reluctantly.”

This new bipartisan consensus holds that onshoring manufacturing and energy production, opening the checkbook for military technology, and building lots of AI data centers are all part of the same mission to strengthen America against its geopolitical rival.

General Catalyst’s Hemant Taneja said it was an urgent matter, noting that only 11% of all factories globally will be in the US by 2030. A subtext of the day was that the federal government should support these endeavors.

There was notable silence on the extraordinary corruption that’s characterized the second Trump presidency — a vivid contrast to the virulent criticism the Hill & Valley cohort directed at President Joe Biden for a host of purported sins. The Iran war was also sparsely discussed.

Founders Fund GP Trae Stephens gave what sounded like a political stump speech on the tech industry’s patriotic duty to America and its values. “When founders start companies, they should ask how does this strengthen the country that is making this success possible?” he declared.

Loyalty to the government, though, apparently depends on favored policies. Stephens said in a fall podcast that tech leaders turned on Biden when they realized that “all of this time and effort and money that they put into courting the Democratic establishment didn’t really give them the connectivity to drive any of the change that they wanted to see.”

Another big theme at the conference was worry about Americans’ bad attitude towards artificial intelligence, as evidenced by AI polling worse than ICE.

Vinod Khosla predicted that “the single biggest issue in the 2028 election will be fear of AI.” Sen. Mark Warner warned that youth unemployment could hit 30 to 35 percent within four years and economic anxiety at that scale would fuel populism from both the left and right. That could threaten the entire AI transition. “If we don’t figure this out we’re going to get screwed,” he said.

There wasn’t much reflection on how the industry’s actions, or mounting disapproval of their ally in the White House, might be affecting public perceptions.

Anthropic CEO Dario Amodei, who consistently cites AI risks including mass job-losses, was backhand blamed for spreading doom about AI. Yet at the very same time he was being subtly dragged by his rivals, US District Judge Rita F. Lin in San Francisco was observing that the government’s designation of Anthropic as a “supply chain risk” looked like punishment for disagreeing with the White House on AI regulation.

Late Thursday, Lin blocked the supply chain designation, ruling that the administration’s actions amounted to “classic illegal First Amendment retaliation.” Anthropic, she wrote, was being knee-capped for the crime of “asking annoying questions.”

Perhaps the public doesn’t support the Hill & Valley agenda in part because they see it as self-serving, and don’t appreciate the “connectivity” with an unpopular administration whose affections are based on fealty and money. That’s not just a messaging problem.

AI in Hollywood

With Sora’s Demise, OpenAI Priorities Are Clear

The team at OpenAI that built Sora supposedly named the AI video generation tool after the Japanese word for “Sky.” It was meant to signify its limitless potential.

Naming something after the sky and thinking of it as limitless potential feels like a bad idea — mythologically speaking. Sora flew for about a year before OpenAI melted the wings. The sun in this case was app chief Fidji Simo’s crusade against so-called “side quests.”

In the wreckage there are some important lessons about OpenAI and Hollywood’s relationship with AI video.

OpenAI is now a couple weeks into its hard pivot into being an enterprise oriented company, jettisoning projects like ballast from a hot air balloon. The company is locked in a major battle with Anthropic to win the agentic coding and API businesses — allocating GPUs to a resource-hogging, essentially revenue-free product like Sora is a luxury the company cannot afford.

We were skeptical of Sora 2’s early success last fall. Despite its quick rise to the top of the app charts, it was a gimmick, and in the purest sense, true slop. It failed as a social media tool even though OpenAI execs were already thinking out loud of ways to monetize creators.

In truth, Sora was just another in the long line of short-lived image generation fads out of OpenAI. Remember when everyone was using Dall-E to rip off the “Studio Ghibli” aesthetic for their family photos? Nobody does.

Well, that’s not entirely true. There’s one group of people who remember the Studio Ghibli moment, and Sora as well: Hollywood. It was only last December that OpenAI signed a deal with Disney to incorporate their characters into Sora in exchange for a $1 billion “investment” in the company (really it was just a warrant for Disney to buy up $1 billion in OpenAI shares).

When OpenAI killed Sora, Disney was given a less than 24 hours’ notice, Reuters reported. That deal is a blemish on former CEO Bob Iger’s record — he was a venture partner at Thrive, which is a major OpenAI shareholder. (Iger is pictured above with Thrive’s Josh Kushner.) It’s yet another example of how little people in Silicon Valley think about their counterparts down south. The tradition of disrespect goes back decades.

We were skeptical OpenAI ever cared about Hollywood. A company whose core technology rides on a cavalier interpretation of IP rights is not going to fit well into the temperamental and litigious entertainment industry. Also, OpenAI employees never really knew what they wanted with Sora. One executive at a movie studio told us he’d get different answers from different OpenAI employees — one would say it was all for robotics; another would say it was a consumer AI video play; and another would say it was indeed an effort to infiltrate the moviemaking process.

But this hardly foretells the death of AI’s integration into movies. Sora was not a top-of-the-line video model. Google’s VEO, Runway ML, Luma Labs, and increasingly the Chinese-made Seedance are better. Even if the film purists hate it, Hollywood is embracing AI at an accelerating rate.

OpenAI is just not in the mix. The company and Hollywood have gone their separate ways, with little love lost between the two. Brynn Mooser, the CEO of AI studio Asteria, gave us the epitaph: “Sora 2 was AI slop. Thank God it’s gone.”

Newcomer Podcast

Matt Mahan on Fixing California, Fighting Trump & Running for Governor

Matt Mahan is the Mayor of San Jose and is running a long-shot campaign for Governor of California.

He is also rare among prominent Democrats in his willingness to say out loud that California’s failure to fix housing, homelessness, and energy costs has handed the MAGA movement its best ammunition.

Eric sat down with Mahan to get into why California has spent $20 billion on high speed rail and delivered nothing, why the billionaire wealth tax will backfire, and how San Jose reduced homelessness by a third without raising taxes.

Listen To The Podcast

One Big Chart

AI Startups are Divided on How Best to Price Products

The question for many startups on how to price AI tools has been a big one. Unlike SaaS, running AI for several thousand users can quickly run down margins, so the old playbook of selling subscriptions doesn’t neatly apply.

Still, few AI companies are switching to a fully usage-based pricing structure, data in ICONIQ’s 2026 State of Go-To-Market report shows. The exception is companies whose core product is proprietary AI model development, though even 16% of these companies are using only usage-based pricing, while 50% have a hybrid approach.

AI application companies are also predominately hybrid, but more have stuck with the traditional subscription model.

Read more

For 15 years, Bessemer Venture Partners followed a different strategy from its global peers in India, backing local startups from its global funds, and keeping its in-country team small.

That changed in 2021, when Bessemer raised a dedicated $220 million India fund, followed by a second $350 million fund in 2025.

New documents obtained exclusively by Newcomer shed light on how Bessemer has fared in India over the years. A 57-page presentation prepared for current and prospective LPs in mid-2024 shows the tradeoffs of the firm having invested less capital than rivals Sequoia, Accel, and Matrix during the boom years of Indian startups from 2012-21.

Bessemer missed out on some big wins, but it’s also been able to return proportionally more cash to LPs than its bigger competitors. (The selected slides have been altered to remove identifying information.)

Subscribe now

Among the highlights:

• From its global funds, Bessemer has committed $774 million to India since 2006 (not including the $570 million its India-only funds have raised) and returned $683 million to limited partners as of March 31, 2024. The total value of the holding, payouts included, was $1.6 billion.

• Compared with Sequoia (now Peak XV) and Accel, Bessemer has returned cash at a faster rate. Peak has raised $10.5 billion and returned $7 billion while Accel has raised $3 billion and returned $1 billion. Bessemer’s slow-burn strategy has borne fruit in a market where investors routinely complain of slow cash returns, though its IRR lags.

• As of mid-2024, Bessemer’s best India investments — many in enterprise software or fintech — were worth 8-10x the amount invested, on paper.

• From its first India fund, raised in 2021, no company had broken out yet.

Read more

The legend of Satya Nadella is of an executive who took over Microsoft a dozen years ago after the troubled reign of Steve Ballmer, and saved the company.

He made Azure into a force and Office into a cash machine. In the AI era, Nadella moved faster than rivals to navigate another platform shift. He bet big on OpenAI, giving Microsoft early access to breakthrough models, and recruited a DeepMind co-founder to lead its internal AI effort when the OpenAI partnership was looking shaky.

It’s a legend I know well because I helped create it.

At the Wall Street Journal, I covered Microsoft during its AI renaissance, talking to Nadella and his deputies frequently, while chronicling a bet-the-company strategy that seemed to radically reshape its culture and products. Of all tech giants, how did stodgy Microsoft emerge as the winner in AI while its rival Google was in chaos? The answer was Nadella.

But the problem with legends is they’re unmoored from time and can outlive the truth. And time has not been kind to this one.

Subscribe now

AI Implosion

Microsoft under Ballmer famously missed out on mobile. Nadella, to his credit, was early to generative AI. But age-old company problems of weak product design, questionable leadership selection, and organizational sprawl have Nadella looking like he’ll be run over by it all the same.

Since the beginning of the year Microsoft’s stock is off 19%, the worst performance among the Mag 7. It’s been walloped by a confluence of factors — among them the investor selloff of SaaS stocks and rising competition from other AI players — but the biggest problems are of its own making.

I’d been hearing a drumbeat of critiques from high level Microsoft sources for months, but that grew into a roar last week after Nadella announced he was re-doing its AI team. The company’s face of AI, Mustafa Suleyman was moving out of product oversight and into a new role building large language models. Jacob Andreou, a newish recruit from Greylock Ventures and previously a star product leader at Snap, is now overseeing Copilot products across the company. A collection of executives including Ryan Roslansky, who’s also the CEO of LinkedIn, will be overseeing Microsoft 365 apps.

In a note to the company, Suleyman said he was now able to “focus all my energy on our Superintelligence efforts.” It’s a type of face-saving PR spin that would be amusing for its cynicism if the whole situation wasn’t entirely predictable. Nearly every current and former Microsoft executive I talked to after Nadella brought Suleyman on anticipated the move would be a bust.

Last year I chronicled the issues Suleyman experienced since he joined Microsoft, including zero growth in consumer Copilot and a headstrong style that alienated its key partner, OpenAI. In an incident Microsoft employees still shake their heads about, he yelled at OpenAI’s then-CTO Mira Murati in a meeting, demanding access to its code (to be clear Microsoft has the rights to it, but the topic was sensitive).

To be fair, Suleyman is charismatic and articulates an appealing vision for AI more cleanly than Sam Altman or Dario Amodei. Far from being banished, his new role of overseeing model development is an important one as the company tries to catch up and lessen its dependence on outside models. Last week Microsoft released an image model underneath its MAI large model family that it said made MAI “a top three text-to-image labs in the world.” (It debuted at number three on the Arena leaderboards, which isn’t nothing, but not exactly a Gemini or GPT killer).

Still, his product legacy isn’t pretty. The consumer Copilot app he was tasked with reviving had 6 million daily active users in February, according to CNBC citing Sensor Tower. ChatGPT had 440 million that same month; Gemini had 82 million (so much for making Google “dance”). Even Claude, which is far from a major player on the consumer side, had surpassed it at 9 million by March.

The businesses Suleyman didn’t oversee fared no better. Adoption of Microsoft 365 Copilot — an enterprise product long run by a different team, confusingly — has been a rolling disaster, barely mustering 15 million paying seats on a global commercial customer base of 450 million. GitHub Copilot, the earliest and once-widely hailed AI coding tool, has lost the narrative first to Cursor and then to agentic tools like Claude Code to a point that it’s barely in the conversation.

Microsoft in a statement pushed back on the notion that their AI products are a dud with customers: “The idea that Copilot isn’t growing just doesn’t match the facts. We’re seeing more and more large-scale deployments and increasing daily usage,” a spokesman emailed me. He compared the 15 million Copilot enterprise users to disclosures from Google, OpenAI and Anthropic, all of which had a fraction the number of enterprise seats.

It’s hard to know how to make fair comparisons among these figures, but given the momentum all those companies and their models have, I’d love to check in on those numbers in a year.

Talent Troubles

The number of recent leadership changes at Microsoft, a company of lifers, has been stark. The head of the Azure AI platform, Eric Boyd, is leaving for a new role at Anthropic, where he’ll oversee its infrastructure team, I’ve been told by multiple sources. Also gone is the CEO of GitHub, Thomas Dohmke, who left last year to start his own company. Phil Spencer, the respected head of Xbox, announced his retirement last month, and his deputy Sarah Bond followed him out the door. Rajesh Jha, the longtime EVP who oversaw Microsoft 365 and Windows, is retiring as well.

Read more

Jeff Bezos is the latest to catch AI-buyout fever as OpenAI and Anthropic flirt with private equity. The latest data dump from Carta shows large, recent venture vintages outperforming everything else. Cybersecurity startups Cloaked and Xbow raise shiny 9-figure funding rounds. OpenAI’s plans for X-rated chats face internal objections while xAI is hit with an “undressing” lawsuit. President Trump’s cash grab on the TikTok deal is getting a little scrutiny. A profile of Dario Amodei and the San Francisco AI scene misses the mark. Carl Eschenbach returns to Sequoia Capital as a full-time VC. Defense tech startups aren’t yet playing a big role in the Iran war, with the exception of drone-makers like SpektreWorks and Eric Schmidt’s Merops.

The Main Item

Bezos Eyes $100 Billion Buyout Fund for AI-Enhanced Manufacturing

Venture investors that have bet heavily on the idea of acquiring established companies and injecting them with AI now have some formidable imitators: Jeff Bezos, private equity giants including Blackstone and TPG, and the foundation model leaders themselves.

Bezos is looking to raise $100 billion for what he’s been pitching to global investors as a “manufacturing transformation vehicle,” reported The Wall Street Journal. It will acquire companies in sectors including chip-making and defense and upgrade their operations with AI. (Some of that AI, he hopes, will come from his own Prometheus Technologies, which The New York Times said last November had raised $6.2 billion, and which the Journal says is now raising another $6 billion.)

The big swing from Bezos comes as OpenAI and Anthropic pursue deals with private equity firms to sell AI to their portfolio companies. OpenAI is in talks with firms including TPG and Bain Capital for a $10 billion joint venture that would bring its enterprise products to PE firms’ companies, Reuters reported. Anthropic is in discussions with Blackstone and Hellman & Friedman on a similar type of deal, per The Information.

It all has echoes of initiatives like Thrive Holdings, which is funding startups that acquire SMB service businesses in fields like accounting with the goal of building an AI-driven powerhouse. General Catalyst, Lightspeed, and other top-tier VC firms have also backed roll-up plays, and we wrote last June that they were going mainstream for venture investors.

Now the same idea may be going mainstream for the sovereign wealth funds of Asia and the Middle East, who Bezos is said to be pitching. If you believe in the transformational power of AI, the logic is obvious enough. For big global investors, it’s also a nice hedge against their exposure to SaaS stocks and other expected victims of AI’s rise.

That said, the idea of bringing a top-down AI transformation to a company — let alone entire industries — and harvesting the extraordinary productivity gains that result remains a theory rather than a proven opportunity. The challenge for Bezos and everyone else will be to show that AI can make this kind of investing strategy something more inspirational than the typical private equity exercise in cost-cutting.

Newcomer Podcast

Shardul Shah on the $32B Wiz Sale, Founder Chemistry & VC Conviction

Index Ventures partner Shardul Shah was one of the first checks into Wiz. This month, Alphabet finalized its $32 billion acquisition of the cyber security startup, generating billions in returns for Index.

In this conversation, Eric sits down with Shah to unpack how the Wiz deal came together, what Google really bought for $32 billion, and why mid-sized acquisitions almost always fail.

They get into how Index thinks about doubling down across funds, why Shah refuses to invest in a founder he’s only met over Zoom, and what he saw in the Wiz founders.

Listen To The Podcast

Data Deep Dive

Top Venture Funds See Quick Gains on Recent Funds While Others Lag Far Behind

The gap between the top performing venture funds and everyone else is widening.

We’ve written about this trend extensively when it comes to the ability to raise mountains of capital. Now we’re seeing some of the recent larger top-tier funds enjoy some unusually quick markups.

That’s the major conclusion from Carta’s Q4 2025 VC Fund Performance report, based on reported data from nearly 3000 venture funds that use its fund management software.

Carta reports that funds in the 90th percentile of performance from 2017 and 2018 are showing TVPI that’s nearly double those in the 75th percentile. The best players are running away with it even compared to the good ones, with the 50th percentile funds left in the dust.

Carta’s IRR data broken down by size and year aligns with what we saw in our recent post on UTIMCO fund data: the top percentile of funds with $100 million or more have seen IRR shoot up in recent years. Funds of this size from 2018 are seeing IRR hit 54.7%, while funds from just last year have reached 77% IRR.

Read more

The Power Law is fully on display in the latest venture return numbers from UTIMCO, the money manager for the Texas state university system.1

Some of Thrive Capital’s 2022 and 2024 funds are already showing IRR of 50% or more — extraordinary paper gains for such young funds, built largely on a few concentrated bets, including OpenAI.

Notable Capital (formerly GGV Capital) also saw remarkable paper gains of nearly 100% on two 2023 funds — likely driven mostly by Anthropic.

Altimeter managing director Meghan Reynolds wrote on X Saturday that based on her internal math, VC investors’ gross profit on 3 LLM companies currently equates to ~70% of all VC profits from the previous decade. We certainly see this reflected in the top performing funds of the last four years in UTIMCO’s portfolio.

One caveat to this data: this disclosure is through November 2025. Some of the companies driving these funds’ returns have been marked up even more since this data was compiled.

While venture capital has traditionally been the get rich slow business, quick up rounds of high-flying AI startups have some recent vintages seeing remarkable markups early in the life of their funds.

The IRRs of older fund vintages are more likely to be reflective of how those funds ultimately shake out for LPs. But something amazing is afoot for more recent funds that have investments into Anthropic and OpenAI at the right time.

Our key takeaways from UTIMCO’s return data:

• Thrive Capital’s Fund VIII — its 2022 venture vintage — had an IRR of just over 126%, according to UTIMCO’s disclosures through the end of November 2025. It’s the largest single-year IRR from a venture fund we’ve seen since we started covering these disclosures, driven by early investments in OpenAI, Cursor, and Base Power. Thrive’s growth fund from 2022 also contains investments in OpenAI and Ramp. Thrive’s 2024 venture fund is still showing slightly negative IRR. We recently reported on Thrive’s internal calculations of TVPI and DPI in its pitch deck to investors.

• Notable Capital, which is still recorded internally at UTIMCO as GGV Capital (its former title before spinning out its China funds), boasted the largest IRR swing of any UTIMCO funds we’ve seen: its core 2023 fund IRR rose from -48% to 96% in just one year of reporting. This fund includes stakes in Anthropic and Fal. However, Notable’s older 2016 and 2018 funds lag behind some UTIMCO vintages we’ve pulled.

• Sequoia Capital’s 2020 and 2021 funds have seen solid IRR growth in the last year. Its evergreen Sequoia Capital fund, which launched in 2021, has a positive IRR of 14.78%. Our past coverage of UTIMCO’s funds had many observers chattering about Sequoia’s performance, but this fresh data shows that five to six years in, these funds could be turning the corner. Sequoia’s 2021 Seed fund, for example, saw its IRR rise from just over 6% in last year’s report to 11.3% this year.

• HongShan and Peak XV, on the other hand, have struggled after splitting out from Sequoia. At HongShan, 2020, 2021, and 2022 IRRs are all negative or barely positive by 2025, reflecting the troubles of the Chinese startup sector. Peak XV, whose internal dramas we revealed here, has seen its 2020 funds reach positive territory, but subsequent vintages remain underwater.

We’ve broken down vintages for nearly a decade of data from 2016 to 2024 and split them out by year to see how UTIMCO is marking up (or down) some of its venture capital investments as of November 30, 2025.2

One important thing to keep in mind: IRR figures for recent funds mostly reflect valuation mark-ups on private companies, rather than realized gains, so any broad market downturn or bubble-bursting would bring those numbers down substantially.

Thrive Capital, Notable Capital, and Sequoia Capital declined to comment for this piece.

Subscribe to see charts that capture the net IRR of select UTIMCO venture fund investments in 2025, divided out by vintage year.

Read more

The technology is finally here. Humans can talk to machines without the halting speech and constant mix-ups that defined past approaches.

That capability leap has spawned new voice infrastructure providers and all sorts of compelling voice applications.

The rise of voice as a natural interface with machines feels like a true paradigm shift nested in the broader AI awakening that we’ve been tracking with our twice-annual Cerebral Valley AI Summits.

People started nudging us to think about a dedicated event focused on what’s happening in voice.

So we’re doing just that.

On Wednesday, May 6, in San Francisco we will be hosting a one-day, invite-only event dedicated to voice as a computer interface.

We’re calling it the Cerebral Valley Voice Summit.

The event will have the same high-caliber founders, investors, AI leaders, and members of the media you’ve come to expect from our summits. We’ll just be deeply focused on voice as a computing interface and an unlock for all sorts of new ideas.

Voice is a horizontal technology. It doesn’t belong to any single industry or use case. Voice is how consumers will interact with their wearable devices. It’s how patients will navigate their healthcare. It’s how gamers play alongside their friends and how they will interact with non-player characters in the future. It’s how customer service reaches thousands more callers.

Apply to Attend

We’ve already locked in a stacked speaker lineup, including:

• Bret Taylor, co-founder of Sierra and Chairman of the OpenAI board

• Jake Saper, General Partner at Emergence Capital

• Dylan Fox, founder and CEO of AssemblyAI

• Shiv Rao, founder and CEO of Abridge

• Anastasis Germanidis, founder and co-CEO of Runway

More speakers will be announced over the next few weeks.

I host the Cerebral Valley AI Summit with my good friends Max Child and James Wilsterman. Max and James are the founders of Weekend, which has been building voice-native products since before it was fashionable. They’ve hosted all of our CVAI summits with me — this one will just be especially in their wheelhouse.

We encourage you to apply to attend the summit. We’re eager to have ambitious AI startup founders, general partners at top venture capital firms, members of the media, and other AI insiders in the room. Raise your hand and let us know you want to be there.

Regular tickets are $699. Founder tickets are $99. Sales and Professional Services tickets are $999. Media passes are free. Startup employees and select other junior roles will only be offered half-day admission for $399. To learn more about pricing, click here.

Apply to Attend

Special thanks to our first sponsor, Felicis, for helping us bring together the premier leaders in AI voice.

More speaker and sponsor announcements coming soon.

• To see Newcomer’s full event calendar, click here.

• To reach out with questions or to sponsor the event, email newcomer@newcomer.co.

• To learn more about our past Cerebral Valley AI Summits, go to cerebralvalley.com. We’ll host our mid-year summit in London and then will be back in San Francisco with our anchor event in November.

Apply to Attend