Jensen Huang’s AI exuberance not enough for Wall St. Crypto losses reach $1 trillion. Index, Sequoia see potential returns shrink as IPOs give back gains. Anthropic’s promiscuous alliances show Google’s strength. AI applications give Europe hope. Jeff Bezos enters the foundation model race with Project Prometheus. Lambda & Luma AI lead big funding week. Meta antitrust win could boost M&A. AI giant Yann LeCun & VC marketing star Shernaz Daver move on. Trump targets state AI regs. Cook, Benioff celebrate MBS at the White House as Trump insults murdered journalist Jamal Khashoggi.

Even Jensen Huang & AI Can’t Juice Equities Forever. Crypto Underscores the Uncertainties Ahead.

After months of unsettling but contradictory indicators, the markets this week sent a pretty clear signal that the long AI-driven run-up is over. We’re not going to say that a bubble is bursting — regular readers know that we don’t find the B-word to be a very helpful moniker — but soaring public equity markets can no longer be counted on to drive tech valuations.

With private markets now the new public markets, that has implications for all investors.

A lot of risk gauges are flashing red.

• The week began with the S&P posting a four-day losing streak, its longest in a while, only to be rescued for a moment by Nvidia’s spectacular earnings. Yet even though the AI bellwether’s forecast was as bullish as they come, stocks on Thursday quickly resumed their slide — suggesting there is no kind of good news that will shift market sentiment anytime soon.

• Big increases in corporate VC investing, exemplified by Nvidia’s aggressive moves to finance its customers, tend to signal a market top, per an interesting piece from Reuters. That’s even aside from the question of whether it’s distorting the picture on underlying demand.

• Bitcoin’s slide is accelerating; you know there’s trouble when its defenders resort to talking about “fundamentals.” This wouldn’t have mattered much outside of crypto circles a few years ago, but it now poses problems for a range of investors. The crypto-treasury trade, which gave rise to publicly traded companies whose only business was owning crypto, is probably done for good. And the vaporization of $1 trillion in paper wealth will surely have ripple effects.

• Private-credit markets, which have been taking a lot of the lending business once controlled by big banks and have been central to the AI data center build-out, are bracing for defaults. Relatedly, debt-heavy companies like CoreWeave are seeing their stocks sink.

• The burst of IPOs earlier this year seemed to be laying the groundwork for investors to get some liquidity from the over-valued unicorns minted in 2021 and 2022. That is now in doubt, especially in light of the correction in the big IPO run-ups of the spring and summer. (More on that below.)

There are a few caveats here. Some of the sell-off is surely due to signals that the Fed will not in fact deliver another interest-rate cut this year. Consumer spending and the job market appear to remain resilient. The major AI companies show few signs of pulling back. Big Tech earnings continue to be extraordinary by any historical measure. The big promise of AI, measured by consumer uptake, revenue growth, and enterprise enthusiasm, remains intact.

But this moment is a reminder of a lesson learned many times before: new technologies that fulfill their promise of changing the world usually create many epic fortunes — and sometimes epic losses too.

This post is sponsored by MongoDB

AI agents are becoming essential across every industry. The next challenge? Getting them to work together.

Join MongoDB and top AI builders to create tools and systems that enable true agentic collaboration. With $25,000+ in prizes, teams will build on January 10th at the Agentic Orchestration and Collaboration Hackathon.

The top projects will advance to MongoDB.local San Francisco on January 15th for a live showcase in front of industry leaders, top-tier investors, and crucial technical decision makers. Winners will be announced on stage during the event.

Note: Hackathon is fully in-person in San Francisco. Space is limited. Max team size: 4. Application required.

Apply now →

VC IPO Gains Lose Some Luster as Shares Fall

The market downturn has erased billions in value for some of Silicon Valley’s most influential investors as recent IPOs give back gains.

Index Ventures, the largest shareholder in Figma, has seen the worth of its 62.7 million shares in the design-tool company fall from $7.2 billion the day after the late July IPO to about $2.1 billion as of the Thursday close.

DST Global, the biggest outside backer of Chime with 52.3 million shares ahead of the IPO, has seen its gains fall by about half as shares dropped to $17.95 Thursday from $37.11 on opening day back in June (it’s unclear how much DST Global sold, if any shares, during the IPO).

Sequoia Capital’s 76.7 million shares in Klarna have likewise dropped in value from $3.5 billion to about $2.15 billion since the IPO two months ago.

To be sure, the investments remain big wins for the firms, even if they didn’t win even bigger from dramatic post-IPO run-ups that few anticipated. At CoreWeave, which barely got its IPO out the door, Magnetar Capital’s 107.96 million shares have climbed from $4.3 billion at the first-day close to $7.47 billion Thursday — well below the post-IPO peak, but still far above the debut price.

Partnership Politics

As AI Alliances Shift, Google Seizes the Pole Position

When OpenAI announced in early November that it had signed a major cloud-computing deal with Amazon, the significance went well beyond the dollar figure. Sure, $38 billion was impressive, but the real message was that the industry’s AI alliances were in flux.

It’s evident we’ve entered a new phase of the foundational model wars, where no frontier lab is beholden to any single cloud giant. Indeed, the labs themselves are now competing with them by building their own data-center capacity.

Nowhere was the shift more apparent than at Amazon. The company has invested $8 billion in Anthropic and Jeff Bezos himself internally discussed it as Amazon’s big AI bet, according to a person familiar with the matter. It wasn’t their only one — Amazon has invested in its own frontier model, called Nova — but bringing OpenAI onto AWS has made the Anthropic relationship look like one of many, rather than a special partnership.

Some Anthropic executives on the infrastructure side were concerned that the Amazon-OpenAI deal could deprive them of some AWS resources, sources told us. A couple weeks later Anthropic announced a deal with Microsoft — once OpenAI’s exclusive provider — wherein the tech giant would invest $5 billion into Anthropic, and Anthropic would in turn purchase $30 billion of Azure compute.

Microsoft is now empowering the ChatGPT maker’s largest rival. And Anthropic, which had long worked with multiple cloud providers, is now pushing even farther in its quest for more capacity.

Anthropic also deepened its partnership with Google, securing “tens of billions of dollars” in cloud computing. Though that deal, which will provide Anthropic with one million of Google’s tensor processing units (TPUs), was struck over the summer, Anthropic only announced it in late October — a week before the AWS-OpenAI tie-up became public.

Beyond expanding its computing capacity, Anthropic intended the stepped-up Google partnership as a recruiting signal, another source said. In the war for AI talent, declaring that you’re about to scale up your compute budget is catnip to researchers, whose loyalties often follow the biggest chip clusters. With each OpenAI infrastructure announcement grabbing headlines, Anthropic wanted to signal that it’s keeping up

There’s delicacy to all this; in the press releases for the Google and Microsoft deals, Anthropic took care to note that Amazon remained its “primary training partner and cloud provider,” and that the two companies continued to work together on big AI-chip build-outs.

In this landscape, the company that looks the most formidable is Google. A few years ago, the media loved to slam it as a company that was drowning, dragged down by a wayward corporate culture and poor leadership. Sundar Pichai seemed to have whiffed on large language models (a technology Google had partly invented!) and looked like he might be on the ropes.

Now Google is storming back. Its stock is up 75% in the past six months, trouncing every major tech company — that’s more than double Nvidia’s gains — and its market cap is neck and neck with Microsoft’s.

Google is the only fully verticalized tech giant, with each part of its stack at or near the top of the class. Its TPUs are the only strong rival to Nvidia’s GPUs for training and inference (Amazon’s in-house chips don’t have a great reputation), its data center footprint is large and expanding, and its Gemini 3 model is currently state of the art, according to the AI benchmarks (Sam Altman even tipped his hat to Gemini and in private has reportedly told colleagues Google’s rise will cause OpenAI turbulence).

There’s a precedent for this kind of upstart-versus-incumbent dynamic. At the dawn of the streaming wars, Netflix’s then content chief Ted Sarandos used to say that his goal was to “become HBO faster than HBO can become us.” In AI, the dynamic runs in the opposite direction. OpenAI has been trying to become Google faster than Google can become OpenAI. The problem for OpenAI is that Google is managing to become both.

Two Big Charts

American Companies Dominate Foundation Model Funding, but European Application Startups Are Closing the Gap

Read more

Major New Entrant Heats Up AI Race Even As Trump & Musk Shadow Everything

Like all Americans who pay attention to public affairs, we struggled this week with a split-screen of news, and it was particularly disorienting since we pay close attention to the tech industry and Elon Musk for a living.

On one side of the screen, we had our regular world of business and venture capital news, topped as it has so often been recently by extraordinarily ambitious AI plays.

Former OpenAI CTO Mira Murati, who replaced Sam Altman as CEO during his brief ouster and later quit herself, on Tuesday said that she’d recruited 30 engineers — 20 from her former employer — for her new startup, Thinking Machines Lab. The announcement was light on details, but Murati’s reputation has gone far in vaulting the start-up to the front of the AI race, at least when it comes to expectations.

Meanwhile Murati’s former OpenAI colleague Ilya Sutskever is also playing the great-expectations game, reportedly valuing his new startup, Safe Superintellience, at $30 billion as he seeks to raise another $1 billion in capital. Another competitor, Musk’s xAI, was reported to be courting investors for a cool $10 billion. X itself is in the funding market too.

On the other side of the screen, Musk was speaking incoherently, marching around with a chainsaw, blitzing his platform with disinformation, interfering in foreign elections, and seemingly moving X to a business model based on extortion. President Donald Trump was delivering on his promise to turn the Justice Department into his personal tort litigation firm, bent on extracting financial and political settlements from criminal and civil defendants, and weaponizing the DOJ, FBI, FCC, and FTC against his critics.

On one side of the screen, there’s routine news such as Hewlett-Packard buying a one-time darling for parts — in this case Humane, maker of a not-great AI pin. (More on that below.)

On the other side of the screen, there’s Trump denouncing Ukraine’s President Volodymyr Zelenskyy as a dictator, blaming him for starting the war with Russia, ramping up his attacks on NATO and other allies and continuing to tout territorial expansion, all while praising Vladimir Putin and ordering sweeping cuts at the Pentagon.

On one side, we watch the spread of bird flu. On the other, we see the nation’s once world-leading public health system and biomedical research institutions being ravaged.

Here in San Francisco, on one side of the screen we have our new Mayor, Daniel Lurie, grappling with a huge budget deficit and a fresh uptick in overdose deaths. On the other side, we have Trump ordering the dissolution of the Presidio Trust, with the apparent intention of selling the one-time Army base, now a magnificent park on an unmatched spit of coastline, to real estate developers.

On one side, life goes on. On the other side, Trump muses about an unconstitutional third term, or maybe king for life.

We don’t think this is going to end well. We’re not sure what our Trump-supporting friends in the VC community are thinking.

Brought to you by Vanta

Compliance for Startups: Download the Checklist

As a startup founder, finding product-market fit is your top priority.

But landing bigger customers requires SOC 2 or ISO 27001 compliance — a time-consuming process that pulls you away from building and shipping. That’s where Vanta comes in.

Join over 9,000 companies, including hundreds of Y Combinator-backed startups like Supabase, Newfront, and Fern who streamline compliance with Vanta’s automation and trusted network of security experts. Start with our SOC 2 compliance checklist, which breaks down the process into clear steps — so you can spend less time on compliance and more time growing your business.

Download the SOC 2 checklist

One Big Chart

Pre-Seed Funding Fell at the End of 2024

Venture funding at the earliest stages of a startup’s lifespan had proven relatively resilient throughout the most recent downturn, but that trend is no longer holding, according to fresh data from Carta.

The boom times saw multistage funds increasingly dipping their toes into seed and even pre-seed investing, but this trend has reversed. Investors are clamoring to get into just a few large, growth-stage deals that they see as clear winners.

Carta’s latest report on the State of Pre-Seed investing in 2024 shows that pre-seed funding dropped significantly in Q4 of last year, down to $716 million across 5,382 deals. At the same time, growth stage investing picked up significantly — the value of Series D deals increased by 78.8%, per Carta.

Six Notable Deals

Saronic, Lambda, Together AI, Abridge, Hightouch, Plus Cherryrock Capital Raises its First Fund

Read more

DeepSeek Spurs Lively Debate About AI Innovation, ‘Distillation,’ VC Valuations & China

This week marked a coming-of-age moment for the AI industry: its very own market panic.

Nvidia on Monday saw its value drop by nearly $600 billion on the news that China’s DeepSeek had apparently built a competitive LLM without giant data centers full of cutting-edge GPUs.

That’s more than what the biggest of big-tech companies were worth less than a decade ago. Broadcom and AMD plummeted too, along with energy and infrastructure companies.

Suddenly, it seemed that VCs who’d pumped billions into Open AI and the other foundation model companies — we’re looking at you, Thrive, Sequoia, and a16z — might have made an historically bad bet.

The panic eased as the week went on, and then Thursday came a Wall Street Journal report that OpenAI was raising new funds at almost double its current valuation. The news was perfectly timed to blunt the doomsday predictions about the big American LLM companies. Whether the deal closes at that price will be a litmus test of its own.

Meanwhile, though, Microsoft and Meta, in reporting earnings this week, showed no hints of a pullback in AI spending. By midday Friday, Nvidia and the Nasdaq had regained all their losses and then some.

There’s still a lot of disagreement as to whether DeepSeek, allegedly trained for a fraction of what U.S. companies are spending, is a game-changer or more of an incremental advance, though rivals acknowledged that the achievement is real.

DeepSeek shows that generative AI still has a lot of juice, despite the chatter about “scaling walls” and worries about spending and margins. Models are still improving. A bounty of agents and applications are in the pipeline.

Benchmark’s Eric Vishria called foundation models “the fastest depreciating asset in history,” and DeepSeek does put a finger on the question of how defensible a better proprietary LLM is on its own.

“That market is already fiercely competitive and slowing down or missing a step already means they are in trouble,” Madrona’s Jon Turow told us over text. Even if LLMs don’t end up being profit machines, though, there are still massive businesses to be built.

The “Jevons Paradox,” where the declining cost of a commodity creates a huge surge of new demand, is having a moment in the sun.

The DeepSeek surprise is also a reminder that China is a serious global competitor. Yet its tech community — as opposed to the CCP — remains a friend of Silicon Valley in many ways.

We liked this Bill Gurley line about the DeepSeek team: “Why are they ‘enemies?’ No one that works at DeepSeek is an enemy of mine. The fact that they love open-source makes me think I would gel with them quite well.”

Others were more hostile to the Chinese startup’s triumph. VC David Sacks in his new capacity as President Trump’s AI czar said “they cheated.”

Sacks and others speculate that DeepSeek may have used a process known as distillation, or learning from a different LLM’s answers, to develop its tech so quickly. Joshua Kushner, a major OpenAI shareholder, also accused DeepSeek of training off of leading US models.

There’s also a lot of skepticism of the claim that the model was trained for just $6 million on a small cluster of Nvidia’s less-advanced chips. Scale AI’s Alexandr Wang and others say DeepSeek in fact had access to a cluster of 50,000 Nvidia H100s, raising the issue of whether and how DeepSeek had dodged export controls that bar Chinese companies from acquiring the most advanced chips.

The analysts at SemiAnalysis estimate that DeepSeek’s full hardware spend is much higher than $500 million, and that its hedge fund owner High-Flyer started building out its chip cluster in 2021, well before any export restrictions were in place.

Bloomberg reported Thursday that White House officials are probing whether the company might have obtained Nvidia chips through middlemen in Singapore.

A second category of reactions are from those who say it’s all in a day’s work for the tech industry. Even President Trump himself suggested the advance should be a good motivator for American companies to stay ahead. Marc Andreessen called it a “new Sputnik moment” for U.S. researchers, a panic that can also spark innovation.

Anthropic’s Dario Amodei offered a detailed argument on why DeepSeek, though important, wasn’t quite what it seemed. It’s unsurprising that he’d downplay the threat, but his logic is worth a look. “DeepSeek-V3 is not a unique breakthrough or something that fundamentally changes the economics of LLMs; it’s an expected point on an ongoing cost reduction curve,” Amodei wrote.

DeepSeek underscores that technical leadership in LLMs is only one piece of the puzzle when it comes to value-creation in AI.

OpenAI has what most of its rivals — and upstarts like DeepSeek — do not: lots and lots of customers. ChatGPT hit 300 million weekly users in December, Sam Altman said onstage at the DealBook summit. Anthropic is nowhere near that number, but it’s been popular with business customers, and expects its “Computer Use” agent to provide another boost.

That said, DeepSeek is currently number one in the Apple app store.

Cheaper-than-expected LLMs could shift some of the overall competitive dynamics. Basis Set’s Lan Xuezhao said startups building in the inference layer will be the quickest winners. Industries with specialized data needs — think healthtech and biotech startups using AI for drug discovery, or developers of foundation models for robotics, for example — will have a big appetite for AI products, he added.

“All SaaS apps based on LLMs will benefit,” Gradient Ventures’ Darian Shirazi texted me. He argued that the companies with the most to gain are those that help model producers find compute efficiency through profiling and optimization, like CentML or Unsloth.

“People make this too much of a big deal on US and China,” said Xuezhao. “They told you the recipe, just take it and make your own dish.”

Podcast Highlight

DeepSeek Proves the Power of Open Source

In our latest episode, we make the case that DeepSeek is a net positive for the AI community at home — precisely because of its open source nature.

We’re cosigning Bill Gurley’s take that Silicon Valley and China’s tech communities are actually pretty in-sync here.

Listen To The Podcast

One Big Chart

Seed Startups Aren’t Making the Leap to Series A

Read more

Robotaxis Are Taking Over SF. Scaling Up Will Be a Different Challenge

The self-driving future is more or less here, at least in San Francisco.

A16z’s Alex Immerman posted data on X late last month showing that Waymo’s self-driving robotaxi service in the city is now carrying just as many passengers as Lyft, and is rapidly gaining on Uber.

Lyft CEO David Risher disputed the data, saying Lyft was growing in San Francisco and that it still held 30.6% of the market as of this past November. It’s also worth noting that Waymo only operates in the city proper. Still, it appears that Waymo has crossed a threshold in its flagship city, becoming a routine choice for many of those who might have called Lyft or Uber.

Waymo’s technical and operational achievement is a monumental one, hard-earned over fifteen years.

Running big robotaxi fleets, though, will offer challenges of its own. Uber is jockeying to manage the fleets in some cities while Waymo is seeing if they can do it on their own in others. Profit margins, or the potential lack thereof, remain a daunting issue too.

Now that it’s enjoyed a smooth growth spurt, Waymo has to demonstrate that it can keep its vehicles clean, for one. Since Waymo opened rides up to the general public in mid-2024, complaints about dirty cars have been popping up regularly.

“No matter what, someone has to take care of the cars,” said Josh Mohrer, a former Uber executive who managed the company’s first push into New York. Mohrer told us he’s skeptical of Waymo’s ability to scale, in no small part because of the messy-car issue. In contrast to a house-sharing platform like Airbnb, the turnover between Waymo users is minutes, not days, which doesn’t leave much room for cleaning, he noted.

One solution could be increased monitoring, Menlo Ventures’ Deedy Das speculated, incorporating smart cameras that could detect all kinds of anti-social behavior during a ride. AI computer vision technology could be particularly useful for flagging riders who behave poorly, though it might take away some of the charm.

There are other aspects of fleet logistics too. Cars need to be inspected, insured, and maintained. If they have issues, they’ll need to be decommissioned for repairs.

Building routes and networks requires a lot of relationships with local regulators, and moving fast and deploying self-driving in new cities carries high risks if something goes wrong. Bad accidents were enough to take both Uber’s self-driving effort and Cruise off the road, helping give Waymo its big lead.

Fleet logistics and local knowledge would explain why Waymo is tapping Uber, with its existing network of fleet managers, to be its operating partner in Austin and Atlanta. This kind of outsourcing could point the way to a model where Waymo is a technology provider to various robotaxi services run by others.

The business model question continues to loom large. Configuring a Waymo car is very expensive: its fifth-generation Jaguar taxis cost around $100,000 when accounting for all of the robotics and LIDAR involved. Its newer 6th generation version is reportedly much cheaper, though it relies on Hyundai and a Chinese electric car partner that could be vulnerable to President-elect Trump’s promised tariffs.

Waymo is competing against ride-hailing companies that have no capital costs for vehicles at all, since the cars are provided by the drivers. Ride share apps like Uber take 20% to 40% per ride on average; Waymo of course takes 100% of the fare, but it’s still not clear if the cost of the car and the technology, amortized over time, is lower than what would otherwise be paid to a driver. Uber itself has only recently seen profits.

With all these challenges, it’s no surprise that competitors are still entering various parts of the market despite Waymo’s formidable technical lead. Nvidia, for one, showed off its brand new GB10 chip at CES this week, designed to power self-driving trucks, and announced new partnerships with the autonomous trucking company Aurora, as well as Uber.

Alphabet said last year that it expected to invest an additional $5 billion in Waymo. The company in October closed a $5.6 billion funding round, which included existing outside investors such as Andreessen Horowitz and Silver Lake, at what Bloomberg reported was a $45 billion valuation.

It will need that war chest, and maybe more, for the next phase of our self-driving future.

Listen to the latest episode of the Newcomer podcast for more on why we think Waymo is still in a good position to expand outside of San Francisco, despite some challenges:

Listen

One Big Chart

AI Funding Explodes in Q4, Even as Volume Recedes

Read more

Andreessen Horowitz Raises Billions

For Andreessen Horowitz, it’s time to build, across every hot tech sector.  

The team announced $7.2 billion in new capital raised on Tuesday. With Andreessen’s new fund structure, there are key sectional leaders underneath Marc Andreessen and Ben Horowitz—and some other once-rising stars have headed for the exits. 

The capital is divided between multiple funds this time, each helmed by a general partner:

• $3.75 billion for growth investing

• $1.25 billion for infrastructure (which includes many of the firm’s AI investments)

• $1 billion for “Apps” — made up of enterprise, consumer tech, and fintech

• $600 million for games

• $600 million for American Dynamism 

Commitments to certain segments are more modest than they were previously. A16z’s last growth fund was capped at $5 billion, so this is definitely a step down from that. The firm also chose not to break out a separate seed fund this time around. 

Overall the new funds closely map what’s “in” and what’s “out” in Silicon Valley, with big commitments to defense tech and AI. Since its founding in 2009, a16z has positioned itself as “a different kind of VC,” and now with its massive size and extensive set of services for founders, it may well be that. But its investment theses for these new funds fit right in line with the hype categories of our current moment, and with the sheer number of investments, a16z looks in some ways like an index fund for hype-y startups.

That might be all but inevitable as the firm scales up, but it certainly faces a long-term challenge in maintaining the fat returns expected from a top-tier VC off an ever-growing pile of capital.

This fundraise feels large given the funding drought, but is in line with some of the firm’s mega-fund peers. Lightspeed Venture Partners closed over $7 billion in fresh funds back in the summer of 2022, although most of its fundraise had happened before the downturn fully hit.

As for what’s “in,” it’s no surprise that AI is a key category within the new funds: general partner Martin Casado leads the infrastructure/AI segment, and now his team has a lot more capital to deploy. The firm made a big bet on Mistral in December, leading its $415 million round, and scooped up OpenAI secondary shares in a sale early last year. 

Defense tech and gaming are big priorities too. David Ulevitch leads the American Dynamism practice, which hosted a DC summit earlier this year. 

Andrew Chen, in turn, is in charge of games. Chen switched his focus from consumer to go all in on games in 2022. 

On the growth side, David George, who oversaw the firm’s investment in Flexport, is leading the charge. Scott Kupor, the firm’s first employee, manages investor relations and growth initiatives on the internal operations side.  

The “Apps” team, helmed by Alex Rampell, has lost some GPs in the past few months—Connie Chan, who had been leading the push to get into more China deals, left in late January. Some of Chan’s bets, such as the live shopping startup Whatnot, have since fizzled.  

Kristina Shen, who racked up several early-stage enterprise deals for the firm such as her recent investment in YC alum Tennr, announced two weeks ago that she was leaving to raise her own fund.

Enterprise SaaS and fintech, also part of the apps group, are in a slow patch, with fintech recording its lowest funding totals in 6 years. SaaS investing jumped at the end of 2023 but is still down from its peak.

Notably absent from the new fund announcement are Bio + Healthcare and Crypto, though they are presumably still flush with cash after big raises a few years ago. A16z was a major crypto booster and Coinbase remains one of its biggest wins ever, but they also suffered some losses. A prized crypto investment — Uniswap — is facing potential trouble from the SEC.

As a16z continues to scale up, it may find that it has to cover more bases, as it is doing in politics: new filings showed the firm has written $250,000 checks to Democratic and Republican Super PACs this year. The bigger you are the more incentive there is to hedge your bets.

VC Directory

Andreessen Horowitz

In light of the new fundraise, we’ve decided to update our entry for Andreessen Horowitz in our VC directory. 

Here’s a breakdown of who’s who at the firm, which was founded in 2009 by Marc Andreessen and Ben Horowitz. It’s come a long way since its initial $300 million fund, now boasting dozens of investors and hundreds of employees. The firm is stage agnostic, investing from seed to growth stage. Eric wrote a detailed “unauthorized history” of the firm a couple of years back that’s worth perusing for a more in-depth look at how the firm became the behemoth it is today. 

Outside of his dealmaking prowess, Andreessen is known for his public prognosticating on Twitter and occasional essays, from his famed “software is eating the world” to a more recent declaration that it was “time to build.”   

Big thanks to our intern Janya Sundar for helping compile this entry. 

Assets Under Management

Read more

The public should not want, nor should the OpenAI board give Altman unbridled power to run OpenAI as he pleases. Altman has a history of fractious corporate breakups. People need to reflect on Altman’s time at Y Combinator; the Anthropic co-founders’ decision to spin off from OpenAI; and his breakup with Elon Musk.

These board members are not the first people to question Altman’s integrity. They’ve just done so in public.

Many OpenAI employees, OpenAI shareholders, and Microsoft seem to want Altman back in charge. That’s putting a ton of pressure on OpenAI’s nonprofit board — which consists of OpenAI chief scientist Ilya Sutskever, Quora CEO Adam D’Angelo, technology entrepreneur Tasha McCauley, and Georgetown Center for Security and Emerging Technology’s Helen Toner — to relent.

Subscribe now

(Altman and co-founder Greg Brockman were also on the board until Friday’s coup. Altman was ousted. Brockman was removed as chairman but remained at the company until he resigned.)

Altman declined to comment for this story.

I don’t think the board should give in just to escape intense public pressure. But they do need to explain themselves better. They’ve bungled the public communication so far.

The board hasn’t articulated particularly specific grievances. And I don’t think there is any one single skeleton that the board is pinning their decision on.

In its statement, the board said it had concluded Altman, “was not consistently candid in his communications with the board.”

We shouldn’t let poor public messaging blind us from the fact that Altman has lost confidence of the board that was supposed to legitimize OpenAI’s integrity.

Once you add the possibility of existential risk from a super powerful artificial intelligence — which OpenAI board member Sutskever seems genuinely concerned about — that only amplifies the potential risk of any breakdown in trust.

My understanding is that some members of the board genuinely felt Altman was dishonest and unreliable in his communications with them, sources tell me. Some members of the board believe that they couldn’t oversee the company because they couldn’t believe what Altman was saying. And yet, the existence of a nonprofit board was a key justification for OpenAI’s supposed trustworthiness.

I don’t think any of us really knows enough right now to urge the board to make a hasty decision. I want you to consider a couple things here:

Subscribe now

• OpenAI’s board is not meant to be a normal board. This is a nonprofit board meant to keep OpenAI aligned with a mission in the public interest. It makes sense that this board would operate differently than a traditional toady board on a for-profit company. Altman has consistently pointed to its nonprofit status in promoting and defending OpenAI. He told Congress, “We are governed by a nonprofit, and our activities are driven by our mission and our charter, which commit us to working to ensure that the broad distribution of the benefits of AI and to maximizing the safety of AI systems.” This nonprofit board was used to shield OpenAI’s reputation. Shouldn’t that same board be expected to wield influence when it thinks things are amiss? The whole point of having a nonprofit board is that they could act in a way that frustrates the company’s investors.

• There are three key historical case studies here: First, Dario Amodei, Jack Clark and the team at Anthropic felt troubled enough by OpenAI’s approach that they needed to spin off and create their own more safety and alignment-oriented foundation model company. What (or who) exactly got that team so worried that it needed to jump ship? Altman was certainly at the center of that decision.

• Second, as I resurfaced on Friday, Altman’s departure from Y Combinator was more contentious than publicly understood. Altman, then YC’s president, was asked to leave, a source told me. He left YC without any affiliation even though initially Altman was supposed to be chairman of YC or at least an advisor. Some similar issues were at play. At YC, Altman was distracted with OpenAI; he invested aggressively alongside the accelerator; he wanted to expand dramatically, even trying controversially to launch YC in China (a decision the subsequent YC president reversed); and YC’s brand was becoming synonymous with Altman. At OpenAI, Altman has reportedly been talking about starting a hardware company with Masayoshi Son and Jony Ive; he’s pushing for faster and faster expansion of OpenAI; and he’s certainly become the company’s almost singular figurehead. Altman seems to have an almost insatiable appetite to increase his own power and influence. Bloomberg reported that Altman was raising billions for a new chip ventures.

• Third, remember also that Elon Musk and Altman had a breakup over OpenAI.

• The whispered narrative among Altman’s skeptics always went like this: he’d led a failed startup. Paul Graham adored him and Altman was tapped as Graham’s successor. Altman leveraged the YC role to amass great personal fame and fortune. But it wasn’t clear if Altman was extremely talented or if YC was such a powerful institution that he was benefiting from its aura. Then when OpenAI became a success even Altman’s private critics had to accept that it wasn’t just the YC title that made Altman great. Altman ran one of the most important startups of the moment that released truly groundbreaking technology. But maybe Altman’s private skeptics were too quick to change their tune.

• I will say that while the public support for Altman appears all but unanimous, there are some people in private who are more mixed.

I’m sure that writing this cautionary letter will not make me popular in many corners of Silicon Valley. But I think we should just slow down and get more facts. If OpenAI leads us to artificial general intelligence or anywhere close, we will want to have taken the time to think for more than a weekend about who we want to take us there.

Altman could always decide just to start a new company, taking employees and OpenAI co-founder Greg Brockman with him. If OpenAI’s board is only left with a decimated nonprofit, then the board’s poor strategy here may end up giving Altman more unilateral authority over the future of AI — just within another entity. This could end up being a major setback for unique structures to control AI companies. I hope that’s not how this plays out.

As a piece of context on this whole situation, a number of departures from OpenAI’s board helped give the remaining board members more influence. Reid Hoffman, Shivon Zilis, and Will Hurd exited the OpenAI board this year. My sense is that the current board’s decision has been stewing for a while. The departure of those three board members gave the remaining members more voting power.

Altman is clearly a talented recruiter and relationship builder. He’s adored by many employees and investors. He’s become a great public advocate for AI and an influential figure in Washington. He’s helped catapult OpenAI to a potential $86 billion valuation while operating as part of a nonprofit with capped returns for investors.

Altman had been given a lot of power, the cloak of a nonprofit, and a glowing public profile that exceeds his more mixed private reputation.

He lost the trust of his board. We should take that seriously.

Let’s give the board and Altman’s critics some time to explain themselves and to articulate a vision for how OpenAI might move forward without putting Altman back in charge.

Here’s the announcement:

Chief technology officer Mira Murati appointed interim CEO to lead OpenAI; Sam Altman departs the company.

Search process underway to identify permanent successor.

The board of directors of OpenAI, Inc., the 501(c)(3) that acts as the overall governing body for all OpenAI activities, today announced that Sam Altman will depart as CEO and leave the board of directors. Mira Murati, the company’s chief technology officer, will serve as interim CEO, effective immediately.

A member of OpenAI’s leadership team for five years, Mira has played a critical role in OpenAI’s evolution into a global AI leader. She brings a unique skill set, understanding of the company’s values, operations, and business, and already leads the company’s research, product, and safety functions. Given her long tenure and close engagement with all aspects of the company, including her experience in AI governance and policy, the board believes she is uniquely qualified for the role and anticipates a seamless transition while it conducts a formal search for a permanent CEO.

Mr. Altman’s departure follows a deliberative review process by the board, which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities. The board no longer has confidence in his ability to continue leading OpenAI.

In a statement, the board of directors said: “OpenAI was deliberately structured to advance our mission: to ensure that artificial general intelligence benefits all humanity. The board remains fully committed to serving this mission. We are grateful for Sam’s many contributions to the founding and growth of OpenAI. At the same time, we believe new leadership is necessary as we move forward. As the leader of the company’s research, product, and safety functions, Mira is exceptionally qualified to step into the role of interim CEO. We have the utmost confidence in her ability to lead OpenAI during this transition period.”

OpenAI’s board of directors consists of OpenAI chief scientist Ilya Sutskever, independent directors Quora CEO Adam D’Angelo, technology entrepreneur Tasha McCauley, and Georgetown Center for Security and Emerging Technology’s Helen Toner.

As a part of this transition, Greg Brockman will be stepping down as chairman of the board and will remain in his role at the company, reporting to the CEO.

OpenAI was founded as a non-profit in 2015 with the core mission of ensuring that artificial general intelligence benefits all of humanity. In 2019, OpenAI restructured to ensure that the company could raise capital in pursuit of this mission, while preserving the nonprofit's mission, governance, and oversight. The majority of the board is independent, and the independent directors do not hold equity in OpenAI. While the company has experienced dramatic growth, it remains the fundamental governance responsibility of the board to advance OpenAI’s mission and preserve the principles of its Charter.

Subscribe now

I’m on a plane back to New York from the Cerebral Valley AI Summit. Laptop is perched half open on my chest, shoulders scrunched awkwardly. Maybe I need to start flying business class.

No one has said publicly yet what the full backstory is. I’ve been pinging sources frantically from the air. Please reach out if you know anything. eric@newcomer.co. I’ll give you my number and we can message on Signal.

Theories abound. Everyone wants to know how much money OpenAI has really been burning.

Brockman’s removal from the board raises lots of questions.

Besides Quora’s Adam D’Angelo (who dropped out of Cerebral Valley at the last minute), OpenAI’s board isn’t that well known. Reid Hoffman stepped off OpenAI’s board earlier this year.

Subscribe now

Wanted to highlight a couple tidbits for you while the world tries to figure out what happened.

This isn’t Altman’s first awkward departure. From a June 2021 story in Newcomer about Y Combinator:

YC has been pretty successful at keeping its drama inside the tent. That’s been to YC’s benefit for sure.

Here’s a telling case study: It’s stunning how little scrutiny Sam Altman’s various title changes at Y Combinator received in the big business publications.² This is someone who was profiled in the New Yorker and who ran the most powerful startup manufacturer in the world.

Altman went from president, to chairman, to being an advisor, to having no affiliation with Y Combinator without much detailed reporting from the press. YC successfully swept, what seems to be a real schism, under the rug.³

First YC announced that Altman would become chairman. On March 8, 2019, the firm wrote, “Sam is transitioning to Chairman of YC.”⁴

Then that same month, YC announced that Geoff Ralston would become president of YC. Altman wrote at the time, “I will, of course, remain an advisor to YC, and will be around to help Geoff as he takes on his new role.” It seems that was YC’s way of saying that Altman would not in fact be the chairman of YC, despite headlines all around the web that had run earlier that month saying that he would be.⁵

But Altman wouldn’t become a formal advisor either.

On January 27, 2020, Wikipedia user Opazazzyzen updated Altman’s Wikipedia to say that Altman is “no longer affiliated with YC.”

If you Google “Opazazzyzen,” an Instagram account for Lindsay Amos is the first result. She is the director of communications for Y Combinator. Based on my internet sleuthing, Amos seemed to have quietly announced Altman’s role change on Wikipedia. Her update didn’t include a citation (and still doesn’t).

I emailed Amos about the weird Wikipedia post. She wrote, “In May 2019, we shared that Geoff Ralston would take over as President of YC from Sam Altman. Sam noted he would be an advisor to YC and would be transitioning to CEO of OpenAI. There was every intention for Sam to become an advisor but the plans never moved forward. So, the title/role never happened. I updated the information on our end (i.e. Wikipedia, Crunchbase, etc.) to combat confusion.”

A source tells me that Altman was asked to leave YC because he was too absentee.

Altman made an appearance at our Cerebral Valley speaker dinner Wednesday night. He was well-dressed and seemed in good spirits. But we didn’t talk much. I gave him a brief download on the takeaways from some of the panels.

Notable Tweets

Moderating six panels in a day in front of a 300-plus person audience is a trip. The day is flowing through you. You’re hyper aware of the world in front of you. The outside world disappears from your mind. Time moves strangely.

I had the privilege yesterday of talking on stage at the Cerebral Valley AI Summit with Mustafa Suleyman, Vinod Khosla, Reid Hoffman, Chase Lochmiller, Ali Ghodsi, Naveen Rao, and Clara Shih. Now I’ve been able to take a few hours and reflect on the whirlwind event.

Thousands of enterprises around the world rely on Oracle Cloud Infrastructure (OCI) to power applications that drive their businesses. OCI customers include leaders across industries, such as healthcare, scientific research, financial services, telecommunications, and more.

NVIDIA DGX Cloud on OCI is an AI training-as-a-service platform for customers to train complex AI models like generative AI applications. Included with DGX Cloud, NVIDIA AI Enterprise brings the software layer of the NVIDIA AI platform to OCI.

Talk with Oracle about accelerating your GPU workloads.

Talk with an AI/ML expert

Some takeaways:

• Suleyman is one of the top AI thinkers in the world, having founded DeepMind in 2010, before the world shared his obsession with artificial intelligence. Now he’s the CEO of Inflection, an emotionally intelligent foundation model company that he co-founded with Hoffman. It’s as close to a therapist as you’re going to find in an AI chatbot.

• You probably know of Hoffman, he’s the LinkedIn co-founder, Inflection co-founder, famous investor, and major Democratic donor. He told us on stage he met with Joe Biden the day before our conference. Hoffman said he was impressed by the president’s intellectual curiosity on AI.

• Khosla was his spicy self, firing off ripostes. He won over the crowd with a call for welcoming high-skill immigrants into the United States and walked away a pretty unanimous crowd favorite.

• We started the day with Ghodsi, the charming and feisty Databricks CEO, and Rao, the founder of MosaicML, who have been Cerebral Valley stars from the beginning. Their introduction at Cerebral Valley’s first speaker dinner in March has become a thing of lore, earning a detailed portrait in Fortune and driving Business Insider to call Cerebral Valley “the AI conference that spawned a $1.3 billion deal.” Ghodsi and Rao dished on how they met and their plans in artificial intelligence. They’d just announced that customers could query their data in the cloud via plain English.

• My co-host James Wilsterman got a rare interview with Waymo co-CEO Dmitri Dolgov. James had been riding around in Waymos for the past couple weeks ahead of the interview. On stage, he pressed Dolgov about Waymo’s reaction to competitor Cruise’s pause of operations in San Francisco after an accident involving a Cruise vehicle and a pedestrian.

The first Cerebral Valley AI Summit, back in March, began with my bachelor party. I was headed to California for a weekend away with my friends, including Volley co-founders Max Child and James. Max and James had a big open office in the heart of Hayes Valley. So, they said, why not host a small event at their office to turn the trip to California into a genuine business trip? That small event quickly spiraled into an all-day elite tech conference when people like Emad Mostaque at Stability, Bucky Moore at Kleiner Perkins, and Miles Grimshaw at Benchmark signed on to participate. On March 30, we had a 200 person conference at Volley’s office. It was a blast. But we didn’t have much space for people to sneak away and do deals, so I had to spent some of my time shushing some very rich people.

This year, felt so crazy in AI that we couldn’t wait a whole year to host another AI conference. This time we rented out the SFJazz Center (no shushing necessary) and hosted more than 300 top AI startup founders, investors, and tech operators. HP and Oracle signed up to be our top sponsors.

Max interviewed Writer CEO May Habib and Snorkel AI CEO Alex Ratner on one panel. And he fanboyed with programmer god Chris Lattner on another. Lattner, a former senior Apple executive, predicted that his former employer would finally fix Siri, which has fallen behind other AI applications, next year. However, he was not so bullish about another former employer, Tesla. The once head of Tesla autopilot said he did not think Tesla would reach parity with Waymo by 2025.

James, Volley’s CTO, jammed with another former CTO, Dmitri Dolgov, the co-CEO of Waymo. He also sat down with vice president of Alexa devices at Amazon Nedim Fresko. I tapped some reporter friends — Alex Konrad, Ellen Huet, and Casey Newton to moderate the remaining panels.

Subscribe now

The conference was a true clash of ideas and ideals. Hoffman said that he was going to spend a big chunk of his time supporting Biden’s reelection, that he didn’t want to ban TikTok, and that he isn’t particularly worried about existential risks of AI. Hoffman’s Inflection co-founder, Suleyman, is more worried about the potential risks, having penned a book that warns of many otherwise smart people’s “pessimism aversion.” The Coming Wave can feel like it was written by someone who was locked away in the pandemic worried that their life’s work was going to bring about the end times. Meanwhile, Khosla wanted to ban TikTok. He also said he was frustrated about Lina Khan’s antitrust efforts at the FTC thanks to Biden. And even though Khosla doesn’t think AI is going to cause society-ending problems, he’s opposed to important research in AI happening in open-source AI projects because he doesn’t want China to be able to benefit from research in the United States. Holden Karnofsky, the director of AI strategy at Open Philanthropy and husband of Anthropic co-founder Daniela Amodei, was the strongest voice for caution about potential risk in AI. He called for strong red lines that would force companies to halt work on AI if their foundation models exceeded their abilities to contain them.

OpenAI CEO Sam Altman made a brief appearance at our speaker dinner. I overheard one startup CEO say, essentially, “I don’t want to compare him to Taylor Swift but—”

You can read on for an in-depth summary of the panels. As soon as we can we’ll begin posting videos of the talks in this newsletter. I want to thank Riley Konsella, Newcomer’s chief of staff and Volley’s Gaby Caliendo for their tireless work making both Cerebral Valleys a reality. Also, thank you to Volley’s Kelly Chronis, Christina Morris, Laura Cheney, and Riyadh Drebika.

Read more

He looks at GitHub stars, ChatGPT usage, AI application users per month, and many other metrics to track AI activity bubbling (and sometimes waning).

Coatue has shared its full 115 page report with Newcomer readers first.

Coatue Artificial Intelligence Full Report

13.9MB ∙ PDF file

Download

Download

• In the presentation, Viswanath argues that fast-moving incumbent technology companies are best positioned in the coming AI wave. But notes that in past technology waves new startups were founded just as technology emerged and far later once a new technology was well established. There isn’t a wrong time to build a disruptive technology company — if it’s a novel use of new technology. Uber emerged early in the life of the smartphone app stores but TikTok climbed to dominance many years later.

• He shows how foundation models are open and closed in a wide variety of permutations. LlaMa 1 shared its training data but didn’t offer a commercial license; whereas, LlaMa 2 didn’t share its training data but opened up a partial commercial license.

• GitHub stars on AI projects are down, indicating that the AI hype cycle may be waning. But AI commits to projects are going strong, showing that people are still hard at work coding. Meanwhile, ChatGPT usage can rise and fall when newer versions come out, drawing users back to the chatbot to try out fresh features.

I’m including some of my favorite slides from Viswanath’s presentation at the Cerebral Valley AI Summit below. Coatue is also publishing the slides on its website. Vibhor Khanna and Yijia Liang worked with Viswanath to create the presentation for Coatue.

Subscribe now

Coatue Artificial Intelligence Full Report

13.9MB ∙ PDF file

Download

Download

More to come soon on the rest of Cerebral Valley!