Menlo Partner Murphy Reviews Anthropic Investment: Three LPs Each Earned Over $1 Billion, Most Later LLM Entrants Unlikely to Fare Well
Menlo Ventures partner Matt Murphy, on The Peel podcast, reviewed how the firm made an exception to invest in 2023 when Anthropic had no product or revenue and was valued at over $4 billion; it later raised a special fund of over $500 million to lead the Series D, and the three LPs that followed each earned over $1 billion based on the latest round's valuation. He said Anthropic's operating cash flow has turned positive, at positive $3.93 billion in the first half of 2026, and that covering Google's 19-year journey from zero to $100 billion in revenue may take less than a third of the time. He believes open-weight models may account for 20% to 30%, while most of the more than 70 new labs will not have good outcomes.
Why read this
AI score 52
A Menlo partner reviews the decision process behind investing in Anthropic during its zero-revenue stage, and offers judgments on open-weight model share and the fate of later entrants.
Topics
Body · AI translation
[[**Menlo made an exception to invest in Anthropic when it had zero revenue and a valuation of over $4 billion, later raising a dedicated fund of more than $500 million; the co-investors each made over $1 billion based on the latest valuation. Murphy says most later entrants in large models will struggle to achieve good outcomes.**
**Key Points**
**1. Zero-revenue investment built trust by making an exception**
Menlo first joined the C round as a co-investor, then sought to lead the D round; when raising its first dedicated fund of over $500 million to lead the round, Anthropic would not permit disclosure of financial projections, and the 30-to-45-day fundraising at one point risked failing, but in the end the three investors each made over $1 billion.
**2. Anthropic's growth rate may far exceed Google's**
It took Google 19 years to go from zero to $100 billion in revenue, and Murphy expects Anthropic may take less than a third of that; its operating cash flow has already turned positive, at positive $3.93 billion in the first half of 2026.
**3. Open-weight models will not replace frontier models**
When developing new applications, the strongest model is used first, and costs are optimized only after scale grows; Murphy estimates open-weight models may account for 20% to 30%, and users will ultimately still pay for the best product.
**4. Most later entrants in large models will struggle to achieve good outcomes**
More than 70 new labs are raising $1 billion in Series A and $2 billion to $3 billion in Series B, but there will be spectacular acquisitions, and there will also be many big craters smashed into the ground.
**5. The current AI wave is different from the 1999 internet bubble**
This article is from the WeChat public account: AGI 2050, author: AGI 2050, original title: "In-depth Interview with an Early Anthropic Investor: Three Investors Each Made Over $1 Billion, Most of the 70-Plus Later Entrants in Large Models Will Not Have Good Endings"
This week, Anthropic officially publicly filed its IPO prospectus. But in 2023, this company still had not a single dollar of revenue, and its product had not yet been publicly released; investing in it at a valuation of over $4 billion was an extremely unusual risk in the venture capital industry at the time.
The key investment was led by Matt Murphy, a partner at Menlo Ventures, a 50-year-old veteran Silicon Valley venture capital firm. As a seasoned investor who lived through the 1999 internet bubble and, while at Kleiner Perkins, sat in on every Google board meeting from investment to IPO, Murphy not only made an exception to participate in investing in Anthropic during its zero-revenue stage, but later, in the round at a $18 billion valuation, launched a dedicated fund of more than $500 million without being able to disclose future financial projections to investors. Today, the three co-investors have each made more than $1 billion based on the latest round's valuation.
On the tech podcast "The Peel," Murphy had an in-depth retrospective with host Turner Novak. Both Murphy and the host have invested in Anthropic, and the company is currently in a pre-IPO quiet period, so he did not say much about specific financial figures, but he clearly explained how the decision was made back then, and also discussed the business rules of the large-model era and the changes taking place in venture capital.
This episode is from The Peel podcast, aired on October 26, 2026; the original video is 73 minutes long.
Below is selected dialogue content; the full text is about 10,000 words, with a reading time of about 6 minutes.
Highlights of This Episode
●
In my early years at Kleiner Perkins, I sat in on all of Google's board meetings from the first round to the IPO. Back then, Google took 19 years to go from zero to $100 billion in revenue and was seen as a growth miracle; today's Anthropic may need less than a third of that time to cover the same path.
●
When we entered Anthropic in 2023, they had no product and no revenue, and the valuation was as high as more than $4 billion. In a conventional VC model, this investment would not fit the rules of any fund, but when facing a real technological shift, you have to dare to make an exception.
●
When raising more than $500 million in a dedicated fund for Anthropic, for 30 to 45 days I kept worrying it would not get done. Anthropic supported us leading the round but could not share any financial information, and we had to ask institutions like state pension funds to put in money at an $18 billion valuation.
●
Sam Altman is more of a business figure; making big deals is his superpower. Dario Amodei's superpower is building great technology and recruiting great researchers; many of the world's best researchers only want to work for people who are like themselves.
●
Many people say open-weight models will completely replace foundation models; that is overcorrection. What consumers and enterprises ultimately want is the best product, and frontier intelligence is worth paying for.
●
Of the more than 70 newly emerging large-model labs on the market, many are raising $1 billion in Series A and $2 billion to $3 billion in Series B; this clearly will not end well. We will see spectacular acquisitions, and we will also see many big craters smashed into the ground.
●
Ten to fifteen years ago, few people specifically went looking for researchers; they were usually not good entrepreneurs. But in today's technological wave, top researchers have become the absolute core of startup teams.
●
I was there when the crash came in 1999, and today feels completely different. Back then only a small fraction of people had broadband and computers, and the other 98% could not get in; today's AI can be used immediately by every consumer and every enterprise.
Reading the Prospectus: Anthropic's Growth Rhythm and Google Back Then
Q: Anthropic has just released its prospectus, and many analysts are discussing the financial figures. As an early investor, how do you view these publicly disclosed data?
Matt Murphy: The company is still in a quiet period, so I can't comment too much on specific numbers. But when reading this prospectus, some outside interpretations can easily be misleading. For example, one headline said the company burned $42 billion, but much of that is non-cash adjustments under GAAP, such as the book accounting for warrants or convertible notes, and is not the company's actual cash burn at all.
Many people don't really know how to read a prospectus. The metrics that truly reflect a company's fundamentals are operating cash flow, gross margin, and growth rate. Many analysts and mutual funds also focus most on operating cash flow, and Anthropic's operating cash flow has already turned positive and is expected to stay that way. (Editor's note: The prospectus shows that Anthropic's operating cash flow was positive $3.93 billion in the first half of 2026, and negative $2.64 billion for all of 2025.)
Matt Murphy: If you look carefully at the data, you will understand that this claim doesn't hold up. Gross margin and operating cash flow are what truly matter to a company's underlying quality. I was very excited that the prospectus was made public; before this, there had been far too much speculation in the market about their business scale and growth rate.
Early in my career I worked at Kleiner Perkins. From our investment in Google to its IPO, I attended every board meeting. Even today, people probably still think Google is the fastest-growing company in business history: it took 5 years to go from zero to $1 billion, 8 years to reach $10 billion, and another 11 years to reach $100 billion, 19 years in total.
Q: Even by today's standards, that is still an extremely impressive achievement.
Matt Murphy: It certainly is remarkable. But today's Anthropic will likely complete that journey in less than one-third of Google's time. From this you can see that the growth speed of this wave of technology is completely different from the past.
Q: You are one of the very few investors who has observed early Google and Anthropic up close. What are the similarities and differences between these two companies?
Matt Murphy: The biggest similarity is that neither company, at the very beginning, anticipated what business model would ultimately make it big.
When Larry Page and Sergey Brin first founded Google, they were strongly opposed to advertising. At first they thought about charging by number of searches, and only later were they pushed into the advertising business.
Anthropic is the same. Dario Amodei and the team knew they had excellent models and APIs, but initially they had no idea they would become a company focused mainly on enterprise customers. At the time, OpenAI already dominated the consumer market, so Anthropic instead allied with Amazon and Google, obtaining not only compute but also important distribution channels.
This step brought them into the enterprise market, which proved to grow faster and be more profitable, helping them surge to first place in the industry. Top technical teams often have strong underlying technology, but exactly how to monetize it must be figured out continuously in the process of going to market. The founders of both companies, Dario, and Larry and Sergey back then, are all technical founders of this kind.
Raising a valuation of more than $4 billion with zero revenue: an exceptional moment in venture capital rules
Q: When you first invested in Anthropic, did they not have any revenue yet?
Matt Murphy: No revenue, and the product had not been released at all. When we invested in Google back then, it had also not yet begun to monetize, but at least we could see very high usage. When investing in Anthropic, none of that existed.
Q: At the time many investors gave up on that round. How did you decide to participate?
Matt Murphy: We initially participated in the Series C, and then served as lead investor in the Series D. During the Series C, we originally had the opportunity to lead, but at that time the company had no product and no revenue, yet its valuation had already reached more than $4 billion.
Today many people may think $4 billion is nothing unusual, but if you turn the clock back to 2022 and 2023, there were almost no companies like this in the market. In the entire history of venture capital, a portfolio company eventually being sold or going public at a $4 billion valuation is already an extremely brilliant result.
Q: And that was just your entry price.
Matt Murphy: Exactly. The most difficult question at the time was: which fund should this money actually come from? The early-stage fund thought the valuation was too expensive; even if it doubled, it would not help the overall fund's return much. The growth fund looked at it and saw no product and no revenue, so it simply could not get approved. It was stuck right between the two types of funds.
But fortunately, our partners were highly aligned on AI. The consensus at the time was very clear: first, this was at least the second-best company in this track, with the industry's top technical team, after all, these were the people who made ChatGPT; second, the model's evaluation results were very impressive; third, this was an enormous market.
So we made some adaptations to the existing rules. Although this investment was not perfect by the standards of any single fund, we decided to participate first, with the core goal of securing the lead position in the next round. It later proved that this decision was crucial. After becoming a shareholder, the two sides established a completely different relationship of trust. We helped them recruit people, meet customers, and deal with various strategic issues faced by early-stage companies.
Soon after, we saw them finalize partnerships with Amazon and Google one after another, and the business flywheel began spinning at high speed. On one hand, they sent deployment engineers on-site to help large telecom companies build proprietary systems; on the other, there was an endless stream of long-tail developers. When the product launched in April, the company's revenue was zero, but by year-end it was reportedly already around $85 million. We saw this momentum in September and October and immediately decided to compete to lead the Series D. Without that first unconventional check we wrote at the outset, we could never have built enough mutual trust to win that lead.
Q: Around 2022 the capital markets were very cold. In that environment, when you made the call on an investment like this, was the internal debate fierce?
Matt Murphy: The process was indeed full of pressure. The initial doubts were all typical VC arithmetic: "The valuation is already over $4 billion. Even if it rises to $9 billion, the lift to the fund's overall returns is limited. VCs usually chase 10x-plus returns—are we supposed to expect this company to be worth $40 billion or $45 billion someday?" At the time, that sounded crazy.
But the speed of this wave broke all prior experience. After we invested, large exits like Cursor emerged. Every previous technology wave had something holding it back—whether hardware devices were widespread, whether broadband was fast enough, whether enterprises were comfortable putting data in the cloud. This time was different: the models were open to everyone from day one.
Models are the heart and brain of the entire AI economy, and all applications are built around them. If you want to be a leading AI investment firm, you have to invest in one of the top two model companies. At first I also worried my partners would think I was crazy, but the team trusted one another and was willing to dream this dream together, so they were actually very supportive.
By the time we led the Series D and launched a special fund of more than $500 million, things were much clearer. Although the valuation had jumped to the $18 billion range, the company was growing fast, the models really were better, and both Amazon and Google chose it as their preferred partner, forming a landscape of OpenAI plus Microsoft versus Anthropic plus Google and Amazon. It has changed now, but that was how it was then. In the past, a financing cycle was usually one round every two to two and a half years, but these partnerships, and revenue going from zero to about $100 million, all happened in less than six months.
Q: Why are the world's top research talent willing to follow Dario Amodei?
Matt Murphy: Because he is a pure technologist. Many of the world's best researchers only want to work for people who are like themselves. Sam Altman is more of a business figure; closing big deals is his superpower. Dario's superpower is building great technology and recruiting great researchers. Researchers feel Dario is one of them and want to join his vision. In addition, Anthropic has always had a great culture with less drama, which stands out among peers.
The $500 million raise: a test where you can't look at financial data
Q: Let's talk about that special fund of more than $500 million. I recall your main fund at the time was $750 million?
Matt Murphy: It was $750 million. By the way, that was the first special fund Menlo had done in its 50-year history.
Q: Many established firms never do special funds. Why did you make an exception?
Matt Murphy: We used to focus on early-stage and early-growth investing, with early growth meaning the stage of $3 million to $10 million in annualized revenue. In the past, it took a company a year to a year and a half to pass through that range; now many companies do it in a week, and on the last day of a quarter they might book $10 million in a single day. So our strategy is changing too.
The turning point came at our LP annual meeting in November 2023. We invited Anthropic executive Neerav to share. He is an all-around operator; when the company was small, he managed both fundraising and major customers. At the time, people had not yet truly understood AI's capabilities, especially LPs from all over the country, who had no intuitive sense of what problems it could solve. Neerav explained it extremely clearly, describing where the technology was headed and what problems it would solve, and the response on site was very enthusiastic.
After the meeting, many LPs proactively asked whether there was an opportunity to invest alongside. Indicated funds quickly accumulated to $200 million to $300 million. We wanted to lead the next round, but that round was very large, and our $750 million main fund could put in at most $50 million to $100 million. So we decided to combine the main fund's money with the LPs' money to form a special fund of more than $500 million to lead the round.
But once we really got going, we found we had underestimated how difficult this would be. Many people expressed interest, but when you actually shook the tree, some fruit fell and some did not.
Q: What twists and turns happened along the way?
Matt Murphy: Over the following 30 to 45 days, I was constantly on edge, always feeling we had overpromised and might not be able to pull it off. Although Anthropic supported us leading the round, it set one condition: no financial information could be shared with LPs.
That meant we had to ask institutions to put in money at an $18 billion valuation while the only things we could show were a few pages of public product introduction and the previous year's revenue figures. No financial projections could be provided at all.
Q: That must have been hard for many large institutions to get through approval, right?
Matt Murphy: Many large state pension funds made it clear that without seeing any financial projections, they could not responsibly commit capital. We held several online information sessions and also communicated with them one by one.
But in the end, three invested. Based on the valuation from the most recent funding round, all three of those LPs made well over $1 billion. Fortune favors the bold.
We signed the term sheet in mid-December, basically finalized it in March of the following year, and closed in early April. This experience made me deeply appreciate how hard it is for founders to raise capital. During those months, a partner and I essentially became Anthropic's full-time fundraisers, going around to institutions on their behalf so that Dario and the technical team could focus on R&D.
At the time, it really did feel like a life-or-death moment for Menlo. If we had judged wrong and lost the state pension fund's money, the consequences would have been severe. But big bets always come with that kind of uncertainty, and I'm glad we ultimately pulled it off.
Breaking VC dogma: the "barbell strategy" and a new generation partner team
Q: Menlo has been around for 50 years. In venture capital, generational transitions often cause a firm to decline. How did you get through this one?
Matt Murphy: Poorly handled generational handoffs are the main reason many established firms fall behind. Usually the older generation stays too long, the best young partners leave, and in the end the firm misses an entire technology wave. If you're late to AI and you're not a top-tier firm, I don't know how you turn things around.
Menlo got through this step smoothly. The founder handed management to my business school friend Mark Siegel, and Mark in turn brought in Venky Ganesan, Shawn Carolan, and me. At the time, Menlo had a great brand but lacked a bit of edge: Were we hungry enough? Could the market feel Menlo's presence? Could we get ahead of others? That was our internal criticism. We made an important decision: voluntarily shrink funds that had already reached $1 billion or more down to $400 million, return to early-stage investing, and rebuild with a smaller team.
We also asked founders what they value most in a VC. Beyond a reliable partner, it's help with hiring and with negotiating deals. We had neither, so I brought in Jordan Ormont, who had worked with me at Kleiner Perkins for eight years, to build a talent team, and brought in Hooman from Qualcomm to handle business development.
Then the new technology wave arrived. We touched neither crypto nor VR, and thankfully we didn't. But when AI came, we felt this was our moment. Conviction across the firm was extremely strong, and we decided to go all in. That was late 2022, shortly after ChatGPT launched, about half a year before we invested in Anthropic.
English text is being prepared.
So in recent years we've brought in a group of partners with deep technical backgrounds, such as Tim Tully, who served as Splunk's CTO for a long time, Geoff Redfern, who was Atlassian's chief product officer, and a Stanford machine learning PhD. When we decided to invest in Anthropic, my conviction was strong, but it was Tim who spent a lot of time in deep conversation with Anthropic co-founder Tom Brown and figured out their compute multiplier: why they could train high-caliber models at far lower cost, and why that advantage could keep compounding. Otherwise, I would have been judging on intuition alone.
Being able to invest in Anthropic also depended on a high degree of internal trust. If everyone worked separately and someone brought back an exception, the answer would be "we don't do this," and then you start building fake guardrails.
Q: Many firms insist on getting 10% to 20% ownership in every investment. What's your view?
Matt Murphy: Sometimes even if you're a highly reputable firm, you can lose the competition. What do you do when you lose? Do you invest a small amount to build a relationship first, or do you simply give up because you can't get 10%, 15%, or even 20%? In this industry, as long as you stand behind the best founders and the best companies, you won't go wrong.
After leading Anthropic's round, we partnered with them to set up the $100 million Anthology fund, investing $100,000 to $1 million per deal, with the goal of broadening our view.
Our first investment in the LLM routing tool OpenRouter was only $100,000. I've been in this industry for more than 25 years, and I'd never invested in a company like this: 14 months after the investment, word came that Stripe was going to acquire it for $8 billion. This fund also invested in Wispr, a voice input product valued in the billions of dollars, and Goodfire—altogether it produced four or five very valuable companies.
Now we've institutionalized this approach: the fund has full-size seats and exception seats. Invest a little smaller first, hoping to scale up later. Of course, it can't all be exceptions, otherwise the fund is just a pile of scattered investments, with no board seats, and you can't build the kind of reputation that makes everyone want to put you on their board.
We now lean more toward a "barbell strategy": on one end, going earlier and doing true seed investments, spending more time at universities like Stanford and at technical labs, finding researchers who have just come out; on the other end, once a clear leader emerges in a sector, concentrating capital in a heavy late-stage position.
What we do less of is the middle ground between seed, Series A, and later-stage growth. At this stage, many companies grow very fast at first and have high valuations, but it's completely unclear who will actually win. There were a bunch of competitors in Lovable's space, and then it broke out and became a household name—you very likely invested in another one. Same with the legal space: Legora and Harvey sucked all the oxygen out of the room, and the other companies became irrelevant.
How to tell when a company has become the leader
Q: What signals do you usually rely on to judge that a company has already become the industry leader?
Matt Murphy: Growth rate is of course critical, but customer diversity matters even more. If a company's revenue comes mainly from two large customers with tens-of-millions-of-dollars orders, that doesn't prove the market is big enough; a true leader usually has a substantial long tail of customers.
Short sales cycles are also a very strong signal. Take the legal industry—it has always been the hardest field to sell software into. But Legora, which we invested in, had an absurdly short sales cycle, closed deals fast, and had extremely high usage, which usually shows that the market really is large.
Another intuitive signal is sales productivity. In the past, enterprise sales quotas per rep were usually $1.2 million to $1.8 million; among this batch of the best companies, a single salesperson can close $3 million to $5 million. And a lot of the business doesn't rely on sales pushing hard—the product itself drives growth.
Q: Anthropic is the most obvious example of this—the scale of its large customers is astonishing.
Matt Murphy: That precisely proves the value of the technology. If one large company can use it this way, why haven't more large companies? Probably they just haven't tried enough and haven't gotten their organizations to adopt it.
More importantly, look at the Fortune 5000—penetration of this technology is almost zero. Right now they've only done some coding and efficiency improvements; not a single one has truly redone its entire set of applications. The room ahead is still very large.
Why this isn't 1999
Q: Many people compare the current AI boom to the 1999 internet bubble. You lived through that crash—what are the essential differences between the two?
Matt Murphy: The 1999 boom was actually built on a very narrow group of people. At the time, people with high-speed broadband, a personal computer, and the willingness to use a credit card online were only a small fraction of the population. That group gave those companies a fast start, but after that the industry hit a ceiling, and the other 98% of people couldn't get in for a long time.
At the same time, many internet companies back then relied on massive ad swaps and had poor business models. In hindsight, there were three reasons 1999 hit a wall: first, the business models weren't durable; second, the addressable population was too small; third, there wasn't enough time to scale up.
Today's AI wave is completely different. From the moment it appeared, every consumer and every enterprise could use it immediately, and from the start it found scenarios on the enterprise side where people were willing to pay. Although some application companies still need to improve gross margins, the overall business foundation is far more solid than in 1999.
How frontier models and open-weight models will ultimately split
Q: There's now a view that open-weight models are rapidly catching up and that eventually no one will use expensive frontier models. What do you think?
Matt Murphy: That view is clearly an overcorrection.
The logic is very simple: when developing a new AI application, the team's only priority is to make the product extremely good first. At this stage, they don't care about cost and use only the strongest model available. Only once the product is really running and scaled up will the engineering team sit down and start turning the knobs.
Teams will repeatedly test: what happens if 70% of it uses frontier models? What about only 20%? I don't know what the final ratio will be—maybe more than half will be mainstream foundation models like Anthropic's, with open-weight models accounting for roughly 20% to 30%, depending on the nature of the application itself.
The most fundamental demand of consumers and enterprises is always a good product. Give them a bad product and they'll go elsewhere. You can't compete with second-rate intelligence; as long as frontier models can make the product noticeably better, the money is worth spending. Many companies are also currently lowering costs through caching and matching the right chips—that's a natural engineering optimization.
Q: But right now more than 70 so-called "new labs" have sprung up in the market, all wanting to follow the path of OpenAI and Anthropic—and you've invested in some of them?
Matt Murphy: We've invested in about ten, but not to directly compete for Anthropic's revenue—rather to find complementary models in adjacent fields.
For example, Skild, which we invested in, does robotics models and has to handle physical problems like gravity, which is completely different from training language models; Chai Discovery uses its own models for antibody development and drug discovery; Axiom Math does models for mathematics; Suno in audio and Wispr in voice both built their own models in order to create a good product experience.
Some teams are complementary in their methods—for example, Inception, which we invested in, uses diffusion methods to generate text, which is extremely fast and suited to scenarios like coding.
But of these 70 companies, many are raising $1 billion at Series A and $2 billion to $3 billion at Series B—that clearly won't end well.
But right now everyone wants this technology and these people so badly that we really will see some astonishing acquisitions. Take Fei-Fei Li's company—I don't know exactly what they've built or how much revenue they have, but I believe it was acquired for $8 billion before it had much revenue at all. At the same time, we'll also see a lot of big craters blown in the ground.
Where the $1 billion raised goes, and the rewriting of the software development process
Q: Many people don't understand why large-model startups routinely need to raise $1 billion. Where does all that money go?
Matt Murphy: The logic for model companies is relatively simple: training is extremely expensive. To compete for talent with top labs, you have to offer absurdly high compensation; and before generating meaningful revenue, you have to buy or rent large amounts of GPUs to train.
Moreover, the model itself is only half the battle—it's just a ticket to the game. Once you have a good model, you still have to figure out how to bring it to market. What's impressive about Anthropic is that very early on they weren't just selling models; they rolled out Claude Code for developers, Claude for Life Sciences for the life sciences, and Cowork, drawing users in step by step.
Application companies raise so much money often because growth itself is expensive. When the business expands quickly, you need a lot of forward-deployed engineers, salespeople, and customer support. These companies spend a lot not because they're inefficient, but because they're growing so fast.
Matt Murphy: The goal in the first stage is simple: "I need a model, whether OpenAI or Claude, to get the application built and get one use case working end to end."
In the second stage, once the company is up and running, the focus shifts to optimizing infrastructure and rewriting the entire software development process.
As large amounts of code are written by tools like Claude Code, new engineering bottlenecks emerge: how do you merge all this code? How do you test it? How do you ship it without crashing the system? How do you ensure security? These problems are spawning a whole new set of tools around testing, deployment, and security.
Q: Anthropic has filed its prospectus, and OpenAI is going public next year. What change will this bring to the venture capital market?
Matt Murphy: I don't think much change is needed right now. There have already been enough strong M&A deals to validate some of the companies we invested in and their valuations. If there were no active M&A market, and everyone had put their chips on the table with no exit, that would be scary. Acquisitions become a pressure-relief valve, and money returning to LPs is a good thing.
The big labs going public has another benefit: attention extends further down the line. Right now everyone is watching the two biggest companies; once they go public, people will start asking, who are the next 10 companies?
Q: Speaking of which, I seem to hear less about SpaceX after it went public.
Matt Murphy: That won't happen with Anthropic. SpaceX is a multi-asset business; our Silicon Valley echo chamber doesn't talk about rocket launches every day. What people are really talking about is the xAI part, but it's not as exciting as Anthropic.
Getting out of the "afraid to make mistakes" mindset
Q: After 25 years in venture capital, what's the biggest change in you?
Matt Murphy: The biggest change is learning to be flexible and not be bound by ingrained experience.
By nature I'm a rule-follower; I grew up within the rules of this business. There's a lot of valuable experience in that, but when times change, you need to recognize that the old rules may no longer apply and dare to make exceptions.
At the same time, listen more to the judgment of younger partners and technical partners, and don't think you know everything. Investing is a team sport, and they make me better.
More importantly, you need courage. Early in my career, at least for me, the biggest worry was that I must never make a mistake. The famous quarterback Alex Smith described a similar experience: early on, his coaches kept telling him "don't screw up"; then he met Coach Harbaugh, who told him: "Go out and play, don't worry about making mistakes. We'll give you the plays; if it doesn't work, that's on us, not you."
It also took me a long time to get out of that fear-of-making-mistakes state. Venture capital requires you to play more to win, not to be afraid of losing. Win as a team, accept that this business is hard and there will be many losses, and face it with grace.
Source: The Peel podcast, Turner Novak in conversation with Matt Murphy, aired October 26, 2026, https://www.youtube.com/watch?v=kQcX4sOaW90AI Industry Signals Channel: Frontier Technology. This content comes from the internet, original link; the views represent only the author and do not represent the position of Huxiu. For copyright issues, please contact hezuo@huxiu.com, and we will verify and handle it promptly. Everyone who is changing the world and wants to change the world is on the Huxiu APP[ ]