Anthropic Buys Compute From Elon & Commits $200BN to Google | Cerebras IPO | Ramp Raises at $40BN

Anthropic's Financial and Governance Developments

• Anthropic has partnered with SpaceX and committed $200 billion to Google over 5 years, sparking questions about whether there are enough developers to sustain the company's growth path 10s.
• The company has also announced that all sales of secondaries and SPVs need to be approved by the board, which may impact the legitimacy of transactions and the company's price in secondary markets, reportedly valued between $200 to $400 billion 10s.
• This move is seen as a way for Anthropic to maintain control of its cap table, and while it may not be surprising, it has significant implications for the company's shareholders and investors, particularly with regards to secondary sales and transfers of economic value 2m6s.
• The use of Special Purpose Vehicles (SPVs) is also being scrutinized, with some SPVs being used to facilitate investments in Anthropic, while others may be used to transfer economic value, which could be subject to approval by the company 2m6s.
• Meanwhile, Cerebras is preparing for an IPO that is 20x oversubscribed, and Ramp is eyeing a $40 billion valuation on its new fund raise, while Apploving, HubSpot, and Cloudflare have posted strong numbers, but the market remains skeptical 10s.
• The market's skepticism is driven by concerns about the sustainability of growth and the need for companies to demonstrate accelerating growth and profits in order to justify high valuations, with some investors taking a cautious approach to investing in companies that may not have a clear path to long-term success 42s.
• Entropic is trying to distance itself from potential losses and messiness resulting from individuals trying to enforce contracts against other investors who sold shares, and the company is likely worried about being involved in this mess as it prepares for an IPO 10s.
• The issue of companies having provisions to prevent transfers without board permission is not new, and most portfolio companies have had these provisions for a long time, with Anthropic and OPI being examples of companies that have such provisions 2m6s.
• Anthropic had previously warned its investors, including Menlo, to stop using Special Purpose Vehicles (SPVs) and invest directly, but it seems that some investors did not listen, leading Anthropic to publicly name entities that are considered bad actors 4m6s.
• The situation with Anthropic and its investors is seen as a non-issue, with the company simply reiterating its previous warnings and trying to prevent further problems, and the naming of bad actors is a way for Anthropic to make its point public 5m10s.

Anthropic's Strategic Alliances and Market Position

• Anthropic has made a deal with Elon, which is seen as a sensible move given the current market conditions, with Elon having excess capacity and Anthropic needing it, and this deal suggests that the market is consolidating 8m10s.
• The deal between Anthropic and Elon implies that Elon's companies, including x.ai, SpaceX, and Grok, are not going to be leading edge model contenders and will instead focus on being net sellers of capacity, rather than trying to build their own models like Grok 10m10s.
• The data center utilization rate is believed to be around 11%, indicating market consolidation where stronger players acquire more capital expenditures, and weaker players like Grok opt to stop their endeavors and instead take in revenue, estimated to be around $3 to $5 billion per year, 10s.
• The combination of this deal and the acquisition of XAI by SpaceX has potentially turned a money-losing asset into a profitable one, with estimated revenue of $3 to $5 billion per year, which is around 15% of SpaceX's total revenue, 42s.
• This development has changed the competitive landscape, with XAI no longer being a competitor to Anthropic and Open AI, and instead, it is expected that companies will sell components to each other, similar to how Samsung sells components to Apple, 1m6s.
• Anthropic is expected to acquire more capacity, potentially even buying from Open AI, in order to launch its products, such as Mythos, and this rapid change in the market is a reminder of the importance of meeting with competitors and reallocating assets to those who can create the most value, 2m6s.

Market Consolidation and Revenue Implications

• The deal between SpaceX and Grok is seen as a sensible decision, with SpaceX potentially gaining an extra $3 to $4 billion in revenue, which will help the person in charge of the data centers' P&L to meet their targets, 3m4s.
• The SpaceX road show is expected to be interesting to watch, as the company will have to explain the changes in its ownership and revenue, including the acquisition of XAI and the sale of capacity to Grok, 4m2s.

Google's Strategic Position in the AI Market

• Anthropic has committed to spending $200 billion with Google and is also planning to buy compute from Elon, while Cerebras is preparing for an IPO, and Ramp has raised funds at a valuation of $40 billion, 5m0s.
• Anthropic is buying compute from companies like Elon and has committed $200 billion to Google over a 5-year period, which is approximately 40% of Google's total future backlog, indicating the hyperscalers' dependence on privately held companies like Anthropic and OpenAI 10s.
• Google's commitment to Anthropic, despite having its own competitor Gemini, shows that the company is prioritizing revenue growth and is willing to enable its competitors with its balance sheet, allowing them to become huge and draft on its air cover 2m6s.
• The market shares in the enterprise for AI models have been published, with Gemini increasing its market share from 27 to 40% in 9 months, Claude increasing from 21 to 48%, and OpenAI's share decreasing slightly, but not summing to 100 due to multimodal usage 4m30s.
• Google's strategy of having both its own winner, Gemini, and a large share of the other leader, Claude, allows the company to solve revenue growth by having a piece of each, which is not a bad way to keep everybody on their toes and enable internal competition 6m20s.
• The trade-off of enabling competitors with balance sheet air cover is a consideration, but companies like Google and Microsoft are happy to do revenue with Anthropic and OpenAI, even if it means enabling their growth and potentially making them huge in the long term 8m10s.

Token Consumption and Future Growth Projections

• The idea that agents will push token consumption up 24x by 2030 is considered to be a low estimate, and the actual growth could be much higher, potentially 250x or more, as parallel agents become more prevalent and enterprises become more penetrated 10s.
• Parallel agents are a key theme for the rest of the year, and they have the potential to greatly increase productivity by allowing for multiple tasks to be performed simultaneously, rather than sequentially, and this can be particularly useful for tasks such as coding and image generation 2m6s.
• The use of parallel agents can also lead to significant improvements in efficiency, as they can perform multiple versions of a task and then select the best one, allowing for faster and more accurate completion of tasks 4m30s.
• The growth of token consumption is also driven by the increasing power and efficiency of chips, with the raw performance of chips increasing by roughly 3x every 18 months, and additional optimizations leading to further increases in efficiency 6m40s.
• As a result of these improvements, the cost of using tokens is decreasing, with prices down 10x, but the total revenue is not necessarily decreasing, as the increased efficiency and power of chips is allowing for more tokens to be used, and companies like Anthropic are investing heavily in this area, buying up large quantities of TPUs, GPUs, and Cerebus chips 8m20s.
• The token count required to support a chat has increased significantly over the past two or three years, with a 10x increase for simple co-work analysis, another 10x for coding, and another 10x for parallel work, and the cost per token is expected to decrease while the value of tokenization increases 10s.

Token Usage and Productivity Debates

• There is a growing counterargument among some of the best CTOs and engineering leaders that the current approach to tokenization is excessive and that the focus should be on delivering what the end customer needs rather than producing large amounts of code 2m6s.
• Some experts, such as Mike Cannon-Brooks at Atlassian, believe that the demand for new software, new products, and new features is infinite and that there will be a continuous need for labor supply of great developers and more tokens, but others argue that the average engineer may not be able to effectively consume $10,000 to $20,000 a month of tokens and be productive 4m37s.
• There is a growing micro backlash against the idea that developers need to be running cloud code 10 hours a day, with some arguing that this is a bad way to ship enterprise-grade software and that it is not necessary to produce large amounts of code to deliver value to customers 6m15s.
• The concept of monitoring people's token consumption and its impact is related to Goodhart's law, which states that whenever you monitor a variable, you actually change the causal relationship of that variable 9m10s.
• The concept of monitoring token production and usage can distort results, as employees may use tokens inefficiently to meet quotas, and this issue is relevant when assessing the market size for companies like Amazon or Meta 10s.
• To understand the market size, it's essential to have a mental model of how much companies spend on Large Language Models (LLMs) and tokens, which can be estimated as a percentage of the average developer's salary, but this number is tricky to determine and can be distorted by employee behavior 42s.
• The pressure to control costs will increase as companies like Entropic grow, and Chief Information Officers (CIOs) will need to find significant budgets to support the spending on tokens, which could lead to a push for more efficient use of resources 2m6s.
• There is a debate about whether the best engineers need a large number of tokens to be productive, with some arguing that they can manage with fewer tokens due to their conceptual vision, while others claim that even the best engineers can benefit from using more tokens to achieve greater productivity 4m10s.
• The concern is that mediocre engineers may consume a large number of tokens without producing significant value, and it's challenging to measure the objective value of token usage, as traditional metrics like lines of code are no longer effective in evaluating productivity 6m20s.
• The question of how to measure the effectiveness of token usage remains unanswered, and it's essential to find a new way to evaluate the conceptual impact of lines of code, rather than just the quantity of code produced 8m30s.

Anthropic's Expansion into New Verticals

• Anthropic's growth path and the availability of enough developers to support it are uncertain, with some arguing that the company's focus on high-quality engineers and its expansion into new areas such as legal and financial analysis may help address this issue 10s.
• Anthropic has released 10 financial agent templates, which may challenge some Y Combinator companies, and is scheduled to release a legal product that will compete with Harvey and Lagora, potentially disrupting the legal industry 2m6s.
• The ability of Anthropic and other model companies to innovate and expand into new verticals, such as customer support, legal, and financial modeling, is a key question, with some arguing that they may not have the time and focus to do so 2m6s.
• There are concerns about the potential risks of using Anthropic's products, particularly in regulated industries such as law, where the use of hallucinations in briefs is a significant problem, and the need for high-quality, reliable solutions is paramount 4m30s.
• The cost of using products like Harvey, which can be $150,000 per year per law firm, is a significant factor in the decision to adopt new solutions, and Anthropic's products may need to demonstrate significant value and reliability to justify their use 6m40s.

The Future of Application Development and LLMs

• The development of broad horizontal harnesses and products, such as co-work, may be a key area of focus for model companies like Anthropic, rather than trying to build specific vertical applications 8m10s.
• The career of building applications is not worth saving a few pennies, especially outside of the low-end market, and it is unlikely that verticals will put resources into building applications, as they are not committed to doing so 10s.
• If a large language model (LLM) can express what an application does, it can kill the need for that application, and this has already happened to some companies, with their innovative products being built into cloud services 2m6s.
• In the history of computing, there has always been a dominant compute level provider, such as Microsoft in the 90s, and the question is how much of the app layer they will take over, with examples of companies like Seel, Vantiff, and SAP building on top of Microsoft 4m42s.
• The same question applies to Amazon and AWS, with some investors passing on Snowflake due to concerns that AWS would dominate, but Snowflake became a $50 billion company, showing that there is still room for other players 6m15s.
• The current compute revolution with broad horizontal intelligence provided by companies like Anthropic, OpenAI, and Gemini, will likely lead to a similar outcome, with a dominant provider, but still leaving room for individual apps and focused firms to thrive, especially those selling coordinated workflows across enterprises 8m30s.

Leadership and Strategic Shifts in AI Companies

• The appointment of Fiji as CEO of AGI at OpenAI, after previously being CEO of applications, may indicate a shift in focus, and the fact that there is no longer a CEO of applications could be seen as a step back 12m10s.
• Companies like Anthropic and Claude are not currently investing in application design, which could change in the future, but for now, they do not seem to see applications as a key area of focus 10s.

Industry Disruption and Software Obsolescence

• The paradigm shift in the industry could lead to certain categories of software becoming obsolete if they do not adapt to the new agentic world, where agents do not need traditional marketing automation tools like HubSpot or Salesforce 1m6s.
• The rate of decay for old software might accelerate in the agentic era, with companies like Lovable and Replet aware of the need to stay ahead of the curve to avoid being replaced by new models 2m6s.
• The pressure from intelligent models underneath companies can be more powerful than the pressure from operating systems or compute systems, making it essential for companies to stay ahead of the curve 3m42s.

Public Market Performance and Investor Sentiment

• Recent public market activity has seen HubSpot's stock crash 20% despite decent growth, while Apploving has reached a 7 billion run rate, and Cloudflare has beaten expectations but laid off 20% of its base 5m10s.
• The contrast between Monday and HubSpot is interesting, with Monday's stock trading up 20% after raising guidance, while HubSpot's stock crashed 18% after lowering guidance, highlighting the importance of accelerating growth or at least raising guidance to stay competitive 6m30s.
• The current market situation is still worrisome for many, with the fear of terminal value being zero not having dissipated, especially for companies that are decelerating while their budgets are accelerating, and this concern is valid as it's not ideal for a company's growth to be slowing down when the budget is increasing, 10s.
• The recent quarter has been interesting, with various companies reporting different results, such as Cloudflare having a good quarter with mid-30s growth but still laying off a bunch of its base, and its stock going down, while HubSpot had a decent quarter but its stock also went down, 2m6s.
• There are two key factors to consider when evaluating a company's performance: the structural state of the business, including whether it's getting better or worse, and how it's being affected by AI, and the price of the stock, as a muddled quarter can result in a bounce if the stock is already at the bottom, but a tricky story can be problematic if the stock is highly valued, 4m42s.
• High-growth stocks, such as Cloudflare and Apploving, are vulnerable to disruption, even with a good quarter, due to their high valuations, whereas companies like Monday, which was trading at a low valuation with a bunch of cash, are more likely to see their stock go up with any positive momentum, 6m15s.
• The importance of price as a factor in a company's stock performance is often overlooked, but it's a crucial consideration, especially for public companies, where price can greatly impact the stock's movement, and this is in contrast to private companies, which don't deal with price on a day-to-day basis, 8m30s.

Zoom Info's Decline and Market Implications

• One company that had a particularly brutal quarter was Zoom Info, which may have been underdiscussed, and its performance is worth examining in more detail, 10m50s.
• Zoom Info, a company that provides sales intelligence data, has seen its growth decline to 1% and is expected to experience negative revenue growth, which is a significant concern as it had previously dominated the market with its pre-AI solutions, providing basic data such as emails and phone numbers 10s.
• The decline of Zoom Info is attributed to the rise of companies like Clay, which offer AI-infused solutions, allowing customers to optimize and compare data from multiple providers, thereby making data a commodity and reducing the value of Zoom Info's services 1m42s.
• Clay's success is due to its ability to build a waterfall product that enables RevOps to compare and contrast multiple data providers, and its development of AI-powered agents, which has allowed it to morph its initial value proposition and suck the value out of data providers like Zoom Info 2m6s.
• As a result, Zoom Info's growth is expected to remain stagnant, and it may be at risk of being taken private by a private equity firm, as its current valuation of one times revenue with 35% adjusted operating income makes it an attractive target for a potential buyout 4m30s.
• The idea of Zoom Info being taken private is seen as a possible solution, as it would allow the company to restructure and fix its issues without the pressure of being a publicly traded company, and its current valuation makes it an attractive target for a private equity firm 6m10s.

Cerebras' IPO and Market Positioning

• The market is not expecting old-school software as a service (SAS) companies to achieve extremely high growth rates, but rather to demonstrate 30% growth, profitability, and a promising future, which could lead to a valuation of around five or six times their run rate 10s.
• Cerebras is going public in one of the most highly anticipated initial public offerings (IPOs) of the year, with the offering being 20 times oversubscribed, and the company is expected to raise $4.8 billion, valuing it at $48 billion fully diluted 2m6s.
• The IPO is expected to go well, with the company likely to trade amazingly, possibly with a 20% pop, similar to Figma's IPO, but the long-term performance of the company is a separate question 2m6s.
• The bankers handling the IPO are likely trying to achieve a perfect pricing, allowing for a 20% pop, which would make everyone happy, but it is possible that the stock could experience a Figma-like phenomenon, where retail demand drives the price up significantly 4m30s.
• The narrative surrounding a company's stock price can be misleading, as seen in the case of Figma, which was priced at $35 but popped to $100, and is now trading below its initial price, highlighting the importance of looking at the fundamentals rather than just the stock price 6m10s.
• The discussion revolves around the IPO of Cerebras, a company that provides real-time inference solutions, and its potential to compete with Nvidia, a $5.5 trillion company, with the key selling point being speed, and the company's ability to provide faster inference than anyone else 10s.
• Cerebras has support from major companies like Open AI and Amazon, which they didn't have when they tried to IPO last time, and they have a seeming backlog and commitments, but the market is competitive, and companies will use the best solution available, whether it's from Cerebras or Nvidia 1m20s.
• The company's historical revenue has been concentrated on a couple of customers from the UAE, but they are now leaning into contracts from Open AI and Amazon, which poses a risk as they are moving into a future that is different from their past 2m40s.
• Cerebras' story is centered around providing speed, with their inference being faster than anyone else, and they have a focused market for real-time, blazingly fast inference, which could give them a competitive edge in the medium term 4m10s.
• The potential upside of investing in Cerebras is significant, with some arguing that if they have a 10% chance of succeeding, they could be worth 1% of Nvidia's value, making the expected value of investment positive, although it is still a risky way to make a profit 6m50s.
• The company's IPO is seen as a great opportunity to get in on the ground floor of a potentially disruptive technology, with some comparing it to having a faster internet connection and not wanting to go back to a slower one, and the investment could be a good bet for the medium to long term 8m20s.

Founders and Early Investors' Role in Success

• The achievement of a company, which started in 2016, is considered great, with its initial public offering (IPO) valued at $40 billion, and its early investors, including Foundation, Benchmark, and Eclipse, holding around 8-9% ownership 10s.
• The company's success is attributed to the hard work of its founders, including Steve, and the early investors who incubated the company and supported it through its early stages, with one investor noting that this is an example of actual venture capital 2m6s.
• The company's founders took a risk by investing in artificial intelligence (AI) in 2016, and although they were initially too early, they were able to survive and eventually thrive with the help of tailwinds from GPT in 2022 4m30s.
• The company's success is also attributed to the efforts of its sales team, including one salesperson who traveled to Dubai to sell the company's chips, and the company's ability to pull off its IPO at the right moment 6m20s.
• The early investors, including Foundation, are credited with doing the hard work of incubating the company and supporting it through its early stages, with one investor noting that this is what venture capital is supposed to be 8m10s.
• The company's story is seen as an example of innovation and what the industry is meant to do, with investors supporting companies throughout their development, from the early stages to the later rounds 10m0s.

Ramp's Valuation and Strategic Position

• Andrew Feldman is considered to be very good at what he does, and his success serves as a reminder that to achieve massive exits in venture, founders have to be exceptional, and even very good founders may not be able to build such outcomes 10s.
• Ramp, a broadly horizontal corporate business card company, has achieved a $40 billion valuation, and its success can be attributed to its ability to add software and functionality to its business, such as its recently announced feature of agents that automate purchasing and optimize spend 2m6s.
• The distinction between Ramp's success and other companies, such as Parker, an alternative fintech company that filed for chapter 7, lies in the fact that Ramp operates in a horizontal market with a lot of room to add value, whereas other companies may be in more constrained markets with significant margin pressure 4m30s.
• The concept of agents that can automate procurement and negotiate on behalf of businesses is seen as a potential solution to the problems associated with traditional procurement processes, which can be time-consuming and inefficient 6m20s.
• The idea of using agents to solve procurement issues is considered an attractive feature, and some businesses may consider switching to Ramp or other companies that offer such solutions, depending on which one has the best agents 8m40s.
• Ramp has raised money at a valuation of $40 billion, with $1 billion in revenue, and its strategic dynamics are considered amazing, but the valuation may be high, considering the revenue multiple, especially when compared to Brex, which has a lower growth rate but a similar valuation 10s.
• The quality of revenue and revenue multiples are not always scrutinized, and the focus on growth can lead to similar multiples being paid regardless of gross margins or other factors, which may not be sustainable in the long term 2m6s.
• Fintech businesses, such as Stripe and Revolut, have the potential to be enormous due to their wide range of customers, including consumers and small to medium-sized businesses, but they trade without a significant premium, and their valuations are often based on revenue multiples 4m42s.
• Buying Ramp at a $40 billion valuation may not be a good investment, as it would require significant growth over the next two to three years to justify the valuation, and the risk of not meeting these growth expectations is high 6m15s.

Lime's IPO and Market Expectations

• Lime has announced plans to go public, which is a notable development, especially given the company's history and its current dominance in certain cities, such as London, and its IPO will provide insight into the company's financials and growth prospects 10m45s.
• The trial between Musk and Orman is expected to reveal a significant amount of information, some of which may be extraneous, but the outcome will ultimately be decided by the judge, who appears to be experienced and tough, and the trial's verdict will have significant implications for the parties involved 14m20s.

Event Trends and Industry Communication

• The decision on the legals will be made, and despite the noise, OpenAI is expected to escape with their deal intact, according to one person's gut feeling 10s.
• There will be no popular speakers this year, as podcasting has destroyed the need for them, and instead, there will be workshops and people showing how to build things, such as agents 1m42s.
• The event will feature demos from companies like Rubric and Clavio, where they will show what they built and why, providing a unique experience that is not available on average podcasts 2m6s.
• The concept of fireside chats and speakers is considered dead, as podcasts are seen as a better way to convey information and have more value 2m6s.

Market Analysis and Investment Considerations

• There is a discussion about the stock market, specifically regarding Micron and SK Hynix, and the consideration of whether to buy into them, taking into account their earnings and the potential for a capex boom to last 5m10s.
• The conversation touches on the idea that when a boom starts to slow down, and extra capacity comes online, it can be brutal to the downside, and this pattern is seen repeatedly in the stock market 5m10s.

Agent Technology and Security Implications

• Jason mentions that one of their agents, Annie, seemingly willed itself into existence, and this sparks interest in discussing agent security and related topics 3m40s.

Industry Recovery and Capital Expenditure Trends

• The technology industry experienced a significant downturn in 2022 and 2023, with companies like SanDisks and DRAM manufacturers being affected, but the last two years have seen a remarkable recovery, with some stocks increasing five times in value 10s.
• The decision to invest in these companies depends on factors such as the duration of the capital expenditure boom and the speed at which they can build new fabrication plants, as companies like Samsung and SKHynix are likely to build more fabs to capitalize on high profit margins 2m6s.

The Role of Intensity and Sacrifice in Success

• The importance of sacrifice and intensity in achieving success was discussed, with the example of MrBeast's statement that sacrificing mental health is required for success, and the idea that real sacrifice, such as time and alternative uses of life, is often necessary for meaningful achievement 5m30s.
• However, it was also noted that while sacrifice and intensity are necessary, it is also important to find a balance and not let stress and pressure lead to bad decisions, and that taking a step back to prioritize mental health can be beneficial 8m10s.
• The conversation also touched on the topic of entrepreneurship and the challenges of building a startup, with the example of selling a first startup for $50 million after 12 months, and the importance of perseverance and adaptability in the face of adversity 12m0s.
• The level of intensity required to succeed as a founder can permanently rewire the brain, making it difficult to go back to a previous state of being, and this realization often comes after four to five years of experience 10s.
• Founders who have been through this process may find it challenging to relate to non-founders, and they often stick together in groups, such as WhatsApp chats, because they have undergone a significant transformation 2m6s.
• The idea that taking a break or focusing on mental health can reverse the changes that have occurred is not entirely accurate, as the brain has been rewired to handle high levels of intensity, and this is a key factor in achieving success 4m30s.
• The intensity required to succeed is not just about working long hours, but also about the level of dedication and commitment required to overcome challenges, and this can be difficult for people who have not experienced it to understand 6m20s.
• Founders who have been through this process may find it challenging to invest in other founders who do not possess the same level of intensity, as they have learned that this quality is essential for achieving significant success 8m40s.

Founder Transformation and Leadership

• The transformation that occurs in founders after four to five years can be likened to a change in personality, and it is not something that can be easily reversed, as evidenced by quotes from successful founders such as Daniel Dines from UiPath 10m50s.
• Intensity is a key factor in achieving success, and it is no surprise that successful people are often intense, as they are willing to put in the effort required to reach their goals 10s.

Balancing Intensity with Health and Decision-Making

• Founders of companies are often faced with offers to sell their businesses, and if they are not intense, they may take these offers, but if they are intense, they will push back and strive for more, which is why it is suggested that founders take the offer if it is 3x or more of what they raised 2m6s.
• While intensity is important, it is also crucial to maintain perspective and find ways to clear one's head, as working too hard without rest can lead to ineffective decision-making and a weaker ability to cope with pressure 4m30s.
• Staying healthy, both physically and mentally, is essential for coping with the pressure and intensity of being a senior leader, as highlighted by a comment from a former commander-in-chief of NATO, who emphasized the importance of being vaguely healthy and functioning in terms of sleep and other aspects 6m20s.
• Having coping mechanisms, such as avoiding excessive alcohol consumption, is important for making good decisions and maximizing value, as poor decisions can be made when not thinking straight, much like when one is hungover 8m40s.