Mamoon Hamid: AI - Where Value Accrues, Startups vs Incumbents & Scaling Laws | E1217

Intro (0s)

• Products that create markets, such as Slack and Figma, have the opportunity to define the playing field and ultimately win the game (0s).
• The top 20 jobs in the US, in terms of highest pay, include doctors, lawyers, and developers, who are highly skilled and scarce, and investing in application layer companies can help supercharge these professionals (11s).
• The current time is considered the most exciting and hardest time to be in venture capital, with the AI super cycle presenting a massive tidal wave of change (51s).
• The AI super cycle is reminiscent of the rise of the internet in the late 1990s, but with a much greater magnitude and the presence of incumbent companies with significant spending power (1m15s).
• Incumbent companies such as Google, Microsoft, Amazon, Meta, and Oracle are expected to spend hundreds of billions of dollars on frontier models, which is a fundamentally different level of spending than in the past (1m56s).
• Despite the rise of corporate investors with different motives and incentive structures, the opportunity for venture investors remains, with the potential to build on top of the infrastructure created by frontier models and create trillions of dollars in value over the next decade (2m34s).
• The current environment is considered a super cycle, with the potential for significant growth and change, and venture investors are well-positioned to take advantage of this opportunity (1m7s).

Where Will AI's Value Accrue, and Where Should Capital Focus? (3m1s)

• The AI landscape is currently over-invested, with both large incumbents and venture investors putting in capital, making it essential to concentrate time and capital in areas where the most value will accrue (3m2s).
• Venture investors have invested in application layer companies that solve specific pain points, focusing on helping highly scarce and skilled professionals such as doctors, lawyers, and developers do their jobs better (3m36s).
• To address the scarcity of skilled professionals, companies are building software and AI that helps them, such as co-pilots for doctors, lawyers, and developers (3m49s).
• Examples of companies that have been backed include those that help doctors, lawyers, and developers, such as Harvey, Ambience, and Codium (4m4s).
• Differentiation in the AI space is crucial, especially with multiple alternatives in every category, and it comes down to teams that can out-hustle and outwork others, with a focus on technical depth and quality of output (4m22s).
• The quality of the output of AI models is critical, particularly in fields like medicine, where a medical transcriber needs to be close to 99% accurate (4m51s).
• Successful companies in the AI space often have founders who are extremely technical, with deep machine learning expertise, paired with a domain expert co-founder who understands the market (5m8s).

Is AI Investing Different from Traditional SaaS Investing? (5m28s)

• Investing in AI is not different from traditional SaaS investing, as the core job of venture capital is to invest in early-stage companies that make history and are generational in nature, recognizing trends and tectonic shifts in technology, and investing in the right people and markets at the right time (5m52s).
• The current state of the system has high entropy, with rapid changes making the job of investing both fun and challenging, particularly from a pricing perspective (6m10s).
• Navigating the pricing environment requires not falling victim to excitement-driven investments, instead focusing on getting ownership at early stages with reasonable valuations, such as investing $5 to $10 million for 15 to 20% equity (7m4s).
• To balance this, a small portion of the strategy, around 20%, can be allocated to breaking the rules and investing in exceptional founders and companies, even if it means deviating from the usual investment criteria (7m27s).
• This approach is referred to as the "YOLO bucket," where extreme conviction in a founder and company leads to a willingness to break the rules, but this is done by exception rather than as a norm (7m33s).

Sustainable Growth vs. Sugar High Revenue (7m49s)

• Revenue scaling has increased significantly, with companies now achieving 11x growth, which is unprecedented, and this trend is seen across the board (7m50s).
• In the age of AI, companies are not just providing software, but also labor and capabilities that enable people to do 10x the work, helping to multiply abilities or bring costs down (8m14s).
• The shift from seat-based pricing to labor-based pricing is evident, with companies now charging $300-$500 per month for software, resulting in rapid revenue growth (8m40s).
• Companies like Clara are replacing labor with AI tooling, and the question is whether the next generation of companies will build their own custom tooling and replace existing SaaS solutions (9m25s).
• Building custom tooling can be costly, as seen in the example of Kleiner Perkins building an internal CRM, which cost millions of dollars to upkeep (9m42s).
• However, with AI, it may be more feasible for companies to hire developers to do labor work in the form of AI, and some companies may choose to do this internally (10m26s).
• Nevertheless, there will likely be companies that specialize in providing AI solutions, and businesses will have to pay for outcomes, such as the number of tickets resolved by the software (10m37s).
• The question is how many companies will choose to build their own custom AI solutions, and how many will opt to pay for external solutions (10m55s).
• Currently, many companies are in the proof-of-concept phase, trying out various AI solutions and seeing what works, but this can be challenging, and some may realize it's harder to build custom solutions than expected (11m1s).
• This reminds us of the early days of the internet, when companies spent a lot of money on internal projects, only to later hire consultants to help them (11m23s).
• The current advancements in AI and the internet are similar to the early days of the internet, where people were hesitant to put their credit card information online or find partners online, but now these actions are common and accepted (11m35s).
• In the future, customer support experiences are expected to improve significantly, with the possibility of resolving issues through text messages and automated agents, eliminating the need for human customer support agents (11m57s).
• The use of hold music, even by CEOs, is seen as unprofessional and annoying, and some companies are professionalizing this aspect of customer service (12m48s).
• In the investing world, a lot of time is being spent on the middle layer between foundation models and applications, including middleware that enables the use of these models more efficiently and effectively (13m21s).
• New technologies are emerging in this middle layer, such as vector databases and fine-tuning of models, which allow for faster and cheaper application development (13m45s).
• However, some of the value in this space seems fleeting, and the high rate of change in the early stages of development may lead to over-investment in certain areas (14m8s).

Will Foundation Models Subsume Application Layer Companies? (14m18s)

• Foundational model companies have the potential to subsume application layer companies if they are big enough, with examples including talking translators and talking avatars that can be interacted with in a friendly way (14m19s).
• The business model of hyperscalers, which provide the models, compute, and electricity, and charge by the hour or kilowatt hour, is considered a good business model (14m41s).
• Open AI's positioning is that they cannot do everything and want other companies to build the application layer, indicating a great business opportunity in providing the compute and electricity, as well as in being vertically focused around applications (15m3s).
• The LLM layer may not be a great business due to price dumping and commoditization, with the price of a token going down by 200x in the last 18 months (15m41s).
• Companies providing GPUs, data centers, and LLMs on top of compute infrastructure are part of the ecosystem, with Nvidia being a great business and hyperscalers investing for the future (16m17s).
• The margins of providing LLMs or selling tokens may not be great, but companies like Open AI are expected to figure out how to become highly profitable over time (16m56s).
• The scaling laws are expected to continue, with the price of a token potentially going down by 10x or 20x in the next two years, and the possibility of seeing 10x or 20x better models (17m18s).
• The rapid progress in the field is evident, with many applications utilizing the technology, and the potential for significant improvements in the future (17m29s).
• The article "The $600 Billion AI Question" by David KH at SEOA highlights the significant gap between the costs and revenues of AI companies, with $600 billion spent on capex and revenues lagging behind (17m56s).
• The world's GDP is approximately $1 trillion, with labor accounting for 50-60% and technology making up around 15%, which is expected to grow to 20% over the next decade, resulting in $25 trillion in technology companies and $10 trillion in annual spend (18m21s).
• The growth of technology companies is expected to create a significant market for AI, with the potential for $600 billion in revenue, driven by the need to address labor shortages and automate tasks that humans find unpleasant or time-consuming (19m0s).
• The impact of AI will be felt in various industries, including healthcare, where there is a shortage of doctors, and the legal profession, where there is a shortage of lawyers, and AI can help address these labor shortages (19m30s).
• The nature of the venture landscape is changing rapidly, with more capital available than ever before, but this has also led to overfunding of some companies, while others may be underfunded (20m27s).
• Despite the changes in the venture landscape, the core business of backing incredible founders who are building the right product at the right time remains the same (19m57s).

Where Does Kleiner Perkins Fit: Boutique or Capital Accumulator? (20m50s)

• Venture landscape can be divided into boutiques and capital accumulators, with boutiques including USV and Benchmark, and capital accumulators including Tiger, Coatue, Andreessen Horowitz, General Catalyst, Lightspeed, and Sequoia (20m50s).
• Kleiner Perkins is primarily early-stage focused, with an $800 million fund for early-stage investments and a $1.2 billion growth fund (21m11s).
• The team at Kleiner Perkins consists of seven members who invest out of both the early-stage and growth funds (21m19s).
• Kleiner Perkins is characterized as a boutique due to its small team size and focus on the craft of venture capital, with the belief that the business does not scale through people (21m29s).
• Despite being a boutique, Kleiner Perkins has the scale of capital, with half of the growth fund dollars invested in the best companies from the early-stage funds (21m44s).
• This approach allows Kleiner Perkins to double down on successful companies, such as Rippling, Glean, and Figma, without requiring a large team (22m2s).

Lessons in Reserves Management & Capital Concentration (22m7s)

• Reserves management is a challenging task, and it's essential to get good at it, with a general rule of thumb being to invest more than half of the total investment in the first check, and reserving the rest for subsequent rounds (22m20s).
• Typically, the investment is split into 60% for the initial round and 40% for subsequent rounds, with the ability to move dollars around as needed, such as when a company gets acquired or shuts down (22m59s).
• Communicating reserves management to founders can be done through signaling, such as participating in follow-on rounds, even if it's a small amount, to show support and avoid being wiped out due to pay-to-play provisions (23m21s).
• There are different scenarios for reserves management, including pay-to-play, where not participating in a round can result in being wiped out, and having to rebalance reserves when a company's valuation changes (23m42s).
• The importance of believing in the people, such as founders Aaron Levie and Dylan Smith, and being willing to take risks, even in a down market, is crucial for successful investments (25m3s).
• Market dislocations, such as the global financial crisis, can create opportunities for investments, but also require rebalancing reserves to support existing companies, such as Box, which required three bridge rounds in 2008 (24m25s).
• Founder dilution is a concern, and investors should be aware of the impact of bridge rounds and subsequent investments on the founder's equity stake, as seen in the case of Aaron Levie, who has spoken publicly about the issue (26m2s).
• Investors and board members have a responsibility to ensure that founders receive fair compensation and re-up their equity stake when it dips below a certain threshold (26m45s).

Why Do Breakout Companies Plateau? (26m53s)

• Breakout companies often plateau because they stop innovating or fail to disrupt themselves, leading to a loss of revenue as new companies disrupt the market and take their place (27m3s).
• Big companies tend to focus on protecting their turf rather than innovating, which can make them vulnerable to disruption (27m11s).
• The right time to sell a company is often when it reaches a "local maximum" in terms of perceived value by the market, where the company is seen as a leader but there are questions about its standalone potential (27m52s).
• This can be a good time to sell to a strategic acquirer who can do more with the company, but this doesn't happen as often as it used to (28m5s).
• One successful example of selling at the right time is the acquisition of Yammer by Microsoft for $1.2 billion, which allowed the investors to avoid conflicts of interest and invest in Slack later on (28m30s).
• When a company is acquired, it can be easier for investors to distribute the cash and move on, rather than having to decide whether to hold or distribute stock (29m13s).
• Even when companies go public, it's often a good idea to distribute stock to limited partners (LPs) to return value to them (29m23s).
• In general, LPs are happy when they receive a 10x return on investment, but they often wish they had held on to generate more multiples (29m47s).
• As a general partner (GP), it's possible to hold on to stock forever, as John Doerr has done with Google stock over the last 25 years (30m10s).

What’s Mamoon’s Best Performing Investment (30m19s)

• The best performing investment on a pure multiples basis was Slack, with a multiple of around 250, considering dilution over time and a valuation of around $27 billion (30m19s).
• Another successful investment was Figma, with an initial investment at a post-money valuation of around $100 million, and Rippling, which was also invested in at a post-money valuation of around $250 million (30m36s).
• A notable investment was made in a company that was later sold to SE for $700 million, resulting in a 70x multiple, but the internal rate of return (IRR) over 15 years was around 15% due to holding onto the investment for an extended period (30m55s).
• The M&A market is currently slow, with big companies being gun-shy about making acquisitions due to regulatory and legal concerns, making liquidity challenging (32m9s).
• There is a next tier of companies that may be interested in buying, such as Adobe, which acquired Figma, and other high-priced companies that are acquiring smaller companies in stock deals (32m46s).
• The regulatory environment and the people running organizations are not the primary reasons for the slow M&A market, and changes in leadership, such as the replacement of Lena Khan, are unlikely to significantly impact the market (33m47s).
• The IPO markets are also challenging, and the CMA (Competition and Markets Authority) in the UK has been involved in regulating acquisitions (34m4s).
• The current IPO market is slow, with many waiting for the election to pass before going public, but next year is expected to be a good year for IPOs, potentially sparked by a successful listing from a major company like Stripe or Starlink (34m24s).
• The $100 million funding round for Figma was notable, as it occurred before the company had significant revenue or scaling, with investors like Greylock and Index already on board (35m0s).
• The key factor in Figma's success was its product, which took time to build but eventually became a powerful design tool that could work in multiplayer mode inside a browser, with designers using it almost every workday (35m39s).
• Early metrics for Figma showed strong usage, with designers using the product 15-18 days out of a month, indicating a successful product that was just starting to scale (36m20s).
• The initial assumptions about Figma's potential were conservative, with some investors thinking its total addressable market (TAM) would be similar to that of Envision or Sketch, but the company's actual growth potential was much higher (37m35s).
• The growth of Figma was driven not only by designers but also by adjacent seats, such as marketers and engineers, which expanded the company's TAM and potential for growth (37m23s).
• The investment in Figma was successful due to a combination of factors, including the strength of the product, the vision of its founders, and the preparedness of the investment team (37m54s).
• Invision was larger than Sketch for a considerable period, and both Invision and Sketch were significantly bigger than Figma at one point (38m5s).
• Many investors view competitive markets as desirable because they indicate the presence of Enterprise Value (38m19s).
• However, an alternative mindset is to prefer products that create new markets, as they get to define the playing field and often emerge victorious (38m21s).
• Examples of companies that created new markets include Slack, which created a market for collaborative communication, and Figma, which created a market for collaborative design software (38m31s).
• Glean is another example of a company that created a new market, specifically for Enterprise search (38m41s).
• Companies that create new markets often have an advantage, as they get to establish the rules and ultimately win in that market (38m48s).

What’s Mammon’s Founder Type? (38m55s)

• There are two types of founders that are preferred: the first-time founder who is hyper-obsessed with building a product in a new market, and the repeat founder who has had a successful outcome and is doing it again (39m27s).
• Examples of the first-time founder type include Owen from Intercom, Dylan from Figma, and Matti from Front, who are young, product-oriented, and building in a newish market (39m51s).
• Examples of the repeat founder type include Stuart Butterfield from Slack and Parker from Rippling, who have had successful outcomes and are doing it again (40m16s).
• The founder profile that is not preferred is the one that takes a top-down approach to building a company, where they look at the market landscape and decide to build a business based on that, rather than being driven by a passion for building a product (40m49s).
• A premium is paid for experience, but it's not just about the price, it's about working with people who see the world in the same way and are willing to be partners in creating a bigger pie for everyone (41m16s).
• Examples of paying a premium for experience include backing Arvind from Glean at a $35 million post-money valuation, and Ali from Al at a similar valuation, where the founders were exceptional and had a strong track record (41m42s).
• The goal is to work with people who are long-term oriented and willing to create a bigger pie for everyone, rather than being short-term oriented and focused on getting a high price (42m29s).

Should Founders Maximize Fundraising and Valuation Only? (43m14s)

• Founders are often told that their job is to raise as much money as possible at the highest price, but this approach can be damaging for the next round of funding if the company doesn't scale into it, making it harder to meet the high watermark set by the previous round (43m15s).
• At the seed or Series A stage, it's more important to surround yourself with the right people rather than focusing solely on raising money (43m21s).
• A CEO's job is to ensure the company never runs out of money, but if the company has a significant amount of capital, such as $300 million, it may not be necessary to raise more funds (43m45s).
• Great founders can overcome bad markets, but it's much harder when the structure of the industry has compressed margins and difficult customers (44m2s).
• It's unlikely to back a great founder in a bad market, as the challenges can be too great (44m17s).
• Great founders often find their way to great markets and are able to pivot or change to succeed (44m30s).
• Some successful companies, such as Uber and Slack, were not obvious home runs when they started, and their success was not solely based on revenue multiples (44m41s).
• Slack's early success was based on usage patterns, with 10,000 users, a third of whom used the product every day for multiple hours (45m16s).
• Approaching early-stage companies with a revenue multiple basis can be the wrong mindset, as it's more important to consider engagement data and the potential for scalability (45m38s).
• Looking at revenue multiples for early-stage companies with low revenue, such as $500,000, can be a mistake, as it's more important to consider the potential for growth and scalability (46m5s).

What VCs Do Today That They Shouldn’t (46m10s)

• Venture capital firms (VCs) often operate within an "Echo chamber" where everyone has access to the same information, which is no longer a source of arbitrage (46m14s).
• Many big firms claim to have built data platforms that provide them with proprietary information, but this is mostly not true, and using a data-oriented approach to investing in startups has often failed (46m38s).
• At the end of the day, investing in startups is about the founders and their willingness to work with a particular VC firm (46m52s).
• LPs (Limited Partners) need to understand that VCs in the valley likely see the same deals, but what matters is being aspirational capital to the best founders in the world (47m9s).
• Founders often have multiple term sheets to choose from, but they will choose the VC firm based on its reputation, body of work, and what it can offer them (47m27s).
• A VC firm's reputation is built on its past work, what the firm represents, and what people who have worked with the firm before have to say about their experience (47m46s).

Thoughts on Voting Structures in Decision-Making (47m53s)

• Venture firms often have complex voting structures, but in some cases, there is no voting structure at all, allowing any partner to write a check, as seen in a firm with only four partners (47m55s).
• This approach is based on the idea that the best deals are often non-consensus, and every partner should be able to put themselves on the line with their conviction (48m27s).
• Partners should be allowed to lead with their conviction, and the team should discuss and test each other's conviction without a formal vote (48m40s).
• The discussion process involves seeing each other's body language and reactions to questions, which helps to build conviction and test each other's excitement about a potential deal (48m43s).
• If a partner disagrees with a deal, they can express their concerns, but no one has ever vetoed a deal in this particular firm (49m21s).
• Deals often get crushed when good questions are asked about the market, competition, or risk-reward ratio, which can make a partner's excitement wane (49m39s).
• Outcome scenario planning is not commonly used, as it is seen as "false precision" and can feel like busy work (50m14s).
• The team has different picking styles, with Josh being a precise picker, and Ilia and another partner often being on the same page, which can be both an advantage and a disadvantage (50m38s).

Mamoon’s The Most Contentious Deal (51m21s)

• The most contentious deal was Figma, which was a contentious investment due to its valuation of around $110-115 million at the time, with questions surrounding the investment, as the company had been around for five years and wasn't an obvious choice (51m21s).
• The investment was part of a fund that was deployed quickly, within 12-18 months, which was a fast deployment pace, especially considering the team had just joined and didn't have board seats (51m54s).
• The fast deployment was partly due to the team's drive and the fact that they had a fund ready to deploy, with a focus on series A investments and real ownership (52m48s).
• The fund's performance is expected to be amazing, despite the fast deployment, which could have led to missing out on a high valuation environment if deployed over a longer period (53m3s).
• The counterpoint to fast deployment is that it can lead to missing out on opportunities, but the team's approach has worked well for them, with a focus on series A investments (53m10s).
• The ability to move between stages or firms is not common among investors, and it's hard to have the neuroplasticity to think about different stages and types of investments (53m33s).
• The team has a structured approach to investments, with a focus on early-stage and venture investing, and some team members also do select investing, but they don't expect everyone to have the neuroplasticity to switch between different types of investments (54m10s).
• The team has a pipeline meeting and an investment team meeting for both early-stage and venture investing, but they don't burden everyone with the need to switch between different types of investments (54m6s).

Mamoon’s Biggest Loss (54m24s)

• A significant loss was experienced with the company Tally, resulting in a capital loss of around $30 million, which is multiples of the largest loss prior to that (54m48s).
• The company Tally was founded by Jason Brown, who had been known for 15 years, and had been featured on the show (54m43s).
• The loss was attributed to the challenges of consumer lending businesses, which became even more difficult when interest rates rose from 0 to 5% (56m0s).
• Reserves were done, but it was not enough to prevent the loss, and subsequent funding rounds from other investors, including Angela at Andrew and Horwitz, did not help (55m25s).
• The experience has led to a realization that lending businesses are hard, and consumer lending is particularly challenging (55m22s).
• As an investor, the goal is to stay open-minded and not let past experiences weigh too heavily, in order to be able to identify the next big opportunity (56m16s).
• The biggest area for improvement as an investor is knowing when to cut losses and stop believing in a company, as there is a tendency to believe for too long (56m48s).
• The importance of VC value add is acknowledged, but it is also recognized that the wrong advice can lead to destruction of value, and the goal is to provide the right help to supercharge companies (57m17s).
• The worst board is one that is not run effectively by the CEO, and this can lead to a range of negative outcomes (57m27s).

How Do the Best CEOs Run a Board? (57m33s)

• Effective CEOs run a board by starting with a high-level overview of the company's performance, followed by in-depth discussions led by capable leaders, and a mix of cheerleading and hard questions (57m33s).
• Successful board meetings focus on one or two key issues that can significantly impact the company's trajectory, rather than discussing multiple topics superficially (57m52s).
• The ideal board meeting structure involves diving deep into a limited number of critical topics, as this allows for more meaningful discussions and decision-making (57m55s).
• Some CEOs may prefer capital-efficient businesses, but can still support and invest in companies with different profiles, such as Uber and DoorDash (58m28s).
• The goal for a company like Finder is not to become a capital accumulator, but rather to maintain its current position and focus on early-stage venture investments (58m46s).
• Early-stage venture investing is considered a beautiful asset class, especially when following the power law and focusing on the few companies that truly matter (58m55s).
• A successful investment strategy involves identifying and investing in a select few companies that have the potential to make a significant impact, and then concentrating efforts on those investments (59m4s).

Quick-Fire Round (59m11s)

• Most people around believe that venture is an easy job and glamorous, but in reality, it's hard, as seen from the reactions of operator friends who find it challenging compared to what they expected (59m14s).
• The most memorable first founder meeting was with Aaron Levie, who brought along Karen Page, thinking he needed to bring an executive, and was hyper nervous, but showed a deep understanding of the problem of cloud storage in 2007 (59m36s).
• Public market revenue multiples need to reflate for venture to be a sustainable business, but getting back to normal historical levels would be good enough (1h0m19s).
• The Rule of 40, which emphasizes growth and profitability, is important, and companies like Box, with $1 billion in revenue and a $4 billion market cap, are examples of this (1h0m20s).
• The venture investor most respected and learned from outside of Kleiner is Matt Cohler from Benchmark (1h0m45s).
• If given the chance to be CEO of any company, the choice would be OpenAI, to see what's going on and how far AGI has progressed (1h0m55s).
• Geopolitics and polarization are the biggest concerns, with too much of an "us versus them" mentality in countries and between countries (1h1m7s).
• When it comes to investing in seed, series A, and growth firms purely on multiples basis, the choice would be seed firms, specifically Gili Raanan's Cyberstarts in Israel (1h1m53s).
• What is known now that wasn't known when joining USVP 19 years ago is that venture is a grind, but it's still loved, especially with the current AI super cycle (1h2m12s).
• A question that should be asked more often is how faith impacts the way one works, as it can inform how to treat people, the earth, and show up in meetings with humility, empathy, and care (1h2m38s).