The Power Law

A reminder for new readers. That Was The Week collects the best writing on critical issues in tech, startups, and venture capital. I selected the articles because they are of interest. The selections often include things I disagree with. The articles are only snippets. Click on the headline to go to the original. I express my point of view in the editorial and the weekly video below.

This Week’s Video and Podcast: There is no video this week as Andrew is traveling.

Content this week from @kteare, @douglassshapiro, @mercury, @circle, @uentrepreneurs, @scottehartley, @marcepntoja, @credistick, @Angels_Cube, @I_Am_NickBloom, @abialecki, @benedictevans, @KimberlyMutandi, @amir, @aaronpholmes, @rhodgkinson, @signalrank


Editorial: The Power Law

Essays of the Week

Power Laws Have More Power Than You Think

Raise Less, Build More

Are There Too Many VCs: Why 98% Are Average Or Mediocre?

Zero-Billion Dollar Markets

The Pocket Guide of Essential YC Advice

Are YC Valuations Really Too High?

Family Offices and Venture Capital – Market Pulse

Nicholas Bloom predicts a working-from-home Nike swoosh

Survey: Remote Work Isn’t Going Away — and Executives Know It

Video of the Week

Klaviyo CEO Andrew Bialecki

AI of the Week

Generative AI and intellectual property

Free AI tools are killing South Africa’s web designer job market

OpenAI Passes $1 Billion Revenue Pace as Big Companies Boost AI Spending

News Of the Week

15.2% of Rounds Completed in Q2/23 were down rounds

Nvidia vs. AMD vs. Intel: Comparing AI Chip Sales

Startup of the Week


X of the Week

Who Are The Best Seed Investors Today?

Editorial: The Power Law

Most data is evenly distributed. That is to say, even though there are points along a range, the number of individual nodes at any segment tends to be the same as any other segment.

But in a lot of markets, that is not true. Venture Capital is one of those. The number of investments that return a fund, or better, is infinitesimally small compared to the set. This creates a power law curve similar to the one in this week’s image (MidJourney plus my pencil).

We can see the law in the data. Below is the SignalRank data for all series AA rounds since 2012. The chart shows the value growth and how many of the 38,000 companies funded achieved that growth.

The plot illustrates the heavy reliance of venture capital on outliers. Here’s how to interpret the graph:

The blue line represents all the companies in the dataset, ranked by their Multiple on Invested Capital (MOIC).

The red points represent the top 5% of companies, which are the outliers in this dataset. These companies have significantly higher MOIC values compared to the rest.

The X-axis uses simple integers to represent the rank of each company based on its MOIC value.

The Y-axis shows the actual MOIC values, and the red points are annotated with their MOIC values for further emphasis.

A small number of companies (those in the red) contribute significantly to the overall MOIC, reinforcing the venture capital industry’s reliance on outliers for substantial returns.

The Power Law is best understood as counting the percentage of high returners in a portfolio. But there is also a power law within the power law.

The best funds take an uneven share of the power law outliers.

For example, 10% of Series B Rounds result in a unicorn (between 2012-2023). But really good investors have much better than 10%. SignalRank’s AI-based B investor achieves over 25%.

So there are a small number of successful companies and an even smaller number of successful funds and fund managers.

Allocating capital to venture is, therefore, a hazardous exercise. The odds of losing capital are very high.

This week there are a lot of really good essays covering this. Dan Gray in Crunchbase News says:

A well understood concept in venture is that the majority of any fund’s returns will be driven by a handful of companies. Maybe 1% of investments will be a 100x return, 5% will be 10x returns, and 50% will lose money. Identifying those outliers is the whole ballgame, so even a marginal improvement to selection can have a huge influence on fund performance. This factor has driven significant investment into VC “platform teams” since 2010.

With that in mind, consider that roughly 4.5% of Y Combinator startups have achieved “unicorn” status since 2010, according to analysis by Inside. That’s head and shoulders above similar accelerators, which makes it such an appealing target for investors.

Scott Hartley explains that outliers create new markets that others fail to see until the startup owns them. Zero to Billion Dollar Market is an excellent read. here is a snippet:

So be the one just crazy enough to invent the Zero-Billion Dollar Market you see, and then just sane enough to dominate the category you create for yourself.

Really good seed investors excel at finding these outliers. My colleague Rob Hodgkinson’s X posts this week make “X of the Week” because they use the data to demonstrate the reality of the power law. He calls out the top ten highest conversions of seed to Series A, where investors made 20+ seed investments since Nov 2021.

This is important because value grows in Venture Capital when an early-stage investment gains value due to a subsequent funding and valuation event.

There are two pieces talking about Y Combinator’s numbers and approach. According to an analysis by Inside, 4.5% of YCs investments become unicorns historically. That compares to only 0.099% of all seed rounds in the past ten years (over 80,000 of them).

So the power law is important because it demonstrates that finding good outcomes is super hard. The advent of data intelligence in channeling capital into de-risked investments becomes very important in that context.

No video this week as Andrew is away.

A call out to Nick Bloom, who is mentioned in 2 essays covering hybrid work. He and I are both advisors to Dan Bladen’s company, Kadence, building the enterprise operating system for hybrid teams. Congrats on the acknowledgement Nick.

Essays of the Week

Power Laws Have More Power Than You Think

Why hits will persist in a world of infinite content


Source: Hurca!/

10-second summary:

In an apparent contradiction, the internet both fragments and concentrates attention.

The reason for the former is intuitive: more stuff, less attention per unit of stuff. The reason for the latter is not. It happens because networks are subject to powerful positive feedback loops. On a network, people’s choices are influenced by others’ decisions, amplifying “hits.”

There are two underlying mechanisms: information cascades (when people treat others’ choices as signals of quality) and reputational cascades (when people conform with the group decision). As choice has exploded on the Internet and it has become easier to both observe others’ choices and share your own, these mechanisms have become more powerful.

Consumers also rely heavily on recommendation algorithms to make choices, intentionally and unintentionally. Depending on how they’re constructed, these systems can either boost or dampen the social signals arising from the network.

The result is that the distribution of consumption in almost all media persistently, and in some cases increasingly, looks like a power law: a few massive hits and a very, very (very) long tail. I provide a framework for thinking about the “extremeness” of the distribution and show a few examples: box office, Netflix original series, Spotify streams, and Patreon patrons.

There are a number of important implications for media companies. The good news is that there will likely always be big hits, even in a world of practically infinite content. The bad news is just about everything else: the lucrative middle is being hollowed out; the randomness—and therefore risk —in producing hits is climbing; the tail is become more competitive for hits; more economic rent will likely shift to talent; content producers are increasingly at the mercy of curators’ algorithms; and paid media is being devalued.

Almost 20 years ago, Chris Anderson published The Long Tail, which accurately predicted that the Internet would fragment attention and consumption would shift into the “tail.” But Top Gun: Maverick generated over $700 million at the domestic box office, Bad Bunny had 18.5 billion streams on Spotify last year, and 142 million households reportedly watched Squid Game in its first 28 days. Why are there still hits in a fragmenting world?

Imagine hiring an AI executive assistant, who understands your tools and workspace like a seasoned pro. Welcome to Lemmy – your ideal coworker! Lemmy syncs with your daily tools like Google Docs, Slack, Notion, Meta Ads, and GitHub, and even supports web browsing.

Instead of wasting precious hours digesting long threads or documents, let Lemmy summarize them and catch you up to speed. Save time with content creation by starting your work and letting Lemmy autocomplete it. Let Lemmy’s analytical skills take over managing your Facebook ad spends or conducting diligent SWOT analysis.

I obsessively think about the future of the media business—I’ve been in and around it for almost 30 years, including about 15 years as a Wall Street analyst and 12 years at Time Warner, where I headed investor relations and held a variety of strategy roles. I recently published an essay about what I call “infinite TV,” in which I made the case that over the next decade, video will follow the path of text, photography, and music, and the quality distinction between “professionally-produced” content and “independent/creator/user-generated” content will increasingly blur. The result will be a practically infinite amount of quality video content. Will there still be hits then, or only personalized niches?

Have you ever wondered why so many blockbuster movies are about superheroes? Is Hollywood lazy, or are consumers’ tastes becoming dumber and more homogenized? Or neither?

Why does something go viral, anyway?

Do content recommendations push you to the most popular shows, movies, and songs, or are they tailored just for you? Or do they have a different agenda?

Will web3 really be the savior of small creators?

When Billie Eilish, Lil Nas X, Mr. Beast, or PewDiePie emerged from obscurity, was it inevitable that their talent would be recognized—or just luck?

Are the top-rated reviews on Amazon or answers on Quora the most helpful?

All of these questions are about the distribution of popularity. The same phenomenon underlies the answers: networks.

In this piece, I explain why power law-like distributions—meaning a few massive hits and a vast number of misses—are an inherent feature of networks; describe how recommendation systems can either dampen or reinforce social signals; show some examples of the persistence of power law-like distributions in media across movies, TV, music, and the creator economy; and discuss why all this matters.

The long tail was half-right

The idea that the internet would cause media fragmentation is almost as old as the modern internet itself. (Or maybe older. The line often misattributed to Andy Warhol that “in the future, everyone will be world-famous for 15 minutes” was a pre-internet prediction of fragmentation.) In 1999, Qwest Communications produced an ad featuring a motel with “every movie ever made in every language” (Figure 1). The Long Tail, published in 2004, argued that because the internet dramatically lowered the cost to store and transport information goods, it would result in practically unlimited shelf space. Faced with far more choice, consumers would shift most of their consumption to the “tail,” heralding the end of mass culture and waning importance of hits. If anything, Anderson underestimated the size of the tail because he didn’t anticipate social media. The tail is not Icelandic synth pop, as it turns out, but an endless amount of user-generated content

Figure 1. Qwest envisioned media fragmentation 25 years ago

Source: Qwest Communications print advertisement, 1999.

That the internet would yield more choice and, therefore, more fragmentation was intuitive then and is indisputable now. But it only tells half the story. Though it seems contradictory, the internet both fragments and concentrates attention

Understanding those dynamics matters. The contention that there are still hits may seem uncontroversial and certainly feels right intuitively. We know that when Beyonce or Taylor Swift releases an album, or the next season of Stranger Things or Game of Thrones drops, the collective attention of popular culture, much like the eye of Sauron, will be trained on it—at least until the next thing comes along. But understanding why there are still hits provides insight into whether this will persist as the supply of content keeps growing faster than demand.

The internet concentrates attention because it connects everyone in a big network. And networks are subject to powerful feedback loops. Since consumers increasingly both discover and consume content through information networks, their decisions are increasingly influenced by other people’s decisions. These feedback loops amplify the popularity of a small number of choices—hits.

The net result of these opposing forces—fragmentation and concentration—is that media consumption, and culture more broadly, is persistently, and in some cases, increasingly, observing power law-like distributions. Few TV shows, movies, songs, books, video games, journal articles, newsletters, short-form videos, and tweets will be wildly popular, while the vast majority will be hardly consumed at all.

What is a power law?

One of the first statistical concepts taught in school—right after mean, median and mode—is the “bell curve,” aka the normal or Gaussian distribution. The intuition behind a normal distribution is that if you have enough random independentobservations, most observations will be relatively close to the average (or mean) and equally distributed on either side of it. Many independent natural phenomena approximate this distribution, especially when the extremes are bounded, like height, weight, test scores, or rolling two six-sided dice.

Figure 2. Normal and power law distributions

Power law distributions, by contrast, look different. A power law simply means that the dependent variable is a “power” of the independent variable. For instance, the volume of a cube is a “power” of the length of the sides, because volume increases three units for each one unit in length. Generally, they can be expressed as:

In a power law probability distribution, the exponent is negative, which results in a downward sloping curve (as illustrated crudely in Figure 2). As shown, power law distributions are characterized by a large number of very small observations and a small number of very large observations.

For our purposes, the main point of this comparison is shown in the graph furthest to the right in Figure 2, which superimposes a power law distribution over a normal distribution: the likelihood of both extremely small and extremely large observations is much greater in the former than the latter.

…. more

Raise Less, Build More

August 20, 2023

by trohan

The conventional venture capital funding path – from raising an institutional Seed, Series A, B, C, D, E, etc, all the way to exit via IPO – has long been treated as gospel. Its verses are most heavily preached by VC board members, whose business model it also supports. 

But there is an influential tide of founders on the rise that is opting out of this path and quietly plotting a new one that leads to building generational companies.

It’s a hybrid path, combining the growth of targeted venture funding with the durability found in bootstrapping (i.e. profitability). It’s a path with less venture capital and more self-reliance. 

And it’s the direct result of founders emerging from a tumultuous period of feast (with 5x more venture capital offered to startups over the past decade) and a brief flirtation with famine from the recent pullback that has left some venture-dependent companies in starvation mode. 

For many founders, a steady reliance on venture capital, as it is heavily prescribed today, is often seen as unhealthy, if not risky.  

Increasingly, these founders are seeking freedom from the risk and control of the perpetual pursuit of venture capital. Instead, they’re ready to reroute their time, efforts, and attention to building enduring companies on their own terms. 

These founders are choosing to raise less and build more.

Foie gras” venture capital 

The clearest trend in the venture industry over the past decade is VCs offering startups more money. 

To gain visibility into this trend, we can examine the ratio of VC capital raised to new startups in any given year. An imperfect but close proxy for viable new startups is the annual number of Seed deals (inclusive of Angel, Pre-Seed, and Seed deals). When you take total VC dollars raised, divided by the number of new companies, you’ll see the average startup today has 5x more VC capital available than its counterpart did in 2013:

The primary driver of this trend is VCs increasing their fund sizes, particularly larger firms. Nearly every single major fund has implemented this change. No other strategy has been so universally adopted.

As a result, every single type of round, from Seed to Series C on, has also increased in size. Here are some averages over time:

While 2023 will likely prove to be a down year for total VC capital raised, both the average and median fund size will remain up. The average fund size went from $336M in 2020 to $386M in 2022 to $538M in 2023 (PitchBook). The trend has become the new baseline.

More isn’t always better 

Why is this trend so pervasive? 

Do today’s startups need more money?  

Well, no. 

The main driver of a startup’s burn is salaries. Salaries over time have increased, but not by 5x (the rate VC capital has increased). Even the new increased infastructue costs with AI will ultimately come down as companies scale and the industry matures. None of this justifies the rate and breadth of this capital increase from VCs. 

Contrary to popular myth, statistically venture backed startups aren’t staying “private longer,” either. Between 2011 and 2021, the median time from venture funding to IPO exit in the US has fluctuated between 5 and 7 years, with no visible increase in time (Statistica 2023).  Granted, SPACs may have impacted these numbers, but it’s nonetheless interesting that the data does not align to the myth.   

Do bigger funds perform better?

Again, no. 

Smaller funds consistently outperform large funds (on a multiple basis). More on this, below. 

Does giving a startup more capital make it reliably perform better? 

Another no. There is no conclusive data to support this. There is no empirical study that I know of that suggests that raising more capital increases a startup’s odds of success. 

Looking at case studies, for every Stripe ($8.7B raised), there are countless startups who raised too much and did not make it. To avoid speaking ill of the dead or the fatally wounded, I’ll withhold names (the headlines will tell this story).

The lack of conclusive evidence seems to have had little impact on the VC bias that startups – as a general rule – should raise more VC capital. 

Consider some of the more persuasive funding conventions prescribed by VCs:

If you can, raise more capital as opposed to less capital 

For founders, the case for more capital is easy to hear – hire faster, be offensive, save for a rainy day. But too much capital, especially too early, can depress urgency and innovation, all while contributing to unnecessarily high headcount and burn. These can be company killers. For every Figma, who raised a large but not unreasonable Seed round (for its time), there is a Seed company that raised that mega round (read $10M), which may do them more harm than good. Once you’ve raised for multiple years of runway, including hiring well and adding a buffer budget, what is any extra capital really doing? Who is that mega round really serving? 

Raise at every round, or when you have explosive growth 

For many founders, taking on regular capital can be extremely valuable, as can taking on capital for hyper growth. But for startups that are already profitable and growing, raising can be a costly distraction. It can swallow founder time during company-making moments, resulting in potential loss of discipline, premature hiring, and unnecessary spending. It also can alert competition, hinder hiring (via ‘less attractive’ stock options), impact dilution, and potentially even affect board composition. Companies like Vanta chose to delay Series A to focus on building instead of raising, with others raising their huge Series As at the ‘standard’ time, because they ‘should,’ before fading into history. You have to wonder – how many of those momentum Series A companies would still be around if they just chose to build (profitably) instead of raise? 

Prioritize growth over profitability to achieve outlier success

For many founders, opting for growth over profitability has clear advantages: you can hire more, build more, market more; in short you can be generally more offensive. But for founders with the goal of creating a sustainable and durable business, prioritizing profitability growth over growth-at-all costs makes more sense. During a time when VCs were fawning over him, Ivan at Notion famously told investors that he’d “prefer to raise from his customers” as opposed to taking outside capital. Countless high-potential startups would still be around today if they had the same discipline.  

No doubt more venture capital can help, but it can also certainly hurt. 

VCs sell a product to founders. That product has ballooned over time in ways so that it can now potentially do harm. Like any other consumers, it falls to the founders of the companies to understand what they’re buying, and why it’s being sold in a particular way.

Are There Too Many VCs: Why 98% Are Average Or Mediocre?

Dileep Rao

Finance Secrets of Billion-Dollar Entre’s (

Aug 31, 2023,11:40am EDT

The prevailing wisdom is that there is a shortage of venture capital (VC). Is this “wisdom” true? It depends on how you measure the shortage. If the “shortage” is measured based on entrepreneurs seeking capital, then yes. There is a shortage.

Entrepreneurial hopes always exceed the capital available. Entrepreneurs want growth. Growth requires skills or capital or both. Instead of skills, most entrepreneurs seek capital. Specifically, they seek early VC, which, unfortunately, is the wrong strategy:

· Early VC is scarce and has been used by only 6% of billion-dollar entrepreneurs. Entrepreneurs should be using the vast array of potential financing sources that are more readily available.

Is there a shortage of capital?

As long as entrepreneurs focus on capital over skills, there will be a shortage of capital. Is there also a shortage based on the productivity of capital, i.e., based on financial returns?

#1. Only about 2% of VCs earn 95% of VC profits. 98% are average or mediocre.

20 VCs are said to earn about 95% of VC profits. Since the number of VC funds in the U.S. is estimated at about 1,000, this suggests that about 2% do very well and 98% are average or mediocre – they fail to live up to the lofty reputations of financial genius that VCs have self-promoted. Interestingly, SPAC promoter Chamath Palihapitiya notes that only about 10% of VCs make money. The rest are said to be money losers with a lot of their profits being phantom profits that their investors really do not see.

#2. VCs need homeruns if they want to succeed. VCs finance very few home runs.

Even the top VCs fail on about 80% – 90% if their ventures, according to one of the most successful VCs in the U.S. The top 2% earn high returns because they finance home runs. VCs need home runs to do well, and most VCs stink because they do not fund home runs. If there were a real shortage, wouldn’t more VCs finance home runs?

#3. VCs mainly succeed in Silicon Valley. VCs outside Silicon Valley are not as productive.

Most data shows that the Top 20 VCs are in Silicon Valley. This suggests that VC outside Silicon Valley do not do well. Silicon Valley has developed an ecosystem that churns out unicorns. The others have many experts and governments wasting money hoping to emulate this ecosystem.

#4. Entrepreneurs need to get to Aha! VCs do not know how you can get to Aha!

VCs finance after Aha, i.e., after potential is evident. Before Aha, many can point out all the flaws – but identifying potential winners is a guess – even Steve Jobs and Google were rejected by more than 10 of Silicon Valley’s finest VCs. You must get to Aha on your own – with your strategy and your skills to beat your competitors and create venture value. The problem is exacerbated by entrepreneurs who follow the VC method of focusing on the opportunity, entry strategy, and VC – rather than on the Unicorn-Entrepreneur method of finance-smart skills and bootstrapping growth strategies.

#5. Business schools focus on the VC-method, which helps about 20/100,000 ventures.

Can business schools be more productive? Most business schools teach opportunity analysis, strategy development, and VC financing. As noted above, this VC-method helps few entrepreneurs and few VCs, mainly in Silicon Valley.

#6. VC analysis seems to be deteriorating. Is too much VC creating FOMO?

VC Brian Grossman invested $96 million in Theranos and lost a lot of it. His due diligence is said to have raised a number of questions. But he still went with his instinct, due to FOMO (fear of missing out). Are these VCs sacrificing their analysis due to desperation – due to too much VC chasing hype?

#7. Are VCs sticking to the knitting?

VC has succeeded in emerging industries, such as Uber, or in high-margin ventures, such as Google. Masayoshi Son has lost $32 billion in VC. Without the circumstances of an emerging economy (China) or an emerging industry (telecom), even a great entrepreneur like Son has struggled. Is that because of too many VCs chasing too few great deals?

#8. VC returns and funding fluctuate with stock market exuberance. Is this skill or luck?

The Top 2% seem to have the talent to build unicorns at all times. The others seem to need Wall Street exuberance.

Zero-Billion Dollar Markets

Scott Hartley

Co-Founder & Managing Partner at Everywhere Ventures

September 1, 2023

In HBO’s Silicon Valley the obnoxious investor Russ Hanneman talks about the “Three Comma Club” in reference to the number of commas in a billion. In a recent Category Design workshop we held with Peter Goldie and Deborah Kattler Kupetz for our portfolio companies, we were introduced to the concept of the zero comma club, or the “Zero-Billion Dollar Market,” a concept they’d explored over past years that was popularized by Steve Vassallo of Foundation Capital, as well as endorsed by leading investors at Floodgate and Sequoia.

The notion of a Zero-Billion Dollar Market is that it’s something that you can see as an entrepreneur, but no one else believes. It’s counterintuitive, and it’s true. As we’ve talked about before, being both consensus and correct is wonderful in many parts of life, but not in entrepreneurship. If you’re trying to create something new and valuable then it’s generally down the less traveled path, against the grain, and certainly not wildly obvious to others. Value is located at the intersection of being both non-consensus, or contrarian, and also being correct.

My friend Tommy Dyer recently alerted me to the fact that Don Quixote has been taught at Stanford Graduate School of Business (GSB), with particular focus on his “Lessons for Leadership.” Among other lessons, Quixote creates his own reality through action. Though often considered mad, particularly as he begins to see the windmills around himself as monsters, it’s perhaps all only in his imagination, and as Tommy delves into in his recent writings, perhaps it’s even him living in his own “reality distortion field,” to use the Steve Jobs turn of phrase. Indeed entrepreneurs must be Quixotic, or exceedingly idealistic, even impractical or “crazy” to invent new categories. “Here’s to the crazy ones. The misfits. The Rebels…” This is the edge between vision and reality where contrarianism dwells.

This is the land of Zero-Billion Dollar Markets, where some people will call you crazy for “tilting at windmills,” or imaginary monsters. Andy Grove, former CEO of Intel, wrote the 1999 book entitled “Only the Paranoid Survive.” Only those crazy enough to be out there in the arena, or on the journey, and also paranoid enough to tilt at windmills, survive long enough to create their own reality and new market.

So what’s better than a hundred billion dollar TAM? A Zero-Billion Dollar Market, a category that hasn’t been invented yet.

This is the one you’re inventing from the ground up, like Apple debuting the iPad, or Chrysler creating a new category of car in the Minivan. Five Hour Energy decided not to play the game of battling for the beverage aisle and instead decided to sell at point of sale, with a straight-forward name, and new form factor. Other examples abound, but one of particular interest over the last decade was the emergence of Qualtrics as the clear winner over Survey Monkey. Whereas at one point these “online survey” tools were indistinguishable, and in many ways, Qualtrics, sitting in Utah rather than Silicon Valley, was at the disadvantage, until Qualtrics decided to invest in defining their own category. In order to do this they looked at “the gap” between what CEOs or employees thought their company or product did, and what their customers did.

This “gap” between perception and reality has to be closed by better understanding your customers. Of course these could be called surveys, but by reframing the conversation around a market opportunity that hadn’t been tapped, and inventing the category of “Customer Experience” or CX, they became the creators of a Zero Billion Dollar Market. They were no longer playing the game of competition. They were the monopoly player in their own new invented world of CX, and through the course of an acquisition and later IPO spin out, they grew to nearly 15x the market cap of Survey Monkey over only a few years. This clear eruption in growth came from one thing: Narrative. Category creation isn’t simple new words; it’s internalizing this Quixotic spirit of inventing the world around you through action, even when others only see Windmills and think you’re crazy. This isn’t to say you’ll be right, but that you’re at least being non-consensus.

A few key steps to finding your Zero-Billion Dollar Market:

Where is there a gap in society between reality and perception? Where has the hype cycle overshot, or what’s the underbelly of a major trend?

What unique skills or lenses predispose you to see two markets converging where others do not think there is overlap or opportunity?

Play what I call the “Animal Flip Book Game” of mixing and matching three things, namely industries you understand deeply, new technologies coming of age, and business models (SaaS, marketplace, etc). Do this to come up with non-sensical combinations that might lead to identifying Zero-Billion Dollar categories everyone else thinks are too crazy to chase. There’s nothing like an Elephant head, Giraffe body, Duck foot company that will weird people out to the point that you just might be contrarian, and right.

At Everywhere Ventures, one of our portfolio companies, Radicle Science, did the same. They observed what they call a “proof gap” between what customers believe supplements like melatonin do, and what they actually scientifically do. Radicle Science now offers clinical trials, or “Proof as a Service,” for companies in the less regulated supplements industry to prove to consumers and retailers what they do. Radicle Science created the market of democratized clinical trials. My friend Catie Cuan did in creating the field of Choreorobotics, or the blending of dance and robotics. She noticed a “trust gap” between man and machine that could be filled by adding graceful movement into the orchestration of robotics, particularly in medical environments where the goal of machines is to improve patient wellness, not increase anxiety. For years as a professional dancer no one understood her interest in both ballet and mechanical engineering until she completed her PhD at Stanford, and invented an entire new field of study now offered at universities such as Brown. She created an entire new category for herself, and other artists in technology, and is evidence of the value of plurality of interests, and the categories that might exist in novel overlaps.

In Joseph Heller’s Catch-22 if you asked not to fly dangerous missions you were deemed sane, and therefore had to fly. And if you were insane you were happy to fly and never thought to abstain. So the only ones who could get out of missions where those unfit to fly, but if you identified as unfit to fly you were deemed sane and therefore were required to fly. In entrepreneurship it’s not an entirely dissimilar logical loop. On some level you need to be wildly Quixotic, or “crazy” enough to try something, and make your investosr believe you are “swinging for the fences.” But at the same time you have to be measured in how you communicate this mild insanity. You have to be fit to fly, and assemble a team of passengers alongside you, knowing that only crazy ones will sign up for this decade-long mission.

So be the one just crazy enough to invent the Zero-Billion Dollar Market you see, and then just sane enough to dominate the category you create for yourself.

The Pocket Guide of Essential YC Advice

This is good advice for any startup founder.

Though YC does not explicitly say it, what’s implied as you build your company are two things:

· Build Product, Sell Product


· Control Your Destiny

Are YC Valuations Really Too High?

By Dan Gray

August 31, 2023

With each cohort that graduates from Y Combinator, the same debate emerges: How can such early-stage startups justify such high valuations?

According to YC President Garry Tan, 75% of the current summer cohort is pre-revenue and 81% are looking at raising their first external capital. Many of the founders will have entered the accelerator without much more than an idea.

Despite this, YC has developed a reputation for launching startups into the investment market at eye-watering prices. This is particularly striking in H2 2023, coming off a real downturn for startup fundraising. Investor Erik Bruckner has reported $15 million post-money caps as the most common terms among the sample of startups he’s met from this batch, at a time when the median U.S. pre-seed valuation is closer to $8.7 million.

How can YC justify promoting these terms to a more conservative venture market?

YC is not setting any valuations

While the partners at YC can make recommendations on terms, the terms of any subsequent raise are up to the founding team. Founders will attempt to raise at a price they think reflects the opportunity they offer, and — for YC startups — the market is on their side.

Secondly, startups raising out of YC are typically doing so with a capped SAFE (simple agreement for future equity) agreement, which determines the maximum price at which capital converts to equity in a future priced round.

Conflating caps and valuations is an easy mistake to make. They share many of the same characteristics, and if it is set reasonably, a cap may well end up reflecting the valuation of a startup. But they aren’t the same. Fundamentally, a valuation is a determination of value while a cap is a ceiling on price.

While it has implications on value, a cap is effectively meaningless up until a startup actually goes through a proper valuation for a priced round — and not all YC startups will manage to raise at or above that cap.

Venture is a power law game

A well understood concept in venture is that the majority of any fund’s returns will be driven by a handful of companies. Maybe 1% of investments will be a 100x return, 5% will be 10x returns, and 50% will lose money. Identifying those outliers is the whole ballgame, so even a marginal improvement to selection can have a huge influence on fund performance. This factor has driven significant investment into VC “platform teams” since 2010.

With that in mind, consider that roughly 4.5% of Y Combinator startups have achieved “unicorn” status since 2010, according to analysis by Inside. That’s head and shoulders above similar accelerators, which makes it such an appealing target for investors.

The increase in price is worth every penny if it can increase your hit rate.

The end justifies the means

Ultimately, if YC didn’t deliver high-quality opportunities for investors, they wouldn’t be able to secure investment at above-market rates.

In fact, YC’s performance even stands up outside of power law shenanigans. Analysis by Jared Heyman of Rebel Fund, a VC that exclusively targets the top 5%-10% of YC startups each year, shows that an index of all YC startups would yield an impressive annual return of 176%, net of dilution.

There has been speculation over the years that YC has lost its way, that it has grown beyond some optimal size. That criticism doesn’t yet appear to be supported by any data, and comes largely from the group which would benefit from lower entry prices: the VCs who will inevitably be queuing up to invest in the next cohort.

Family Offices and Venture Capital – Market Pulse

Alexandre Covello, Managing Partner at 109 Capital

I’ve spoken to 30 family offices over the past six weeks about their venture capital portfolio allocation.

Here are the 7 most controversial (and surprising) things I learned:

𝐕𝐞𝐧𝐭𝐮𝐫𝐞 𝐂𝐚𝐩𝐢𝐭𝐚𝐥 𝐋𝐚𝐧𝐝𝐬𝐜𝐚𝐩𝐞 𝐄𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧:
– Family offices recognise that the landscape has dramatically changed since they made their first investment in venture capital.
– These changes include the need for more diversity and inclusion, higher allocations to impact and climate change, the emergence of new players, from solo-VCs to CVCs, and the rise of alternative funding models, such as funds of funds.

𝐑𝐞𝐜𝐞𝐧𝐭 𝐌𝐚𝐫𝐤𝐞𝐭 𝐕𝐨𝐥𝐚𝐭𝐢𝐥𝐢𝐭𝐲 𝐂𝐨𝐧𝐜𝐞𝐫𝐧𝐬:
– Obviously, recent market volatility and uncertainty, as well as increased interest rates have prompted family offices to seek more stable and liquid assets, moving away from direct venture capital investments.
– They reckon that liquidating part of their most illiquid portfolio offer a path to achieve this.

𝐏𝐨𝐫𝐭𝐟𝐨𝐥𝐢𝐨 𝐎𝐩𝐭𝐢𝐦𝐢𝐬𝐚𝐭𝐢𝐨𝐧:
– Interestingly enough, family offices are revisiting their overall exposure to existing venture capital allocations.
– They are keen to optimise returns, lock in realised performance and reduce exposure to underperforming or stagnant investments.

𝐋𝐢𝐪𝐮𝐢𝐝𝐢𝐭𝐲 𝐍𝐞𝐞𝐝𝐬:
– Some family offices are (unfortunately) experiencing liquidity constraints.
– They see direct and portfolio secondary transactions as a mean to unlock capital tied up in long-term venture investments.

𝐑𝐢𝐬𝐤 𝐌𝐢𝐭𝐢𝐠𝐚𝐭𝐢𝐨𝐧:
– In an increasingly uncertain market, family offices are proactively mitigating risk by diversifying their portfolios and reducing their concentration in an asset class that they now perceive (right or wrong) as bearing more risk.

𝐏𝐨𝐫𝐭𝐟𝐨𝐥𝐢𝐨 𝐑𝐞𝐛𝐚𝐥𝐚𝐧𝐜𝐢𝐧𝐠:
– Family offices are strategically reshaping their portfolios to align with emerging sectors and technologies that offer more stable risk-adjusted returns.

𝐀𝐥𝐢𝐠𝐧𝐦𝐞𝐧𝐭 𝐰𝐢𝐭𝐡 𝐈𝐧𝐯𝐞𝐬𝐭𝐦𝐞𝐧𝐭 𝐌𝐚𝐧𝐝𝐚𝐭𝐞:
– Last but not least, family offices are reevaluating their investment mandates and finding that other investment strategies probably align better with their evolving investment criteria, their ethos and their investment horizon.

The “new way” of managing venture capital portfolios is all about strategic flexibility, diversification, and smart capital allocation.

Nicholas Bloom predicts a working-from-home Nike swoosh

Firms, employees and society will all benefit, reckons the Stanford economist

image: dan williams

Aug 29th 2023

The media are full of stories of how firms from Amazon to Zoom are dragging their employees back into the office. So is working from home (wfh) over? Was this simply a pandemic-era remote-work boom extended by tight labour markets?

No. I believe that, having stabilised, wfh will soon start growing again. Remote working is set to undergo a Nike swoosh, with an initial post-pandemic drop, followed by its current stabilisation and a future long-run surge.

Before the pandemic about 5% of full paid days were worked from home across Europe and America. Working from home was something professionals would do occasionally, perhaps to look after a sick child or let in the plumber. Now about 25% of the workforce is on a hybrid schedule, working from home typically one or two days a week. Another 8% are working a fully remote schedule. Overall, about 20% of all days are now worked from home. Looking ahead, two powerful economic forces will drive WFH up, perhaps to 30% of days worked a decade from now.

The first and most powerful of these is improving technology. In 1965 just 0.4% of days were worked from home in America. The share then doubled roughly every 15 years until 2019, driven by technological advances. These included the personal computer in the 1980s, the spread of laptops in the 1990s, the explosion of the internet in the 2000s and most recently cloud file-sharing and video calls.

These technologies made it easier to work remotely. Imagine trying to work from home without a computer or the internet. I saw this first-hand as a child of two working parents. My parents would occasionally work from home when child-care emergencies happened, and it was a challenging experience. This remote work in the 1980s required carrying wads of paper to the office and back, and being excluded from meetings and key decisions at work. Now we all video-conference on sleek laptops, instantly cloud-sharing files, and connect (fairly) seamlessly with remote colleagues.

The rate of technological progress is accelerating thanks to the Schumpeterian economics of “market-size effects”. When markets grow, firms want to innovate to serve the newly enlarged and more profitable market. As wfh has jumped since 2019, the rewards from producing the best video camera, video-conference package or desk-scheduling software have also shot up.

I see the virtual gold rush that this has provoked in Silicon Valley, where I live: venture capitalists, startups and established technology firms are racing to build the next remote-working gadget or app. One indicator of this is filings with America’s patent office. The share of new filings mentioning words such as “telework”, “work from home” or “remote work” spiked after the world went into covid-induced lockdown and remains at double the pre-pandemic level. The surge in wfh has sown the technological seeds of its own future acceleration.

The second force supporting remote working is business “cohort effects”. Data show that younger startups tend to be more remote-focused. These firms have been born in an era when having an office is optional and meeting customers and business partners online is standard. Many see forgoing offices and using more remote workers as a key cost-saving strategy. As a result, employees at today’s new firms work almost twice as many days from home as those at firms founded 30 years ago.

Society should embrace the Nike swoosh of wfh. Employees gain. They put a value on hybrid working that is equivalent to an 8% pay increase. Indeed, recruiters I talk to argue that the “big two” employment perks—pensions and health care—have become the “big three” with the addition of remote working.

Firms gain, too. Studies find that hybrid working can reduce employee turnover by 30-50%, as well as saving office costs and allowing companies to better tap global markets for talent. Indeed, American firms reported record profits in 2022 and the American economy has seen annual labour productivity growth accelerate from 1.2% in the five years before the pandemic to 1.5% since. There are many factors at play here, of course, but the surge in remote working is one potential contributor to this productivity renaissance…. more

Survey: Remote Work Isn’t Going Away — and Executives Know It

by Nicholas Bloom, Jose Maria Barrero, Steven Davis, Brent Meyer and Emil Mihaylov

August 28, 2023

ER Productions Limited/Getty Images

Summary.   Many CEOs are publicly gearing up for yet another return-to-office push. Privately, though, executives expect remote work to keep on growing, according to a new survey. That makes sense: Employees like it, the technology is improving, and — at least for hybrid…more

Remote work spiked during the pandemic, from about 6% of full workdays in the U.S. to more than 50% in the spring of 2020. Since then, it’s steadily decreased and since early 2023 has hovered around 28%. Many executives believe it’s time to come back to the office: Jamie Dimon, CEO of JPMorgan, has declared himself a remote-work skeptic; Mark Zuckerberg has declared that engineers “get more done” in the office; and Google’s chief people officer recently told employees that office attendance would factor into performance reviews. Even Zoom’s leadership wants employees back in person two days a week.

The only problem? Not even senior management expects this return-to-office push to work.

The Survey of Business Uncertainty is jointly run by the Atlanta Federal Reserve Bank, the University of Chicago, and Stanford. It surveys senior executives at roughly 500 U.S. businesses across industries and regions each month.

The most recent iteration of the survey, conducted in July 2023, asks:

“Looking forward to five years from now, what share of your firm’s full-time employees do you expect to be in each category [fully in person, hybrid, fully remote] in 2028?”

As the chart below illustrates, executives expect both fully remote and hybrid work to continue to grow.

See more HBR charts in Data & Visuals

They’re right to expect remote and hybrid work to increase, for four reasons.

First, as remote-working technology improves, the share of people working remotely increases. In the 1960s, offices were entirely paper-based, and working from home was very inconvenient. By the 1980s, personal computers began to become more widespread and remote work became easier. By the 2000s, the internet and nascent video calls made it easier still. The response followed basic economic logic: As the “costs” of remote work fell (lower inconvenience, for example), more people chose to do it. Work-from-home rates grew steadily over the half-century leading up to the pandemic, albeit from a very low starting point. And this trend will continue: The pandemic significantly increased the amount of research and patenting happening in technologies that support remote interactions.

Second, remote work will increase because startups born since the pandemic are more likely to use it. As these younger firms grow, the share of jobs offering remote work will increase.

Third, and perhaps least obvious, the U.S. is well positioned for remote work. Already, the U.S. has one of the highest rates of remote work of any country, behind only New Zealand and Canada among the 34 countries we surveyed. That makes sense. Remote work is a form of decentralization and personal autonomy: It gives employees more discretion over how and when they work. Management researchers have long known that for decentralized decision making to succeed, a company must be especially well managed. Separate research by our colleagues consistently finds that U.S. firms have better management practices, on average, than firms in other countries. Those better practices enable U.S. firms to more effectively manage remote work. It also helps that Americans have larger residences, which makes it easier to create a dedicated workspace at home.

Finally, remote work will increase because employees like it. The evidence suggests that working from home is valued by employees about the same as an 8% pay increase, on average. It’s a huge amenity and helps reduce turnover — in one recent, large study, by as much as 35%.

What about worries that remote work lowers productivity? Research suggests that fully remote work is up to 10% less productive than onsite work, on average. But it’s also much cheaper, because it cuts space needs and enables hiring from anywhere. Productivity in hybrid mode differs across jobs and people and with management practices. On average, however, hybrid work seems to have little net effect on productivity and may increase it. Hybrid arrangements also save on the costs — in time and money — of commuting. If employees accomplish the same amount of work whether they commute into the office two or five days a week, they’re actually spending their time more efficiently in the hybrid arrangement.

Companies and their leaders should seriously consider the merits of working from home, at least a couple of days a week. Managed hybrid, where teams all gather in the office the same day or two each week, may well be the best of both worlds. It can be profitable for companies, popular with employees, and better for the planet due to less energy consumption.

While the future extent of remote work remains uncertain, there’s little chance we will see a big return to the office. Remote technologies will only get better, and employees will gravitate to firms with more flexible policies. The biggest clue that the return-to-office push won’t work, though, is the fact that executives themselves privately predict that remote work will keep increasing.

Nicholas Bloom is a professor of economics at Stanford University.

Video of the Week

AI of the Week

Generative AI and intellectual property

Benedict Evans

If you put all the world’s knowledge into an AI model and use it to make something new, who owns that and who gets paid? This is a completely new problem that we’ve been arguing about for 500 years.

We’ve been talking about intellectual property in one way or another for at least the last five hundred years, and each new wave of technology or creativity leads to new kinds of arguments. We invented performance rights for composers and we decided that photography – ‘mechanical reproduction’ – could be protected as art, and in the 20th century we had to decide what to think about everything from recorded music to VHS to sampling. Generative AI poses some of those questions in new ways (or even in old ways), but it also poses some new kinds of puzzles – always the best kind. 

At the simplest level, we will very soon have smartphone apps that let you say “play me this song, but in Taylor Swift’s voice”. That’s a new possibility, but we understand the intellectual property ideas pretty well – there’ll be a lot of shouting over who gets paid what, but we know what we think the moral rights are. Record companies are already having conversations with Google about this. 

But what happens if I say “make me a song in the style of Taylor Swift” or, even more puzzling, “make me a song in the style of the top pop hits of the last decade”? 

A person can’t mimic another voice perfectly (impressionists don’t have to pay licence fees) but they can listen to a thousand hours of music and make something in that style – a ‘pastiche’, we sometimes call it. If a person did that, they wouldn’t have to pay a fee to all those artists, so if we use a computer for that, do we need to pay them? I don’t think we know how we think about that. We might know what the law might say, but we might want to change that.

Similar problems come up in art, and also some interesting cultural differences. If I ask Midjourney for an image in the style of a particular artist, some people consider this obvious and outright theft, but if you chat to the specialists at Christie’s or Sotheby’s, or wander the galleries of lower Manhattan or Mayfair, most people there will not only disagree but be perplexed by the premise – if you make an image ‘in the style of’ Cindy Sherman, you haven’t stolen from her and no-one who values Cindy Sherman will consider your work a substitute (except in the Richard Prince sense). I know which I agree with, but that isn’t what matters. How did we reach a consensus about sampling in hip hop? Indeed, do we agree about Richard Prince? We’ll work it out. 

Let’s take another problem. I think most people understand that if I post a link to a news story on my Facebook feed and tell my friends to read it, it’s absurd for the newspaper to demand payment for this. A newspaper, indeed, doesn’t pay a restaurant a percentage when it writes a review. If I can ask ChatGPT to read ten newspaper websites and give me a summary of today’s headlines, or explain a big story to me, then suddenly the newspapers’ complaint becomes a lot more reasonable – now the tech company really is ‘using the news’. Unsurprisingly, as soon as ChatGPT announced that it had its own web crawler, news sites started blocking it.

But just as for my ‘make me something like the top ten hits’ example, ChatGPT would not be reproducing the content itself, and indeed I could ask an intern to read the papers for me and give a summary (I often describe AI as giving you infinite interns). That might be breaking the self-declared terms of service, but summaries (as opposed to extracts) are not generally considered to be covered by copyright – indeed, no-one has ever suggested this newsletter is breaking the copyright of the sites I link to. 

Does that mean we’ll decide this isn’t a problem? The answer probably has very little to do what that today’s law happens to say today in one or another country. Rather, one way to think about this might be that AI makes practical at a massive scale things that were previously only possible on a small scale. This might be the difference between the police carrying wanted pictures in their pockets and the police putting face recognition cameras on every street corner – a difference in scale can be a difference in principle. What outcomes do we want? What do we want the law to be? What can it be?

But the real intellectual puzzle, I think, is not that you can point ChatGPT at today’s headlines, but that on one hand all headlines are somewhere in the training data, and on the other, they’re not in the model. 

OpenAI is no longer open about exactly what it uses, but even if it isn’t training on pirated books, it certainly uses some of the ‘Common Crawl, which is a sampling of a double-digit percentage of the entire web. So, your website might be in there. But the training data is not the model. LLMs are not databases. They deduce or infer patterns in language by seeing vast quantities of text created by people – we write things that contain logic and structure, and LLMs look at that and infer patterns from it, but they don’t keep it. So ChatGPT might have looked at a thousand stories from the New York Times, but it hasn’t kept them.

Moreover, those thousand stories themselves are just a fraction of a fraction of a percent of all the training data. The purpose is not for the LLM to know the content of any given story or any given novel – the purpose is for it to see the patterns in the output of collective human intelligence. 

That is, this is not Napster. OpenAI hasn’t ‘pirated’ your book or your story in the sense that we normally use that word, and it isn’t handing it out for free. Indeed, it doesn’t need that one novel in particular at all. In Tim O’Reilly’s great phrase, data isn’t oil; data is sand. It’s only valuable in the aggregate of billions,, and your novel or song or article is just one grain of dust in the Great Pyramid. OpenAI could retrain ChatGPT without any newspapers, if it had to, and it might not matter – it might be less able to answer detailed questions about the best new coffee shops on the Upper East Side of Manhattan, but again, that was never the aim. This isn’t supposed to be an oracle or a database. Rather, it’s supposed to be inferring ‘intelligence’ (a placeholder word) from seeing as much as possible of how people talk, as a proxy for how they think. 

On the other hand, it doesn’t need your book or website in particular and doesn’t care what you in particular wrote about, but it does need ‘all’ the books and ‘all’ the websites. It would work if one company removed its content, but not if everyone did.

If this is, at a minimum, a foundational new technology for the next decade (regardless of any talk of AGI), and it relies on all of us collectively acting as mechanical turks to feed it (even if ex post facto), do we all get paid, or do we collectively withdraw, or what? It seems somehow unsatisfactory to argue that “this is worth a trillion dollars, and relies on using all of our work, but your own individual work is only 0.0001% so you get nothing.” Is it adequate or even correct to call this “fair use”? Does it matter, in either direction? Do we change our laws around fair use? 

In the end, it might not matter so much: the ‘Large’ in ‘Large Language Models’ is a moving target. The technology started working because OpenAI threw orders of magnitude more data into the hopper than anyone thought reasonable and great results came out of the other end, but we can’t add orders of magnitude more data again, because there genuinely isn’t that much more data left. Meanwhile, the cost and scale of these things means that a large part of the research effort now goes into getting the same or better results with much less data. Maybe they won’t need your book after all. 

Meanwhile, so far I’ve been talking about what goes into the model – what about the things that come out? What if I use an engine trained on the last 50 years of music to make something that sounds entirely new and original? No-one should be under the delusion that this won’t happen. Having suggested lots of things where I don’t think we know the answers, there is one thing that does seem entirely clear to me: these things are tools, and you can use a tool to make art or to make cat pictures. I can buy the same camera as Cartier-Bresson, and I can press the button and make a picture without being able to draw or paint, but that’s not what makes the artist – photography is about where you point the camera, what image you see and which you choose. No-one claims a machine made the image. Equally, I can press ‘Go’ on Midjourney or ChatGPT without any skill at all, but getting something good is just as hard. Right now they’re at the Daguerreotype stage, but people will use these to make art that we haven’t imagined, not because we lack the skill, but because we’re not artists. Nadar is coming.

The more interesting problem, perhaps, might be that Spotify already has huge numbers of ‘white noise’ tracks and similar, gaming the recommendation algorithm and getting the same payout per play as Taylor Swift or the Rolling Stones. If we really can make ‘music in the style of the last decade’s hits,’ how much of that will there be, and how will we wade through it? How will we find the good stuff, and how will we define that? Will we care? ….more

Free AI tools are killing South Africa’s web designer job market

People are using ChatGPT to code and create basic designs, while salaries for web designers have dropped.



South Africa’s web designers struggle to stay afloat as new players using cheap AI tools flood the market.

Earnings have dropped significantly as recruitment agents push to hire fewer professionals.

Donald Bengu started freelancing as a website designer in 2018, right after graduating college. By 2022, he had a booming career, with constant gigs through gig work platform Fiverr. His clients included small businesses and large corporations across South Africa, and he earned up to 45,000 rand (around $2,422) per month.

But by early 2023, the tide began to turn. Bengu saw a noticeable decline in the availability of new projects — a stark contrast to his previous thriving workload. Now, he gets barely two new gigs a month, and makes around 5,000 rand ($270). Bengu blames the rise of artificial intelligence tools for his plight.

“Getting web designer gigs is no longer as easy, and the market is now cheap due to AI,” he told Rest of World. “The job market for web designers in South Africa is very fickle. With the introduction of AI … web designers [are] racing to the bottom with their prices.” He is now looking for full-time jobs where he can use his skills to earn a steady income.

South African web designers told Rest of World they are struggling to find work and earn enough as free AI tools are replacing their skills. According to Bengu and others, anyone can become a web designer by using free services like ChatGPT, WordPress and React, leading to an overcrowded market. They now find it hard to ask for reasonable rates for their work. 

“The web design market has become unstable,” Sikhulile Dube, a Johannesburg-based freelance web designer, told Rest of World. Dube has worked on projects through platforms like Indeed and Fiverr for the past 12 years. “Professionals like myself have more to offer in terms of quality. Low prices being charged by new players are killing the market,” he said. He has now been looking for clients beyond South Africa, including countries like Ethiopia and Mozambique, to stay afloat.

Five years ago, web designing was one of the hottest jobs in South Africa, with plentiful opportunities that paid high salaries, according to Jupiter Punungwe, tech researcher at Discovery, a South African financial services company. “Being just a web designer is no longer enough, and you find that new designers can easily create basic websites by using free AI tools,” Punungwe told Rest of World. “Bigger businesses now require full-stack developers familiar with front-end, back-end, and middle-wave web development.” 

The average monthly salary for web designers in South Africa was around 38,000 rand ($2,040) in 2018. Freelance work is now drying up and few local companies want to hire full-time web developers, Tatenda Chatukuta, engineering talent acquisition manager at recruitment agency Takora, told Rest of World. Salaries for web designers in the country have also dropped to a little over 12,000 rand ($643). “Learning new AI trends will increase chances of being employable [as a web designer],” she said.

While more experienced web designers see AI as a threat, others have benefited from the new technology.

Thabiso Malatji, a 20-year-old self-taught web designer from Pretoria, started doing freelance work in January. He told Rest of World he uses ChatGPT to code and create basic designs, and is happy to charge low rates to get work. “My services are cheap because I mostly use free tools online,” Malatji said. “Gone are the days when people needed to pay a fortune for a website.” He learned how to build websites through YouTube tutorials, and works on WordPress. Making an entire website costs him next to nothing, Malatji said. He charges 1,500–2,000 rand ($80.43–$107) for his services, which also covers the costs of domain registration and basic web hosting. 

Web designers risk becoming obsolete if they do not upskill, Scott Wentworth, online marketing analyst and strategist, told Rest of World

“The future of web designing in South Africa is at an equilibrium stage where going forward, web designers should evolve to avoid becoming obsolete,” Wentworth said. “While AI makes web designing easier, there is no substitute for human touch. Web designers should get geared [up] for the future of using AI.”

OpenAI Passes $1 Billion Revenue Pace as Big Companies Boost AI Spending

By Amir Efrati and Aaron Holmes

Aug. 29, 2023

OpenAI is currently on pace to generate more than $1 billion in revenue over the next 12 months from the sale of artificial intelligence software and the computing capacity that powers it. That’s far ahead of revenue projections the company previously shared with its shareholders, according to a person with direct knowledge of the situation.

The billion-dollar revenue figure implies that the Microsoft-backed company, which was valued on paper at $27 billion when investors bought stock from existing shareholders earlier this year, is generating more than $80 million in revenue per month. OpenAI generated just $28 million in revenue last year before it started charging for its groundbreaking chatbot, ChatGPT. The rapid growth in revenue suggests app developers and companies—including secretive ones like Jane Street, a Wall Street firm—are increasingly finding ways to use OpenAI’s conversational text technology to make money or save on costs. Microsoft, Google and countless other businesses trying to make money from the same technology are closely watching OpenAI’s growth.

• OpenAI is signing big deals with customers to access its GPT-4 model
• Its business is a barometer of the commercial potential of language models
• Jane Street hired a prominent AI language model researcher from Google

The company’s bottom line couldn’t be learned, but it lost around $540 million last year as it developed GPT-4 and ChatGPT. Features involving large-language models generally require servers with special chips that draw more power than the servers powering traditional software features. OpenAI CEO Sam Altman did not have a comment.

The percentage of revenue OpenAI generates from ChatGPT subscriptions versus selling access to GPT-4 through an application programming interface also couldn’t be learned. But in March of this year, OpenAI had between 1 million and 2 million ChatGPT subscribers paying $20 per month, said a person with knowledge of the figure.

The technology, also known as a large-language model, is far from perfect but already has proven useful in developing software that suggests code to software engineers as they type, automates customer-service responses, and aids securities-trading firms such as Jane Street. The latter has become a meaningful customer of OpenAI and of Microsoft’s cloud AI services, according to two people with knowledge of the company’s spending.

News Of the Week

Marc Penkala

𝟭𝟱.𝟮% 𝗼𝗳 𝗿𝗼𝘂𝗻𝗱𝘀 𝗰𝗼𝗺𝗽𝗹𝗲𝘁𝗲𝗱 𝗶𝗻 𝗤𝟮/𝟮𝟯 𝘄𝗲𝗿𝗲 𝗱𝗼𝘄𝗻 𝗿𝗼𝘂𝗻𝗱𝘀

𝗦𝗼 𝘄𝗵𝗮𝘁 𝗶𝘀 𝗵𝗮𝗽𝗽𝗲𝗻𝗶𝗻𝗴 𝗮𝘁 𝘁𝗵𝗲 𝗺𝗼𝗺𝗲𝗻𝘁?

𝗙𝗶𝗿𝘀𝘁𝗹𝘆, the median time between VC rounds increased by almost 30% over the past 12 months. 𝗦𝗲𝗰𝗼𝗻𝗱𝗹𝘆, the early-stage demand-to-supply ratio hit an all-time low and rose by 85% over the past 12 months. 𝗧𝗵𝗶𝗿𝗱𝗹𝘆, VCs are applying more protective terms, such as liquidation multiples, cumulative dividends, and other structural terms that are more heavily dilutive for founders and existing investors. 𝗟𝗮𝘀𝘁𝗹𝘆, 15.2% of the rounds completed in Q2/23 were down rounds, an increase of 300% over the past 12 months.

𝗕𝘂𝘁 𝘄𝗵𝘆, 𝘄𝗵𝗮𝘁 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗱𝗿𝗶𝘃𝗲𝗿𝘀?

1. 𝗗𝗲𝗰𝗿𝗲𝗮𝘀𝗲𝗱 𝗖𝗮𝗽𝗶𝘁𝗮𝗹 𝗔𝘃𝗮𝗶𝗹𝗮𝗯𝗶𝗹𝗶𝘁𝘆
The US VC market has experienced a significant decline in available capital compared to previous years, evident by the contrast between the USD 347.5 Bn invested in 2021 and the USD 85.6 Bn invested through Q2/23.

2. 𝗖𝗿𝗼𝘀𝘀𝗼𝘃𝗲𝗿 𝗜𝗻𝘀𝘁𝗶𝘁𝘂𝘁𝗶𝗼𝗻 𝗣𝘂𝗹𝗹𝗯𝗮𝗰𝗸
Crossover institutions involved in large VC deals have drastically reduced participation, decreasing the volume of deal value from over 550 deals and a quarterly peak of nearly USD 50 Bn in 2021 to fewer than 200 deals and a quarterly peak of around USD 15 Bn in 2023.

3. 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗶𝗻𝗴 𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗻𝗴 𝗠𝗮𝗿𝗸𝗲𝘁𝘀
Capital availability has been low across the market, including in seed and early-stage financing. Investors are being more cautious and deliberate in deploying their capital.

4. 𝗦𝘂𝗽𝗽𝗹𝘆-𝗗𝗲𝗺𝗮𝗻𝗱 𝗜𝗺𝗯𝗮𝗹𝗮𝗻𝗰𝗲
The US VC market has been operating for five consecutive quarters with less capital available than what is estimated to be demanded. The inability of companies to exit may lead to a situation where additional private financing is required, adding to market demand.

5. 𝗦𝗵𝗶𝗳𝘁 𝗶𝗻 𝗜𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁 𝗟𝗮𝗻𝗱𝘀𝗰𝗮𝗽𝗲
There has been a shift in the availability of capital, with valuations coming down. Down rounds have seen a surge, with 15.2% of rounds completed in Q2/23 at a lower valuation than the prior round.

These points collectively paint a picture of reduced capital availability, which will not change any time soon (𝗙𝘂𝗻 𝗙𝗮𝗰𝘁 : More than 50% of the dry powder in 2020 to 2022 vintages is stored in just 7% of the closed vehicles).

#VentureCapital #DownRounds #Startups #MarketyDynamics

Nvidia vs. AMD vs. Intel: Comparing AI Chip Sales

Published on August 25, 2023

By Jenna Ross, Graphics/Design: Sam Parker

Nvidia vs. AMD vs. Intel: Comparing AI Chip Sales

Nvidia has become an early winner of the generative AI boom.

The company reported record revenue in its second quarter earnings report, with sales of AI chips playing a large role. If we compare to other American competitors, what do the AI chip sales of Nvidia vs. AMD vs. Intel look like?

In this graphic, we use earnings reports from each company to see their revenue over time.

A Clear Leader Emerges

While the companies don’t report revenue for their AI chips specifically, they do share revenue for their Data Center segment.

The Data Center segment includes chips like Central Processing Units (CPUs), Data Processing Units (DPUs), and Graphic Processing Units (GPUs). The latter are preferred for AI because they can perform many simple tasks simultaneously and efficiently.

Below, we show how quarterly Data Center revenue has grown for Nvidia vs. AMD vs. Intel.

Nvidia’s Data Center revenue has quadrupled over the last two years, and it’s estimated to have more than 70% of the market share for AI chips.

The company achieved dominance by recognizing the AI trend early, becoming a one-stop shop offering chips, software, and access to specialized computers. After hitting a $1 trillion market cap earlier in 2023, the stock continues to soar.

Competition Between Nvidia vs. AMD vs. Intel

If we compare Nvidia vs. AMD, the latter company has seen slower growth and less revenue. Its MI250 chip was found to be 80% as fast as Nvidia’s A100 chip.

However, AMD has recently put a focus on AI, announcing a new MI300X chip with 192GB of memory compared to the 141GB that Nvidia’s new GH200 offers. More memory reduces the amount of GPUs needed, and could make AMD a stronger contender in the space.

In contrast, Intel has seen yearly revenue declines and has virtually no market share in AI chips. The company is better known for making traditional CPUs, and its foray into the AI space has been fraught with issues. Its Sapphire Rapids processor faced years of delays due to a complex design and numerous glitches.

Going forward, all three companies have indicated they plan to increase their AI offerings. It’s not hard to see why: ChatGPT reportedly runs on 10,000 Nvidia A100 chips, which would carry a total price tag of around $100 million dollars.

As more AI models are developed, the infrastructure that powers them will be a huge revenue opportunity.

Startup of the Week

The Next Great SaaS IPO is Officially Coming: Klaviyo

by Jason Lemkin | Blog Posts, Scale

So we’ve been saying for a while that the back half of 2024 could be good for SaaS IPOs, we just needed a few of the break-out winners to IPO to get the engine rolling again.  The SaaS IPO window really closed in late 2021 with HashiCorp as the last great one to IPO in December 2021.

And now we have that first great IPO filing since 2021 — Klaviyo.

You may not have heard of Klaviyo if you are outside of e-commerce, but in the e-comm world, it’s the #1 B2B player.  It dominates the Shopify marketing ecosystem and others as well.

And it’s got the full package:

True Scale:  Almost $600m ARR

Epic Growth:  Still growing 57% at $600m ARR — wow!

Top Tier NRR: 119%

Profitable:  Klaviyo has gotten leaner, and has been profitable the past 6+ months.

There’s nothing to knock here, folks.   And … it was essentially incredibly capital-efficient as well.  Borderline bootstrapped (more to come here on SaaStr).  They only burned $15 million to get to $585m in ARR and an IPO filing!

“Since we were founded in Boston, Massachusetts by Andrew Bialecki and Ed Hallen, we have been able to reach significant scale, with revenue of $472.7 million for the year ended December 31, 2022. Efficiency is part of our DNA. We have raised $454.8 million in primary capital since our inception, of which we have utilized only $15.0 million in the operation of our business as of June 30, 2023, which is net of the $439.8 million of cash, cash equivalents, and restricted cash on our balance sheet as of June”

This is the break-out IPO the markets need to get the engine going again.  Not ARM, or Cava.  No, we need a great SaaS IPO, with strong unit economics, high NRR, and that is efficient.

Klaviyo has it all.  Wow.

A New Marketing Behemoth Klayviyo : How 7 Key Benchmarks Stack Up in the S-1

On Friday, Klayviyo filed their S-1 – one of the first software companies to do so since the beginning of the economic downturn in late 2022. It’s a marvel of business with strong growth, great efficiency & the potential to reignite interest in SaaS IPOs.

Klayviyo is a customer data platform company which ingests marketing data & empowers their customers to automate their marketing with data & AI. The company raised $455m in venture capital, but has used only $15m, growing nearly entirely on profits.

Quoting from the S-1, “Our go-to-market strategy is primarily product-led, and we attract the majority of our new customers through inbound channels, such as word-of-mouth, agency partnerships, and platform integrations.”

The PLG efficiency is tremendous both in cash & sales efficiency, which tops the charts of public software companies at 1.03 for the last 6 months. When the company spends $1 in sales & marketing expense, Klayviyo produces $1.03 in gross profit next year.

Shopify is a key partner & significant investor in the business. “Approximately 77.5% of our total ARR as of December 31, 2022 came from customers who also use the Shopify platform; however, the vast majority of those customers came to us through inbound channels or through other means such as our marketing agency partners. For 2022, approximately 10.6% of our new ARR was attributable to customers that chose to install Klaviyo through Shopify’s App Store.”

Braze is the most recent marketing SaaS company to IPO & we can contrast the two businesses. Braze targets enterprise buyers with a sales-led motion.

At IPO, the companies grew at similar rates, but Klayviyo is roughly 215% larger in revenue.

Gross margins are comparable, with Klayviyo five percentage points higher. Braze may offer more professional services / support to larger customers which may reflect in gross margin.

The difference in customer base is apparent in the ACV : $4.5k vs 169k, a 37x delta. However, NDRs (net dollar retention) are equal. Typically, enterprise customers expand more than SMB.

Klayviyo’s sales efficiency tops the market at 1.04 benefiting from the strategic partnership with Shopify & their PLG motion.

On the profitability & cash flow metrics, this compares Braze during the low interest rate environment when public market investors didn’t factor efficiency heavily. Klayviyo shines : generating profits & positive cash flow from operations.

A basic linear model using Revenue Growth, Sales Efficiency, Revenue, Gross Margin, Profitability, & CFO / Revenue suggests the company should trade at about a $10b valuation with an 11.2x multiple if the stock behaves similarly to the rest of the 75 or so publicly traded software companies.

Klayviyo is a fast growing, incredibly efficient, profitable software company : an ideal specimen for one of the first software IPOs in the fall. Congratulations to the team on building a paragon of a SaaS business!

Klaviyo: Benchmarking the S-1 Data


Klaviyo filed their initial S1 statement today. This is the first S-1 we’ve seen in almost 2 years from a software company! A S-1 is a document companies file with the SEC in preparation for listing their shares on an exchange like the NYSE or NASDAQ. The document contains a plethora of information on the company including a general overview, up to date financials, risk factors to the business, cap table highlights and much more. The purpose of the detailed information is to help investors (both institutional and retail) make informed investment decisions. There’s a lot of info to digest, so in the sections below I’ll try and pull out the relevant financial information and benchmark it against current cloud businesses. As far as an expected timeline – typically companies launch their roadshow ~3 weeks after filing their initial S-1 (the roadshow launches with an updated S-1 that contains a price range). After the roadshow launch there’s typically ~2 weeks before the stock starts trading. So we’re looking at roughly 5 weeks before any retail investor can buy the stock.

Klaviyo Overview

From the S1 – “Klaviyo enables businesses to drive revenue growth by making it easy to bring their first-party data together and use it to create and deliver highly personalized consumer experiences across digital channels.

Our modern and intuitive SaaS platform combines our proprietary data and application layers into one vertically-integrated solution with advanced machine learning and artificial intelligence capabilities. This enables business users of any skill level to harness their data in order to send the right message at the right time across email, SMS, and push notifications, more accurately measure and predict performance, and deploy the specific actions and campaigns that drive the highest impact. Our reviews add-on allows our customers to collect product reviews within our platform to provide a seamless experience across the customer lifecycle, and our CDP offering gives customers user-friendly ways to track new types of data, transform and cleanse data, run more advanced reporting and predictive analysis to drive revenue growth, and sync data in to and out of Klaviyo at scale. By combining easy implementation, rapid time-to-value, and clearly attributable outcomes, which we measure and refer to as KAV, we drive substantial return-on-investment, or ROI, for our customers. We focused on marketing automation within eCommerce as our first application use case, and we believe our software is highly extensible across a broad range of functions and verticals.”

Product Overview

At their core, Klaviyo is a marketing automation solution that covers email, SMS, mobile push, and reviews, tying it all together in a customer data platform. Customers can segment their customers, create specific campaigns that cover multiple channels, and overall engage more efficiently with their audience through the Klayvio platform

The Klaviyo platform is a “vertically-integrated, highly-scalable, and flexible platform unifies the data and application layers with our messaging infrastructure into one modern tech stack.”

“Data Layer. Our highly-scalable platform is optimized for large volumes of data, delivers sub-second-level accessibility, and provides extremely high levels of personalization and attribution. We built our data store from the ground up to be agile, unbound by specific schema or data structures, and, as of June 30, 2023, we are able to process data from over 300 native integrations and open APIs without friction. Our integrations span from retail and eCommerce platforms – such as Shopify, Salesforce Commerce Cloud, and WooCommerce – to loyalty, social media, customer service, and shipping solutions. Additionally, we have begun to launch integrations for new verticals, including Mindbody, Zenoti, and Olo. Our data store synchronizes unaggregated, historical profile data with real-time event data in a single system-of-record. Profile data enables our customers to generate unified consumer profiles with extremely granular segmentation, grouping consumer profiles into precise audiences that update in real-time as consumers interact with our customers. Event data allows customers to send behavior-triggered messages that keep consumers engaged with the right message at the right time. This industry-agnostic, data-first approach represents a new foundational capability in our market and can be applied to several use cases and new verticals in the future that all require the combination of fast performance with real time, predictive intelligence.

Application Layer. We built an application layer on top of our data layer, which provides a comprehensive set of tools and features that enable our customers to easily turn consumer learnings into insights and actions to drive revenue growth without the need to hire sophisticated and expensive in-house engineers. We started with our marketing application, enabling our customers to create and manage targeted marketing campaigns and flows, track customer behavior, and analyze campaign performance to grow revenue. Our advanced data science and predictive analytics capabilities also utilize artificial intelligence and machine learning so businesses can estimate consumer lifetime value, predict a consumer’s next order date, and calculate potential churn risk. As a result, our application helps companies deliver contextually-relevant and personalized experiences throughout the entire consumer journey and across digital channels, such as email, SMS, and push notifications, through our messaging infrastructure. We focused on marketing automation for business-to-consumer, or B2C, companies within retail and eCommerce as our first application, and we believe our software is highly extensible across a broad range of functions for B2C and B2B businesses alike.”

Market Opportunity

From the S-1: “Today, the customers we serve primarily operate within the retail vertical, with retailers spanning both online and offline channels. Our estimated serviceable addressable market opportunity within this vertical is over $16 billion. We calculate this opportunity using business count data sourced from Analysys Mason, focusing on the number of businesses in the geographies we primarily operate in today, including North America, Western Europe, New Zealand, and Australia, and segmenting them into Micro, Small, Medium, and Enterprise segments based on the number of employees. We then multiply the total number of companies in each segment by our respective average ARR per customer per segment as of December 31, 2022. The average ARR is based on customers in each segment in all geographies that have been using our email and SMS offering for more than twelve months.

While our first use cases were focused on the retail and eCommerce vertical, we believe our platform is highly extensible across a broad range of verticals, including education, events and entertainment, restaurants, and travel, as well as B2B companies. As we continue to scale our platform, we expect that our total addressable market will expand to businesses in all verticals that engage with third parties, including customers and clients, through email, SMS or push. Accordingly, we estimate that the total addressable market opportunity for our platform across all of these verticals is $34 billion in the United States alone.”

How Klaviyo Makes Money

Klaviyo has about 130k customers, with an ACV of ~$5k

X of the Week

Leave a Reply

%d bloggers like this: