Curated Insights 2018.12.21

Investing ideas that changed my life

If something has a chance of either destroying you or making you very wealthy and you don’t know how to measure what that chance is, it’s understandable that people default to high levels of credulity.

You can’t believe in risk without also believing in luck, because they are fundamentally the same thing—an acknowledgment that you are one person in a 7 billion player game, and the accidental impact of other people’s actions can be more consequential than your own. But the path of least resistance is to be keenly aware of risk when it affects you, and oblivious to luck when it helps you. Investors adjust returns for risk; never for luck. Companies disclose known risks in their annual reports; lucky breaks are rarely mentioned. The danger is that experiencing risk reduces confidence when it should merely highlight reality, which can make future decisions more conservative than they ought to be. Luck increases confidence without increasing ability, which makes people double down with less room for error than before. Realizing that luck and risk are ever-present and normal makes you accept that not everything is in your control, which is the only way to identify whatever is in your control.

If you think the world is all art, you’ll miss how much stuff is too complicated to think about intuitively. But if you think the world is all science, you’ll miss how much people like to take shortcuts, believe only what they want to believe, and have to deal with stuff that is too complicated for them to summarize in a statistic. Another way to think about this: Investing is not physics, which is guided by cold, immutable laws. It’s like biology, guided by the messy mutations and accidents of evolution, constantly adapting and sometimes defying logic.

“How long will this remain important given my strategy and time horizon?”


The dynamics of network effects

So how can entrepreneurs and founders navigate this era of seemingly diminishing network effects? The trick is to know what your network effects look like today, but also project how they’ll evolve over time. To that end, you’ll need to understand three aspects of your company and how they could change going forward: 1) your value proposition, 2) your users/inventory, and 3) your competitive ecosystem. Otherwise you could get caught flat-footed, claiming that network effects are dead.

Early on, more friends in Frank groups meant more demand and more liquidity, which created a bigger incentive for people to join those groups. But once a group had more than 7 people, they became less likely to lend or borrow: turns out people only have ~7 friends/family members they have that level of trust with! The network effects in this case went from positive to negative as an individual’s network outgrew the value proposition. This pattern has also held true for a number of other highly social products.

Platforms/marketplaces with more differentiated inventory have stronger and longer-lasting network effects, because they have a diversity of inventory that suits the unique preferences of customers (while maintaining just-enough substitutability across that inventory as well). For example, AirBnB can show users every iteration of lodging from $225-$325/night in Los Angeles, which overlaps with someone else’s search for something that costs $150-$250 and has a both a balcony and a hot tub. The platform is therefore more valuable on both sides of the marketplace than a site that just shows a commoditized set of standard and executive rooms. The network effects remain strong not only because it reaches a base level of liquidity across all these different types of inventory (making them valuable to more users), but because it also continues to see increasing returns with new supply.

When you forecast out your network effects — and more importantly, your growth strategy for acquiring and engaging more users — you will need to pay attention to the incremental users you’re likely to attract. Are they network “contaminants”, “neutrals”, or “contributors”? For a social network, adding a troll that disengages other users is a pollutant who removes value. Adding a lurker is neutral since that person doesn’t add or subtract any value from the network. Adding a great content producer contributes an enormous amount of value to the network. So, making sure to incent the users you want while disincenting the ones you don’t want, is key. This is why most great platforms also invest heavily in curation mechanisms to screen and remove bad inventory/users (e.g., Wikipedia’s editors, Airbnb’s reviews/onboarding, etc.). Unfortunately, these screening mechanisms don’t always work and sometimes the cost of finding strong contributors becomes very high, so the calculus of growth relative to cost matters a lot here.

While network effects businesses tend to be more defensible at scale, they are not immune to competition. But for these types of businesses it’s not just a matter of figuring out who your direct competitors are — you also need to think about the network overlap. If someone else has a similar network to yours, there’s always existential risk they’ll move into your market. Because they have a similar network already, they’ll more easily be able to enter your space (Instagram’s foray into Snapchat-like disposable “Stories” is a good example of this). This is also true where the competition may already have registered a superset of your network (e.g., DoorDash and Uber Eats; Didi and Uber in China).

The increasing speed of product iteration, the pace at which networks can scale, and the ease with which competitors can get started has therefore dramatically changed how we project network effects in businesses. Instead of winner-take-all markets where early movers may have once had a seemingly lasting advantage, network effects change more quickly than ever. Especially where specific factors — an asymptotic value proposition, network overlap, increasing number of contaminants, etc. — can lower the platform’s ability to generate a sustainable network effect in the future.

How much is social media worth? Estimating the value of Facebook by paying users to stop using it

As noted previously, Facebook reached a market capitalization of $542 billion in May 2018. At 2.20 billion active users in March 2018, this suggests a value to investors of almost $250 per user, which is less than one fourth of the annual value of Facebook access from any of our samples. This reinforces the idea that the vast majority of benefits of new inventions go not to the inventors but to the users. Further, our results provide evidence that online services can provide tremendous value to society even if their contribution to GDP is minimal. If the billions of people who use Facebook and other free online services derive anything close to $1000 per year in benefits, the productivity slowdown cited by Solow and others may not be reflected in a slowdown in the growth rate of welfare measures like consumer surplus. Many observers have commented on the difficulties of measuring productivity growth in great technological change. While our current study does not offer a solution that can be broadly applied to address this challenge, it does present a methodology and results that provide important insight into the scale of the issue when considering the online revolution of our current era.

Concerns about data privacy, such as Cambridge Analytica’s alleged problematic handling of users’ private information, which are thought to have been used to influence the 2016 United States presidential election, only underscore the value Facebook’s users must derive from the service. Despite the parade of negative publicity surrounding the Cambridge Analytica revelations in mid-March 2018, Facebook added 70 million users between the end of 2017 and March 31, 2018. This implies the value users derive from the social network more than offsets the privacy concerns.


Alibaba stock poised to return 200%, advisor says

Alibaba has a unique business model where it operates solely as a platform, rather than a middle man. The company doesn’t have to purchase inventory, provide logistics, or distribute product – it simply collects fees from merchants for advertising and commissions for completed transactions. This asset light model has allowed BABA to compound earnings more than 43% per year for the last 5 years with very little incremental capital.

The GMV for BABA in the last twelve months was a staggering $4.8 trillion yuan, or $768B USD. This towers over Amazon’s $186B or Walmart’s $495B. GMV is nearly 7% of the GDP of China. In the last 5 years GMV has compounded at an annual growth rate of 29% per year.

Core Commerce is the largest and most profitable division of BABA representing 71% of total revenue and 100% of owner earnings. This division generates revenue by selling advertising to merchants and collecting commissions ranging from 0.3% to 5.0% on sales that occur across BABA’s e-commerce platforms. In the last 5 years revenue from the Core Commerce division has compounded at a rate of 43% per year.

Cloud Computing provides individuals, merchants, and businesses across China online access to the vast computing resources of BABA’s datacenters. Alibaba Cloud offers a complete suite of cloud services, including elastic computing, database, storage, network virtualization services, large scale computing, security, management and application services, big data analytics, and more. Alibaba Cloud has grown at an average pace of more than 100% in the last 5 years. While the business currently does not generate owner earnings due to the aggressive investment in market share, we are confident that the division will be highly profitable in the future.

Digital Media & Entertainment offers an online platform, Youku, where users can watch TV shows, movies, and other content. It is similar to the business model of Netflix, where revenue is generated by selling subscriptions and advertising. While both revenue and daily average subscriber growth has been impressive averaging more than 100% per year for the last 3 years, the business loses money annually due to the high cost of purchasing content. Unlike Alibaba Cloud, which we are confident will be profitable based on comparisons to AWS and Google Cloud, we are less confident in the future profitability of Youku. Comparable companies, like iQiyi and Netflix, have never generated positive cash flow for owners and the path to a successful business model is not presently clear. We are hopeful that this business division will be spun off as a standalone business in the upcoming years.

“As a result of our broad value propositions to consumers, we have seen increased engagement over time. The longer consumers have been with us, the larger numbers of orders they tend to place, across a more diverse range of product categories, and the more they tend to spend on our China retail marketplaces. For example, in the twelve months ended March 31, 2018, consumers who have been with us for approximately five years placed an average of 132 orders in 23 product categories with average spending of approximately RMB12,000 in terms of GMV, whereas consumers who have been with us for approximately one year placed an average of 27 orders in 6 product categories with average spending of approximately RMB3,000 in terms of GMV. In the twelve months ended March 31, 2018, the average annual active consumer on our China retail marketplaces placed 90 orders in 16 product categories with average spending of approximately RMB9,000 in terms of GMV.”

The business case for serverless

The case for serverless starts with a simple premise: if the fastest startup in a given market is going to win, then the most important thing is to maintain or increase development velocity over time. This may sound obvious, but very, very few startups state maintaining or increasing development velocity as an explicit goal. “Development velocity,” to be specific, means the speed at which you can deliver an additional unit of value to a customer. Of course, an additional unit of customer value can be delivered either by shipping more value to existing customers, or by shipping existing value—that is, existing features—to new customers.

Whereas a ‘normal’ cloud server like AWS’s EC2 offering had to be provisioned in advance and was billed by the hour regardless of whether or not it was used, AWS Lambda was provisioned instantly, on demand, and was billed only per request. Lambda is astonishingly cheap: $0.0000002 per request plus $0.00001667 per gigabyte-second of compute. And while users have to increase their server size if they hit a capacity constraint on EC2, Lambda will scale more or less infinitely to accommodate load — without any manual intervention. And, if an EC2 instance goes down, the developer is responsible for diagnosing the problem and getting it back online, whereas if a Lambda dies another Lambda can just take its place.

Although Lambda—and equivalent services like Azure Functions or Google Cloud Functions—is incredibly attractive from a cost and capacity standpoint, the truth is that saving money and preparing for scale are very poor reasons for a startup to adopt a given technology. Few startups fail as a result of spending too much money on servers or from failing to scale to meet customer demand — in fact, optimizing for either of these things is a form of premature scaling, and premature scaling on one or many dimensions (hiring, marketing, sales, product features, and even hierarchy/titles) is the primary cause of death for the vast majority of startups. In other words, prematurely optimizing for cost, scale, or uptime is an anti-pattern.

Herein lies the magic of using managed services. Startups get the beneficial use of the provider’s code as an asset without holding that code debt on their “technical balance sheet.” Instead, the code sits on the provider’s balance sheet, and the provider’s engineers are tasked with maintaining, improving, and documenting that code. In other words, startups get code that is self-maintaining, self-improving, and self-documenting—the equivalent of hiring a first-rate engineering team dedicated to a non-core part of the codebase—for free. Or, more accurately, at a predictable per-use cost. Contrast this with using a managed service like Cognito or Auth0. On day one, perhaps it doesn’t have all of the features on a startup’s wish list. The difference is that the provider has a team of engineers and product managers whose sole task is to ship improvements to this service day in and day out. Their exciting core product is another company’s would-be redheaded stepchild.

One day, complexity will grow past a breaking point and development velocity will begin to decline irreversibly, and so the ultimate job of the founder is to push that day off as long as humanly possible. The best way to do that is to keep your ball of mud to the minimum possible size— serverless is the most powerful tool ever developed to do exactly that.

Huawei ban casts shadow over $100bn economic sphere

Huawei reported sales of 603.62 billion yuan ($87.4 billion at current rates) last year — not far off from Microsoft and Google parent Alphabet, although less than half as much as Apple. Its two biggest telecom equipment rivals, Nokia and Ericsson, had net sales of 23.1 billion euros ($26.1 billion) and 201.3 billion krona ($22.2 billion), respectively, last year. It ranks as the world’s top seller of base stations for wireless networks with a 28% share, the No. 2 maker of smartphones and routers, and the fourth-largest server manufacturer.

Privately owned Huawei spent $14 billion on outside procurement of semiconductors alone last year. Much of this came from American companies, with $1.8 billion in purchases from Qualcomm and $700 million from Intel, according to Chinese media.

Hikvision, which is more than 40% controlled by state-owned companies, is the world’s leading maker of security cameras and offers image-analysis technology powered by artificial intelligence. Privately run Hytera boasts world-leading production capacity for specialized wireless communications technology used by police and the military. Hikvision and Hytera in recent years have both acquired foreign peers — something that Beijing normally puts strict limits on — to expand their technological capabilities and overseas presence.

LVMH inks $2.6 billion deal to buy ‘21’ club operator Belmond

The acquisition is one of LVMH founder Bernard Arnault’s biggest, rivaling the purchases of Bulgari and Loro Piana. It comes as consumers shift spending toward trips, health clubs, restaurants and entertainment and interest in shopping malls dwindles.

The acquisition addresses another challenge facing LVMH and rivals Kering SA and Richemont. They’ve snapped up so many of the world’s leading brands that there are few prominent leather and couture labels left to buy. The Louis Vuitton owner, formed through a merger with Champagne and cognac maker Moet Hennessy, has already expanded into perfume, watches, jewelry and cosmetics retail. Prominent remaining independents like Chanel and Hermes have shown little inclination to sell.

The deal will expand the French company’s high-end hospitality offerings. LVMH formed a hotel management group in 2010 to oversee its operations in the sector, which include properties under the Cheval Blanc name in locations like the Courchevel ski resort in the French Alps. LVMH’s Bulgari jewelry brand has six hotels, including one in Shanghai that opened in July. It plans to open hotels in Moscow, Paris and Tokyo in the next four years.

Belmond, which used to be known as Orient-Express Hotels, owns or has stakes in more than 30 high-end hotels around the world, from St. Petersburg to Anguilla in the Caribbean. In addition to the ‘21’ Club power restaurant in Manhattan, its stable of luxury properties includes a cruise line in France, a London-to-Venice train line and safari camps in Botswana.

What straight-A students get wrong

The evidence is clear: Academic excellence is not a strong predictor of career excellence. Across industries, research shows that the correlation between grades and job performance is modest in the first year after college and trivial within a handful of years. For example, at Google, once employees are two or three years out of college, their grades have no bearing on their performance.

Academic grades rarely assess qualities like creativity, leadership and teamwork skills, or social, emotional and political intelligence. Yes, straight-A students master cramming information and regurgitating it on exams. But career success is rarely about finding the right solution to a problem — it’s more about finding the right problem to solve.

Getting straight A’s requires conformity. Having an influential career demands originality. In a study of students who graduated at the top of their class, the education researcher Karen Arnold found that although they usually had successful careers, they rarely reached the upper echelons. “Valedictorians aren’t likely to be the future’s visionaries,” Dr. Arnold explained. “They typically settle into the system instead of shaking it up.”

If your goal is to graduate without a blemish on your transcript, you end up taking easier classes and staying within your comfort zone. If you’re willing to tolerate the occasional B, you can learn to program in Python while struggling to decipher “Finnegans Wake.” You gain experience coping with failures and setbacks, which builds resilience.

Employers: Make it clear you value skills over straight A’s. Some recruiters are already on board: In a 2003 study of over 500 job postings, nearly 15 percent of recruiters actively selected against students with high G.P.A.s (perhaps questioning their priorities and life skills), while more than 40 percent put no weight on grades in initial screening. Straight-A students: Recognize that underachieving in school can prepare you to overachieve in life. So maybe it’s time to apply your grit to a new goal — getting at least one B before you graduate.

Curated Insights 2018.12.14

The Facebook Fed

I think a good analogy is that Facebook is the Federal Reserve of web publishing. It can turn its dial and blast millions of visitors to numerous publishers, allowing everyone to have more eyeballs to sell to and more rising traffic numbers with which to attract investment. Facebook turning on the traffic fire hose is like loose monetary policy that stimulates the economy for everyone.

Of course, Facebook could tighten policy, pulling traffic (liquidity) and leaving weaker players parched. When the Fed tightens policy, shaky borrowers who depend on ample lending are hit hardest. When Facebook tightens policy, second-tier publishers that totally rely on Facebook are hit hardest.

Digital divide is wider than we think, study says

Over all, Microsoft concluded that 162.8 million people do not use the internet at broadband speeds, while the F.C.C. says broadband is not available to 24.7 million Americans. The discrepancy is particularly stark in rural areas. In Ferry County, for example, Microsoft estimates that only 2 percent of people use broadband service, versus the 100 percent the federal government says have access to the service.

The hardest problem in finance

But here’s the catch, this table assumes you get this rate of return year after year after year. The real world is not so accommodating. The table above shows that you can earn 4% a year for 34 years before running out of money, but let’s look at what happens if a nasty bear market were to arrive as soon as you retire. The chart below shows one way in which an investor can arrive at a 4% CAGR over a 30 year period.

The danger of assuming compound annual growth rates when making long term projections can be seen in the chart below. The black line shows that spending a constant $40k annually, using the returns from the previous chart, an investor would run out of money in the 19th year. Spending 4% and assuming a 2% inflation rate, a more realistic assumption, an investor would run out of money in just 15 years. Side note, if the returns above were to happen in reverse, in other words the bear market comes at the end of the period, an investor with the same spending would be left with $1.3 million.

Curated Insights 2018.12.07

The money and math behind our newsletter headlines

That means that the winning headline results in 4,620 incremental opens per newsletter. Over the course of a month, that is 115,000+ incremental opens. If we apply a conservative estimate based on some of our data from 2015 of how many of those incremental opens try the CB Insights free product (0.09%) and that only 1% of those trials buy a subscription, we’re looking at nearly $625,000 of incremental revenue.

Facebook is undervalued

The result of billions of readers providing free content for each other is a massive network effect and a business that can serve each incremental ad at a very low marginal cost. The company’s 85% gross margin leaves a lot of meat on the bone that the company can spend on hiring new engineers and developing new technology that is needed to maintain and grow the business. But even after paying the engineers, growing the headcount, spending over $9 billion on R&D, and paying the tax bill, shareholders are still left with around 40 cents of profit for each dollar of revenue. Facebook reinvests these earnings into data centers, technology, and the occasional acquisition, but even so, the cash in the till has been piling up, and now exceeds $40 billion and counting ($14 per share). Facebook has spent $10 billion on buybacks in the last year, and this will likely accelerate in the coming years.

Facebook’s business is still growing fast (33% revenue growth last quarter), and despite the company’s large size, the runway is still long. Benedict Evans, a partner at the VC firm Andreessen Horowitz, recently pointed out that roughly $1 trillion is spent each year by businesses who are trying to reach customers who are asking the question: “What should I buy?”. Facebook has roughly 5% of this market, which I think is a share that will inevitably rise over the coming years as advertising dollars continue to shift from traditional media to the much higher ROI advertising platforms like Google, Facebook, and Amazon.

A strong network leads to Facebook grabbing a much greater share of the $1 trillion (and growing) global ad/marketing industry. Revenue from ads alone could easily be twice the current size, which means that Facebook is a $100 billion business before counting any of the possible upside from payments, messaging, or any of the other potential business lines that the company could develop on the back of its large network effect.


Accelerator demand should grow ten-fold according to top 500 supercomputers

If the data center market as a whole were to follow in the footsteps of HPC, accelerators could displace Intel CPUs as the workhorse of enterprise computing.

If processors were to grow to 50% of total server component costs, with accelerators capturing 70% of processors, as is the case in the HPC market today, the market for accelerators could scale from $4 billion1 today to $24 billion. Currently, the size of the computer server market is $80 billion globally, according to IDC. If manufacturers take a 15% cut, then $68 billion accrues to component suppliers such as Intel, Micron, and Nvidia. Today we estimate that processors account for roughly 35% of server component costs, the balance allocated to memory, storage, and networking. With HPC and AI, servers are evolving toward higher compute configurations, pushing processors and accelerators to 80% of total component costs.

As shown by the latest IDC data above, the server market as a whole continues to grow, driven primarily by higher average selling prices (ASPs). With denser configurations, server ASPs could climb from today’s price of $8,000 to $10,000. Even at the current volume of 10 million units per year, that would grow the server market 25% from $80 to $100 billion, pushing the accelerator total addressable market to $35 billion.

Today, the accelerator market is dominated by Nvidia’s data center products, which we estimate will generate $3.2 billion of revenue in 2018. Nvidia estimates that its data center business has a total addressable market of $50 billion—a figure larger than Intel’s data center business, which seems difficult to reconcile. That said, as the latest data from IDC and Top 500 shows, the server market is undergoing a fundamental shift. The decay of Moore’s Law has increased total server spend as compute intensive workloads migrate from CPUs to specialized processors. If CPUs and accelerators were to swap places in the data center, as they did in HPC, the accelerator market would approach Nvidia’s estimate, implying a 10x increase over the next five to ten years.

Artificial intelligence is giving rise to fake fingerprints. Here’s why you should be worried

In the new paper, the researchers used neural networks—the foundational software for data training—to create convincing looking digital fingerprints that performed even better than the images used in the earlier study. Not only did the fake fingerprints look real, they contained hidden properties undetectable by the human eye that could confuse some fingerprint scanners.

Julian Togelius, one of the paper’s authors and an NYU associate computer science professor, said the team created the fake fingerprints, dubbed DeepMasterPrints, using a variant of neural network technology called “generative adversarial networks (GANs),” which he said “have taken the AI world by storm for the last two years.”

Daniel Kahneman: Your intuition is wrong, unless these 3 conditions are met

There are three conditions that need to be met in order to trust one’s intuition. The first is that there has to be some regularity in the world that someone can pick up and learn. “So, chess players certainly have it. Married people certainly have it,” Kahnemen explained. However, he added, people who pick stocks in the stock market do not have it. “Because, the stock market is not sufficiently regular to support developing that kind of expert intuition,” he explained. The second condition for accurate intuition is “a lot of practice,” according to Kahneman. And the third condition is immediate feedback. Kahneman said that “you have to know almost immediately whether you got it right or got it wrong.”

The 80-hour workweek

Every year on Wall Street, first-year analysts who were once courted by college recruiters with lines about how they would be given meaningful, creative work and the chance to learn from top executives are shown the harsh truth: they are Excel grunts whose work is often meaningless not just in the cosmic sense, but in the sense of being seen by nobody and utilized for no productive purpose. Some of the hundred-page pitch books analysts spend their late-night hours fact-checking in painstaking detail are simply thrown away after being given a quick skim by a client. In other cases, the client doesn’t read the deliverables at all, and the analysts’ work is literally garbage.

Most days that winter, he had worked the “banker nine-to-five”—getting to work at 9:00 a.m. and staying until 5:00 a.m. the next morning, at which point he’d trudge back to his Murray Hill apartment, sleep for three or four hours, and do it all over again. Even the money he was making wasn’t cheering him up. Ricardo had once made the mistake of calculating what his hourly wage would be, given his salary and a conservative estimate of his year-end bonus. The result—which he estimated was something like $16 an hour, after taxes—was much more than minimum wage, but not nearly enough for Ricardo to be able to justify the punishment he’d been taking.

The question of in office hours versus out of office hours is a good one in thinking about a career in the finance industry. I knew basically nothing when I came into finance so a massive amount of my learning was done on my own after hours.