Most Large Tech Companies Are Systematically Undervalued

If you guys were the inventors of Facebook, you’d have invented Facebook. - Mark Zuckerberg, as imagined by Aaron Sorkin As a general rule, a large tech company will be chronically undervalued if 1) it produces reasonable free cash flow in its core business, and 2) investors are rolling

If you guys were the inventors of Facebook, you’d have invented Facebook.

- Mark Zuckerberg, as imagined by Aaron Sorkin

As a general rule, a large tech company will be chronically undervalued if 1) it produces reasonable free cash flow in its core business, and 2) investors are rolling their eyes at the CEO’s M&A and R&D priorities. This is a lazy heuristic: it fits Alphabet, Facebook, Uber, Netflix, and Airbnb; it doesn’t apply to Priceline or Microsoft (everyone loves their capital allocation right now) or to Snapchat (the free cash flow isn’t there yet).

This isn’t really a financial argument: it’s a social argument. Good ideas are hard to understand, even if you’re in the business of understanding good ideas. Good companies still get rejected by 90% of the VCs who could potentially invest at a given valuation, so even someone in the 80th percentile among professional venture capitalists will probably be wrong about how valuable a company can get. And they’re not just generically hard to understand. Each company has its own quirks, and the one investor in a hundred who understands what Facebook will be is not necessarily the one investor who knows what Uber will be.

(In fact, that one investor ended up investing in Lyft.)

The one person who definitely understands the company is the CEO. Historically, plenty of smart people thought Reed Hastings was too optimistic about on-demand streaming taking share from linear TV, or about a tech company’s ability to produce better shows that Hollywood. Historically, all of these people were wrong. There are a tiny number of people who saw in 2012 what Netflix could be in 2017, and they’re the people running Netflix.

Large cap + cash flow filters sort out the potentially transformative companies from the once-and-future transformative companies: a good idea that will probably throw off free cash in the future is evidence of a good CEO, but a good idea that has actually thrown off cash flow — which is now being invested in fiber, or virtual reality, or self-driving cars, or flying cars — is proof.

N Is Small

Technology companies succeed because they do a million minor things right and get two or three really big things right. This is most visible in their financials: it’s extremely rare for a growing tech company to have two or three sources of revenue that are similar in size. More common is the situation at Facebook or Google, where 85%+ of their revenue and 100% or more of their profits comes from just one line of business.

Google chose to treat search as a core product rather than a loss-leader for display ads, at a time when this choice was objectively crazy. They cycled through a few monetization options before they settled on AdWords, and since then ~90% of their growth in market cap has come from 1) secular trends towards more Internet use and more search, 2) continuous improvement in search, and 3) raising their ad load from “one or two ads, way off to the side” to “for a typical monetizable query on a typical screen size, maybe one thing you can tap or click that won’t generate revenue.”

It’s comparatively easy to evaluate a company’s execution track record: look at the things every tech company has to do: hire engineers, actually sell stuff, continuously raise prices and ad load without scaring off advertisers or customers — in short, tech execution just means avoiding unpleasant quarterly earnings reports. But it’s hard to tell whether a company’s big calls were skill or luck. Even Facebook has a sample size of four or so company-making decisions:

  1. Your primary online identity will always be tied to your real name and your real-life identity.
  2. Universities form a natural, preexisting network; instead of launching everywhere at once, you can launch to 1k-20k students at a time. (Early on, this also meant that Facebook could defer tough scaling questions. They didn’t have to learn how to run a million-user site; they could just run five hundred copies of the site on five hundred separate databases.)
  3. Time and dollars will move from desktop to mobile, and mobile ads will eventually have a higher CPM because they’re more precisely targeted. In 2012, investors were worried that mobile would blindside Facebook; Facebook was talking about the importance of mobile in 2006, nine months before the iPhone.
  4. A billion dollars is a ridiculously low price for the best photo-sharing app.

Facebook could have gotten any one of these wrong. There’s a near possible world where Mark Zuckerberg is just one more case study of peaking within five years of getting accepted to Harvard. But the fact that Facebook got all of these right, before anyone else was even asking the right questions, is pretty good evidence that whatever mysterious projects they’re dumping cash into now are disproportionately likely to win.

Given the lag time — six years between deciding mobile is a big deal and Facebook’s first dollar of mobile revenue — you’ll never have a big enough sample size to know if someone just got lucky. The best you can do is notice that lucky strategic choices and good execution correlate with each other, and lucky choices tend to be serially correlated, too. Like good traders, good technologists tend to have positive skew: many small mistakes, paid for by a few well-exploited successes.


There are several exceptions to this rule, mostly falling into the category of companies that do things with computers but aren’t actually tech companies. The two biggest categories of fake tech companies are sales companies that use computers and marketing companies that use computers.

Mobile gaming, online travel, and online dating are all marketing businesses: they’re as much a slave to customer acquisition cost as their are to customer lifetime value, and customer acquisition cost is determined by monopolistic channels. If you’re Zynga, you can never earn truly superior returns if you’re paying the Apple/Android tax on all of your monetization and the Facebook/Twitter tax on all of your user acquisition. Tech companies that depend on marketing are ceding data and knowledge to the platforms they market on, and the advertiser’s margin is the ad seller’s TAM.

Enterprise SaaS companies face a different problem. Their economics are driven by dollar retention: all else being equal, if a million-dollar contract this year means an expected $1.1m next year, it’s easy to justify a high multiple on an unprofitable business. But growing those contracts tends to mean building your product around what big customers ask for; the bigger the customer, the more they can ask. So even though the margins start out looking like software, they end up looking like consulting.

The best exception to that rule is Twilio: if your upsells start with API calls and not phone calls, you get another turn on your price/sales multiple.

What’s Next?

I’ve spent most my career, both at a hedge fund and as a consultant for hedge funds, using data to make more informed investing decisions. Knowing that a company has a high probability to succeed over the next five years is valuable, but if you want that sweeet 2 & 20, you also want to know whether they’ll be up or down this quarter. When you zoom out, the long-term chart for Facebook or Netflix looks pretty good: since its IPO, Facebook has delivered 28% annual returns. Since their 2002 IPO, NFLX has returned 39% compounded. But Facebook had a brutal drawdown, from $42 to $18 in its first six months. And NFLX dropped from $5.00 to $1.30, and then from $40 to $10, on the way to hitting their current $144. Every downdraft in that long-term chart is a short story about some smart analyst feeling vindicated, but only after they got fired.

Being able to slice up differentiated datasets is still a great way to generate alpha, and large-cap tech companies are both liquid and volatile: there is serious money to be made having an advantage predicting Facebook or Google’s next quarter. But for each of these companies, the most valuable datapoints are the ones that exist in the CEO’s head.