American political discourse undoubtedly peaked in 1952 when a Mississippi legislator with the wonderfully Dickensian name Noah “Soggy” Sweat, delivered the “If-By-Whiskey” speech. Reading the WeWork S-1, it kept coming to mind.
If by WeWork you mean the capital inferno; the overhyped infernal machine that will only get shareholder funds for as long as Vision Fund I needs to show a high enough IRR to close Vision Fund II; the bizarre multidimensional Flying Spaghetti Monster of nested C-Corps, LLCs, and LPs, all rife for abuse by its controlling shareholder; the commercial real estate operator whose delusions of tech unicorn grandeur have justified a temporary valuation premium and a permanently high level of overhead, I am against it.
If by WeWork you mean a company that can invest in the multi trillion-dollar commercial real estate market and consistently generate better unit revenues and faster breakevens than competitors, while continuously extending its offering into residential real estate, health and wellness, employee benefits, and more; a company whose CEO is a champion salesman and a stellar dealmaker, who can ensure that your high multiple as a buyer is your low cost of capital as a long-term shareholder; then I am for it.
That is my stand. I will not retreat from it. I will not compromise.
(Full disclosure: I have a dinner bet with a friend on where WeWork will price its IPO. Other than that, no interest besides curiosity)
The numbers are certainly impressive. Who wouldn’t kill for a growth chart like this?
The revenue chart looks basically the same, though, and is in the same ballpark: in Q2 ’19, WeWork’s revenue was $807m, up 91.4% Y/Y. Adjusted EBITDA was negative $524m, up… 104% Y/Y.
WeWork is an unusually Hegelian situation. Thesis: it’s a tech company that happens to have some real estate on the balance sheet, and thus deserves a premium multiple. Antithesis: it’s a real estate company that’s trading high overhead for a high multiple in order to gull VCs with more cash than ideas, real estate investors with tech company FOMO, and RobinHood investors with margin accounts. Synthesis: it’s a full-stack tech company that monetizes a high-margin, scalable software and marketing business by earning high returns on commercial real estate, a $29 trillion market.
WeWork as a Real Estate Play
Every CEO needs to be good at sales. The CEO’s job is to convince employees, investors, big customers, and potentially regulators that the company is a good one. The relative importance of sales varies based on what kind of company it is. At one end of the spectrum, there are hard tech companies that sell to just a few big customers. Paradoxically, a company that sells to lots of tiny buyers can have a poor salesman as CEO — marketing is sales for people who don’t like handshakes and eye contact. (I wrote about this a bit in The Socially Awkward Social Network Paradox.)
Real estate may be the most salesly of industries. To get a landlord to lease you a floor or a building, you need to make a convincing case that you can fill it with paying customers so you don’t default. Then you need to actually fill it with those paying customers.
Adam Neumann is, by all accounts, a phenomenal salesman. I’ve heard more than one anecdote about Neumann meeting with skeptical public- and private-market investors, and convincing them all to be WeWork bulls. This may be because the S-1 somehow understates the economics of the business, or it may be because Neumann himself is good at compelling exaggeration.
If it’s the latter, this implies that WeWork is overvalued — but it doesn’t mean WeWork will fail. After all, the converse of “overvalued” is “has a low cost of capital,” so the earlier you buy, the happier you’ll be. In fact, you can tell a reasonable story that WeWork has economies of scale that really kick in when it’s 5x-10x bigger, and it’s more likely to get there if the CEO can sell effectively. If he can convince equity investors to overpay, he can convince landlords to offer generous terms, so as long as the net present Hype Discount on leases exceeds the net present Hype Premium on the asset, it’s a good deal.
This is, naturally, an incredibly nerve-wracking position for an investor to be in: you’re consciously choosing to be the lesser fool whose greater fool is… a lessor fool.
WeWork’s status as a real estate play explains one of the other issues that’s gotten a ton of media attention: the numerous related-party transactions between Adam Neumann, his family, and his company. (It is very much “his” company, given that Neumann has super-voting stock.)
It does look pretty weird that Neumann has been borrowing money from WeWork to buy buildings in order to lease them to — WeWork. But it makes sense from a financial engineering and cultural perspective. On the financial engineering side, WeWork is valued at a very healthy premium to its lease revenues.
So Neumann can efficiently transform cash-on-hand into a higher equity valuation. Essentially he’s keeping the low-multiple assets on his personal balance sheet and putting a high-multiple revenue stream from them onto WeWork’s P&L.
There’s a sense in which this is the entire point of finance. The world is full of assets that throw of streams of cash flows. Sometimes those assets are put to inefficient uses, and the market pays people to put them to better uses. Neumann is acting as a market-maker here, taking some personal balance sheet risk in order to maximize the aggregate value of the assets in question.
There’s also a narrower cultural divide. I haven’t spent a ton of time in the real estate sector, but I get the impression this kind of self-dealing is, well, pretty standard. At least the direction, if not the magnitude. WeWork has hyped itself and priced itself as a tech company; I’m sure many of the journalists covering it have, at some point, worked out of a WeWork.
A tech investor looking at commercial real estate’s corporate governance and tax structure is like a Mormon missionary who gets assigned to Bangkok’s red light district: “Why did you… how did you… and you like it?”
Big companies do some things that also sound like you can only pay for them in Patpong— Double Irish, Dutch Sandwich, etc. But it’s usually, deliberately, sub rosa. The closest real-world example that didn’t rise to the level of criminal behavior is probably Nick Denton licensing software from an offshore corporation to Gawker, for an amount that always, coincidentally, wiped out Gawker’s taxable profits. But Denton is now less famous for his entrepreneurship than for his blindness to which lines one ought not to step over. So he illustrates the tech world’s aversion to complex legal structures.
WeWork’s legal structure is certainly complex. The company is basically a big pile of spaghettilike interlocking subsidiaries, which have slightly different economics and slightly variable ownership. For example, some overseas WeWorks are joint ventures. WeWork India operates under a property management model, rather than an ownership model. And then there’s ARK, which is basically a captive real estate private equity fund with We as the largest general partner, which buys buildings and leases them to WeWork. ARK has raised almost $3bn, 93% of which did not come from WeWork.
It’s a tribute to commercial real estate’s use-every-part-of-the-buffalo ethos. Some cash flows are lumpy, and they go into a pocket that tolerates lumpiness. Some are smooth, and go somewhere else. Some deals need a local front, so they have one.
And then we get to the weird stuff. Like everyone else, I call WeWork “WeWork,” even though it’s officially “The We Company.” It licensed “We,” a two-letter word, for six million dollars (edit: six million total, not six million per year), from… its CEO. I genuinely don’t understand why. If he needs six million dollars, he could ask the board for a bonus plan, or some more restricted stock — that should be an easier negotiation, since it’s not going to show up as a weird disclosure in the S-1.
Back in the wild days of the early 90s, Bill Gates got interested in digitizing art and building a “smart home” with screens displaying famous paintings. This seemed pretty cool at the time, but there was a problem: the project was expensive, and probably not financially viable. But it might be lucrative. If Microsoft did it and it failed, it would be a waste of shareholder money, but if Bill Gates did it and it succeeded, it would look like he had kept his best products to himself. Their solution was for Gates to form a separate company and give Microsoft an option to buy it.
This is not the Neumann family approach. They’re constantly mixing business and family, swapping assets with WeWork, working odd (and oddly well-paid) jobs for the company, etc. While this looks bad, it’s hard for it to be any worse than it looks: every explanation I’ve seen is that Neumann is looting the company. A slightly better explanation is that this is the first real company he’s worked for, so his professional network skews strongly to relatives. Before We, he briefly ran a similar coworking company, and before that he tried to sell toddler clothes and served as a captain in the Israeli navy. (Not at the same time.)
Recruiting people is hard, and when you’re growing at over 100% a year, it’s the biggest bottleneck. So the less ethically challenged view of WeWork’s nepotism is that it was a shortcut to scaling. Not ideal — one reason VCs like to fund people from brand-name companies is that those people can scale by poaching — but not necessarily a catastrophe.
WeWork as a Tech and Branding Play
WeWork, in accordance with securities laws, discloses a lot about their real estate operations. Lease costs, property counts, capex, depreciation, the cost of building out a new office. But the pictures and the commentary tell a totally different story: they look like every single tech company’s “We just raised our biggest round ever” HQ. In fact, many of them are basically that: WeWork provides office space to Salesforce, Slack, Airbnb, Oscar, Glassdoor, Dropbox, Bytedance, and 38% of the Fortune 500.
WeWork doesn’t exactly emphasize the space: they emphasize the service — well-designed offices, app-based conference room booking, “community,” and the like. The document seems to have been written by someone who regularly reads Stratechery, and mentions the term “demand aggregation” six separate times. The S-1 is also full of deep thoughts on the We brand. The document is in fact dedicated “to the energy of we, greater than any one of us but inside each of us.”
One version of the WeWork story — the one that justified a $47 billion valuation on last year’s $1.8bn in sales — is that they are a software company with a great brand.
Can you build a brand on something as commoditized as office space? People care about their buildings, but they don’t generally have a favorite landlord.
There are precedents for taking a commodity product and turning it into a brand. Fiji water is just water; vodkas are chemically identical; the Mast Brothers marked up their chocolate because they had a good story, not because the chocolate was anything special. And in finance, brokers can get nice commissions by selling equity-linked notes to the vast demographic of people who don’t say “Wait, isn’t that just SPY plus a put and a markup?”
If WeWork builds a globally recognized brand in commercial real estate, becomes synonymous with offices, and backs it up with a software stack that continuously improves the process of getting and using space, it’s a huge opportunity.
But if that’s the story, what’s the deal with $33.9bn in lease commitments?
Full-Stack Real Estate Tech
The trick with a new technology is not just to solve a problem, but to capture the value. It’s not always obvious how to do that. Software and M&A advisory were both, at one point, free services used to attract customers for the good stuff (mainframes and underwriting, respectively).
A full-stack approach makes sense if you know you have a good idea, but don’t know how the economics will shake out. Willis Carrier made good money from the air conditioner, but the city-state of Singapore made a whole lot more. If Vision Fund had existed early on, Carrier might have had 10% of its assets in factories and inventory, and 90% in commercial real estate in Singapore, Atlanta, and Dubai.
The Big Idea in WeWork’s case is that real estate should be a lot more liquid. Demand changes on shorter timescales than the length of a typical office lease.
The real estate yield curve is naturally inverted because there are fixed transaction costs. If you want to live somewhere cheaply, buy or sign a long lease. If you want to spend a lot per square foot per month, rent a Breather or visit a Love Hotel. Even coffee shops and bars are basically in the real estate business: they sign a multi-year lease on a venue and rent out individual tables for the cost of a beverage. It’s not a coincidence that the two kinds of dining venues best modeled as real estate plays sell diuretic drinks, which prevent people from camping out too long at one table during peak demand periods.
There’s a broad economic argument that something like WeWork makes sense: when more real estate transactions move online, transaction costs decline, and since transaction costs are a bigger share of the rent premium for short-term rent than longer-term rent, the first company to insert itself into that gap should harvest some of the economic gains. The natural question is: why shouldn’t this company be a pure software player, and keep the real estate on somebody else’s balance sheet? If WeWork just made a market in hot desks and monthly leases on behalf of landlords, taking a 5–10% commission a la Airbnb, it would have a cleaner economic model. There are two reasons this isn’t how it happened:
- Real estate is hard.
- Capital is cheap.
Let’s take the Airbnb comparison: the core of the Airbnb offering is that you’re not staying in some soulless hotel chain — you’re a guest in someone’s home. You Live Like a Local. That model works because people like having nice homes, and they like being good guests. But most people take a lot less pride in their offices. Sure, a swanky office is worth bragging rights, but there are a lot more people who tolerate a merely tolerable office than who tolerate a merely tolerable home. Airbnb also gets to take advantage of the person-to-person nature of the product: a bad review is not just a marketing problem, it’s personal. Whereas I’ve never had a landlord who seemed personally bummed out when tenants complained.
The easiest way for WeWork to maintain quality standards is to impose them top-down by controlling the building. This is also a great marketing move: they can create a WeWork style, impose it everywhere, and tweak it for local markets. (The WeWork S-1 discloses that they have 500 designers and architects, constantly cranking out new features.) A long time ago, Joel Spolsky pointed out that supply chains naturally alternate between fragmented and monopolistic layers. Lots of companies sell corn syrup, water, and aluminum, but only one company makes Coca-Cola. Lots of people make short videos, but there’s just one YouTube.
When transaction costs and financing constraints change, the natural border between the monopoly and the commodity shifts.
WeWork talks about their software as a free ancillary product that helps them sell more real estate:
Our enterprise technology solutions enable members to make informed decisions across the entirety of their real estate portfolio while improving the workplace experience for employees and guests. Our foundational and free product is workplace analytics, which provides a deeper understanding of space utilization, future space needs and associated costs for enterprises.
There is a long and glorious history of tech companies commoditizing the complement. If your ad product gets the best ROI, you should make analytics free so everyone knows. A commoditized complement makes sense if it helps sell something valuable. But why should WeWork get good economics from raising capital to buy and resell real estate?
WeWork owns real estate because there’s enormous global demand for places to put capital, and both real estate and tech companies offer supply to match this demand. Landlords don’t want to be in the business of selling one hot desk at a time for one-month increments, but they don’t want to offer a bunch of space to a third party for a commission and then have that third party fail to fill any of it. If WeWork didn’t have access to cheap capital, they’d have a far slower growth rate (and, given that pre-opening, sales & marketing, and growth & market development expense collectively amount to 67% of revenues, a far higher profit margin).
All this is not to say that WeWork is a pure bubble creation — many great companies start out by taking advantage of the most recent economic dislocation. YouTube benefited from cheap broadband due to overinvestment in the 90s; the car industry got big because of a pervasive US oil glut in the 20s and 30s; the entire East Asian economic boom was partly catalyzed by cheap container shipping. Part of being an entrepreneur is being able to benefit from collectively poor decisions, and right now a series of bad policy decisions, harmful cultural shifts, and bad luck have created a savings glut that’s desperately seeking companies who can put billions of dollars to work.
Really, the only flaw in the full-stack real estate company thesis is that the numbers don’t bear it out. There are few signs anywhere that WeWork is starting to achieve better economies of scale. I broke out their unit costs by comparing them to workstation growth (for building opening costs) and member growth (for sales and marketing). Unit costs are consistently trending higher. Meanwhile, revenue per member is declining, and revenue per workstation is doing worse due to lower occupancy.
They do explain the lower revenue per seat by noting that they’re expanding in lower-priced markets, but I’d expect those markets to have lower unit costs, too. They don’t call this out specifically, but either Q2’s membership and workstation growth were unusually light, or Q2 ‘18’s were very strong — I would look at Q1’s incremental margins as more representative unless WeWork gives us a specific reason to think they’ve decided to dramatically slow their growth.
Ultimately, the full-stack model works under a simple assumption: if WeWork has a better brand, it can acquire tenants more cheaply than anyone else, and if has more tenants hooked up to a software stack that tracks their every move, it can relentlessly optimize to lower the number of square feet needed per employee, giving WeWork a cost advantage. People joke about this stuff, but it’s true: WeWork isn’t just tailoring offices to demand, it’s Taylorizing them, too. WeWork boldly claims a 57% lower cost per employee than traditional office buildings, and the only way they’re getting that is to pay a bit below market per square foot and put way more people in each of those square feet.
CB Insights claims that WeWork has fifty square feet per worker, compared to 250 in a standard office. WeWork has disputed the stat, and they don’t give exact square footage in their S-1. They do, however, say that their pipeline of potential buildings has 40 million square feet with room for 724,000 workstations. That’s 55.25 square feet apiece. It sounds completely insane that WeWork would be almost 400% more efficient than the average office management company, but it’s not out of the question that WeWork knows exactly how much space you need for every perk you’d want, so for anyone who needs an office with a particular amenity in a specific location, they’re the natural low-cost provider.
That’s the hypothetical long-term steady state. To get there, WeWork has to invest upfront to scale up, both to get new properties — adding to its sample size so it can further optimize everything — and to hire the engineers necessary to implement it. As of yet, there’s no evidence that the scalable part of their business is scaling any faster than the fixed costs, at least in the aggregate. We’ll see, though. It will be interesting to watch how their margins develop when growth slows from ~100% annualized to something more modest and sustainable, like 50%. A 100% growth rate means that half of your properties have been open for less than a year, and WeWork says it takes two years for occupancy to stabilize. So, for now, they’re both spending heavily on theoretically one-time costs and diluting their long-term unit economics.
Fortunately, WeWork has given us some disclosure around their underlying profitability. Unfortunately, it’s imperfect.
There’s this folk theory about companies and their non-GAAP metrics: cynical management teams are snookering investors by tricking them to ignoring costs. The actual story is more complicated: most investors I know focus on the non-GAAP metrics everyone complains about, because those metrics are more representative.
What a company will typically do when they report a non-GAAP metric is to a) strip out a bunch of one-time stuff, like legal fees; strip out some non-cash expenses, like depreciation and amortization; and showcase what the company’s unit economics look like if you ignore fixed costs.
This is also what investors do. If a company is growing at 50% a year, and its net profit margin goes from -100% to -50% to -20%, what’s clearly going on is that there are some costs that rise linearly with the revenue growth and some that rise much more slowly. Over the long-term, the company’s overall economics are dominated by those incremental margins. It costs a few billion dollars to build a car factory, but orders of magnitude less to build one more car.
WeWork got razzed in the press for highlighting a metric they called “community-adjusted EBITDA,” which stripped out most of their corporate overhead costs. But if you think WeWork will expand from 500+ buildings to 10,000+, the unit economics will be the dominant factor in what their long-run economics look like. Sure, the home office needs more admins, middle-managers, architects, and software developers at a larger scale, but not in proportion to the company’s size.
WeWork opens their unit economics discussion with a graph:
By contrast, I found WeWork’s quantitative unit economics presentation pretty informative: they showed that on average, the contribution margin from a new building is about 10%, which they hope to get up to 30% long-term. I was a bit triggered by their decision to present some metrics that exclude straight-line lease costs. That gets into some accounting arcana: the way many leases are structured is that in between when WeWork takes possession of a property and when they can actually start leasing to tenants, they don’t pay rent. (They do spend money, on building out the space.) Later on, leases have automatic annual rent increases. (You can see this in their disclosure on page 122, where they show their current lease obligations stepping up by 3.9%% annualized from 2020 through 2023.)
Arguably, cash lease costs are representative of the unit economics — as the lease steps up, WeWork will charge more, due to inflation if nothing else. But ignoring the noncash charge from the initial rent-free period is dubious, because it definitely represents an economic cost. The landlord prices the entire lease to recoup that rent-free period; you can’t argue that the timing of cash flows here meaningfully reflects economic reality.
(And by their own argument, they’re setting themselves up for a lower future improvement in unit revenue. They’ve talked up their increasing efficiency in opening offices: they’ve gotten the time from signing to opening down to four to six months, compared to a typical nine month rent-deferred period. If they tighten that up further, their unit economics will only improve if they’ve been recognizing the implied cost of the rent-free period.)
The way WeWork wants you to model them is like a SaaS company with some irrelevant balance-sheet details: they spend heavily on R&D to build a product, then they spend heavily on sales and marketing to get it to customers. But once they land a customer, they can expand from there, in terms of seats and pricing. They’re certainly expanding seats — WeWork cites a 119% net member retention rate (i.e. for every hundred members they had a year ago, the companies those people worked for, including the ones that left WeWork, have 119 seats a year later).
The revenue side is more ambiguous: revenue per seat declined 7.6% in the last quarter, but, as they note, they’re growing in countries with lower price points. If you start out in New York and expand to SF, it’s hard to keep your rent per square foot high when you add Warsaw. They could have, but did not, provide a comp revenue per seat number, which would have been informative. They do note that contribution margins would have been 3% higher without China. Expect this decline to continue:
The majority of our revenue from locations in the United States was generated from our locations in the greater New York City, San Francisco, Los Angeles, Seattle, Washington, D.C. and Boston markets. A majority of our locations in the United Kingdom are in London.
While the accounting is basically okay, I’m harboring increasing doubts about the high-margin dollar-expansion model. Here’s the problem: lots of growth companies are doing it, and their revenue is some other growth company’s COGS.
If Twilio’s revenue from a given company doubles when that company’s revenue doubles, that cuts into the company’s margins. If Dropbox is growing because it has a piece of Twilio’s growth, and Twilio is growing because it gets a piece of Lyft’s growth, and Lyft is growing because Salesforce reps need more taxi rides, and WeWork is growing because every one of those companies is adding headcount faster than they can sign long-term leases on a new HQ, then every software company’s long-term margin story is contingent on the breakdown of everyone else’s long-term revenue story, or vice-versa.
The denouement of this probably won’t be a collapse of every model. What will probably happen is that some of these products will turn out basically the way the model says, and some will be relentlessly commoditized as their biggest customers get more pricing power. Growing along with your customers is a great problem to have, until one of your customers decides that your 90% gross margin is their opportunity for an easy few hundred basis points of their margin.
The IPO and the Masa Factor
One of WeWork’s biggest investors is the SoftBank Vision Fund. Vision Fund I, to be exact. Vision Fund II is in the process of closing investors, at least one of whom might be reluctant to cut a check. WeWork’s IPO performance is the swing factor in Vision I’s IRR, which is the biggest factor in Vision II.
Masa Son has a lot at stake here, and a lot of cash, not to mention interests in major companies. If it looks like WeWork will price at a low valuation, he might take action — maybe he’ll suggest to portfolio companies that they sign leases at underperforming WeWork offices, or commit to doing so in the future. Maybe he’ll commit to not selling his stock — or lend it out to short sellers, and call it in at an opportune moment.
Masa is a nonlinear thinker, so I don’t know what he’ll decide. Maybe he doesn’t, either; the guy seems pretty spontaneous. But we do know his means (tens of billions of dollars, equity stakes in numerous growth companies) and his motive (make WeWork go up, or at least not go down too much, until Vision II closes). And the IPO represents an opportunity.
I read the WeWork prospectus. A lot of people have skimmed it. So far, it’s changed nobody’s mind. Everyone knew WeWork was losing gobs of money, and the vast majority of people are skeptical.
I think a lot about Warren Buffett’s poker quote: “If you’ve been in the game 30 minutes and you don’t know who the patsy is, you’re the patsy.” Are WeWork’s institutional investors the patsys? Adam Neumann is clearly an excellent salesman and an opportunist par excellence. Not the kind of person you want to trust with your money, perhaps, but not someone you want to bet against blindly, either.