[Note: I originally wrote the bulk of this article as an unpublished memo about 18 months ago. I have updated it to include new information. The views and opinions expressed in this article are mine and do not necessarily reflect the views of my clients.]
The big news this past week was that Coinbase acquired Earn.com (née 21.co, née 21e6 LLC). According to Recode, the offer “was slightly more than $100 million” but also lower than Earn.com’s most recent valuation (in 2015) which was $310 million.
From the current coverage, it is unclear what the revenue for any of the products or services for Earn.com was. Instead most stories have focused on one specific aspect: the current Earn.com CEO, Balaji Srinivasan, will join Coinbase as the CTO.
There have been a lot of questions around why Coinbase would purchase a company that seemed to have poor product-market fit with unknown KPIs. This post will look into several areas for answers.
Taking a step back
Following the official acquisition announcement from Coinbase, Srinivasan published a self-congratulatory Medium post that basically paints him as the savior of 21.co: that it was the previous management that were bad and he came in and turned it all around.1
His revisionism arguably whitewashed what happened, so let’s dive into a little bit of the company’s history.
In May 2013, 21inc (formerly 21e6 LLC) was co-founded by five men including Balaji Srinivasan. According to a story from Nathaniel Popper:
The company was also structured as an limited liability company, rather than the C Corp typical of startups, so that people could invest with their own money.
Why is that important to some investors?
According to Popper:
The 21e6 investment was attractive in part because venture capital firms generally felt that they couldn’t buy Bitcoins directly. 21e6, on the other hand, offered to pay its investors back with Bitcoin dividends, allowing the firm to get Bitcoins without buying them outright.
What does this mean?
Venture funds often have clauses restricting their partners from investing in asset classes that may be seen as a conflict of interest or something that could reduce the firm’s reputation (e.g., cannabis startups). In this case, cryptocurrencies may be seen as a direct speculative bet on a commodity or foreign exchange which could be prohibited by an investment funds by-laws.2
Altogether the 21e6 team, over three separate rounds, raised approximately $116 – 125 million – which at the time was more money than any other cryptocurrency-related company.3 The sum total varied depending on news source but Srinivasan frequently made it a point to casually insert comments such as: we are the “most funded” or “best funded” company in Bitcoin into interviews and talks during 2015-2016.
In the beginning
In its early days 21e6 focused exclusively on designing custom ASIC chips for Bitcoin mining and then integrating and deploying Bitcoin mining hardware for private, non-retail usage. This included installing hundreds of hashing systems in data centers which for several reasons eventually became uncompetitive against those based in China and the Republic of Georgia.45
Based on publicly available information and allegedly leaked slides we know that:6
It closed its Series A for $5 million in May 2013.
Investors included: Peter Thiel, David Sacks, Max Levchin, Marc Andreessen, Ben Horowitz, Naval Ravikant, Winklevoss Capital, Mark Pincus
Estimated $3.8 million revenue in 2013
In June 2013, then-CEO Matthew Pauker filed a Form D, Notice of Exempt Offering of Securities, which stated that 55 investors had already invested in its offering.7 While that may sound unusual for an early stage company to have so many investors, recall what Popper pointed out above, that individual investors could invest directly into 21e6 because of its LLC status.
It closed its Series B for approximately $65 million in December 2013.
Andreessen Horowitz (the VC fund) invested $25 million as the lead investor; and $10 million came from existing investors (such as $100,000 from Pantera)
$30 million also came in the form of “venture debt”
Estimated $41 million in revenue in 2014
19 employees in November 2014
The funds from its first two rounds were used in part to design and deploy “Gandalf” (its 2nd generation ASIC chip) and “Yoda” (its 3rd generation ASIC chip) in the aforementioned data centers.
How much capital is required to build a state-of-the-art ASIC chip? Depending on how much is done in-house or out-sourced as well as the fabrication facilities, it can be upwards of $15 – $20 million.8
First major pivot
The company rebranded from 21e6 to 21.co and announced its Series C on March 10, 2015, with $56 million led by RRE Ventures. 9
That morning, The Wall Street Journal led with the story:
This marked the beginning of its pivot from purely building mining hardware and instead marketing itself as supposedly moving into the Internet of Things (IoT) and API marketplace. Around this time you frequently saw 21.co and its supporters publicly talk about machine-to-machine (M2M) payments as being a killer app.10 One of the 21.co engineers was even interviewed on a (now deleted) podcast where he spoke about how drone owners would pay tolls denominated in bitcoin to cut across airspace over yards in your neighborhood. You know, the usual word salad and shower thoughts on social media.
When I first drafted this memo 18 months ago, based on LinkedIn profiles, 21.co had about 25 full-time employees; as of now their page says 22 employees but most of them are just people adding 21.co in their profiles without formally being affiliated with it. Most of the current employees unsurprisingly have shifted to Earn.com’s official LinkedIn profile. Its tally is 63 people but again, some of these profiles are from people who are likely unaffiliated with the organization.
Alan Chang (Jeff Skoll’s family office via Capricorn Management) in Series B
Richard Tapalaga (Qualcomm Ventures) in Series C
Gen Isayama (World Innovation Labs) in Series C
According to Nathaniel Popper, as of March 2015 when it announced the closing of its Series C round, “the company has paid back all of its investors.” It did so partially via payouts in bitcoin.
In his self-canonization this week, Srinivasan wrote that:
And with this deal, the total value of cash, cryptocurrency, and equity returned to our shareholders is now in excess of the capital invested in the company.
How much of the cryptocurrency above is from the not-yet-released Earnable Token? Get the whitepaper while you still can.
Since March 2015, there has also been noticeable churn at the top:12
Matthew Pauker, co-founder, was replaced as CEO in spring 2015 by Balaji Srinivasan
Albert Esser was the COO from December 2013 through August 2015
Replaced by John Granata from March 2016 to the present
Nigel Drego co-founder, was chief architect from May 2013 through March 2016
Replaced by Jian Li as CTO from March 2016 through 2017
Lily Liu became CFO during summer 2015 to the present
Because of the economic incentives that tilt in favor of mining countries like China, 21.co stopped its operations in the Bitcoin mining sector and those subject-matter experts seem to have left the ranks.
Second major pivot. Or part of the first?
What has it built since the pivot after Series C?
The 21 Bitcoin Computer was their first consumer-facing product that was announced on September 21, 2015 and released with great fanfare as an exclusive to Amazon Launchpad on November 16, 2015 at a price of $400. It also picked up the “toaster” nickname from the Financial Times.13
Several enthusiasts explored the component prices via a piece-by-piece breakdown and found that it likely cost around $247 to build each 21 Computer.14 It was subsequently nicknamed the “Pitato” because the main component at its heart was basically a Raspberry Pi, a popular DIY kit that sells for less than $200.
The only other notable piece of tech was a custom built ASIC chip that could be used for mining. However, ever before it had shipped, the mining chip was already uncompetitive and obsolete. Even if you had free electricity you likely would not generate enough bitcoin in order to recoup the full cost of buying the 21 Computer, especially since the few satoshi you generated would be stuck as dust.15
What were the maths behind this?
In September 2015, after it was announced, Vitalik Buterin crunched the numbers and worked out that:
So you’re paying $399 upfront and getting $0.105 per day or $38.3 per year, and this is before taking into account network difficulty increases, the upcoming block halving (yay, your profit goes down to $0.03 per day!) and, of course, the near-100% likelihood that you won’t be able to keep that device on absolutely all of the time. I seriously hope they have multiple mining chips inside of their device and forgot to mention it; otherwise you can outcompete this offering pretty easily by just preloading a raspberry pi with $200 of your favorite cryptotokens.
Why the relatively large markup for a device? Part of it is that Amazon Launchpad gets a 25% cut.
But like just about all things Bitcoin, sales numbers were so bad that they were never disclosed and it was eventually discontinued. Prior to its discontinuation, 21.co representatives approached multiple well-known Bitcoin developers to help resell the devices. In short, these developers were offered to buy 21.co devices at wholesale prices and expected to resell them at the retail price. It is unclear how many (if any) developers did so.
For real, the second major pivot
On April 1, 2016, 21.co launched an app “marketplace” and initially seeded it with 50 apps that were built in-house. At the time, the only way to externally measure usage or traction is to manually observe the amount of ratings (stars) an app had each day. Interestingly, in early July 2016 the amount of apps stood at 95 whereas six weeks later it fell to 76 and basically fluctuated for the remainder of the year.
In May 2016, Srinivasan took the stage at Consensus and announced his vision of a “machine payable web” and introduced several ideas but notably did not mention the Bitsplit which was rumored to have been in the works for over a year.16
Throughout the remainder of the year, 21.co sponsored and hosted meetups and had an active Slack room, and most of the ideas that were used or borrowed as API and app ideas, languished due to… a lack of users.17 If you are new to my site, one reoccurring observation is that in general: cryptocurrency owners typically are not actual users, but that’s a whole different discussion.
The 21.co Marketplace now redirects to the Earn.com homepage.
On October 27, 2017, 21.co emailed its users that it was ending server-side support for three things: the Bitcoin Computer, 21 command line interface (CLI), and marketplace. 18
Three days later, 21.co announced that it was rebranding as Earn.com and pivoting away from its second vision as a VC-backed quasi protocryptojacking play towards taking on Amazon Mechanical Turk, but with Bitcoin. It also announced a non-ICO ICO called Earnable Token, which as you can tell from its name: was earnable from doing the same kind of tasks as you could before like: filling out surveys or answering bots who email you.
Earn.com also migrated the unique profile pages it first introduced with 21.co, which is basically a static page that users can claim and use a bit like LinkedIn, but with more Bitcoin-related spam.19
This last part is of particular interest in today’s regulatory climate because Earn.com, which hosts these user-controlled accounts, has accidentally assisted and enabled the promotion of alleged unregistered securities (ICOs) as a business line. Recall that Google, Facebook, Snap, Twitter, Mailchimp, and other tech companies have reduced or removed the ability for ICOs and cryptocurrency promoters to solicit retail investors, Earn.com has done the opposite and been a refuge. At what point is this an unsuitable risk profile for a “bank” like Coinbase?20
What does that mean?
In its January 2018 update, Earn.com announced that:
This week we were thrilled to announced the launch of Earn.com Airdrops — a new way for blockchain entrepreneurs to give 100,000+ Earn.com users a free trial of any new coin or token. Airdrops allows token projects to instantly bootstrap your new blockchain project with 100,000+ cryptocurrency early adopters.
We announced our first Airdrop partner, CanYa — a decentralized marketplace for services — as well as the next three upcoming Airdrops: Bloom, Bee Token, and Vezt. Sign up for an account on Earn.com, verify your account, and download the Earn.com mobile apps on iOS or Android apps to become eligible.
I am not a lawyer but in the past – like the dotcom era – companies (including startups) have attempted to give away equity in some very creative ways… and depending on the circumstances, it can be a no-no.21 That’s not to say that the tokens above are securities or that any airdrop is a violation of securities laws. But highlighting this type of feature has inadvertently led to Earn.com becoming a magnet for ICO issuance and promotion.
Where’s the beef?
What was the long term deliverable for roughly $125 million in nearly 5 years?
Throughout 2016 – including at Consensus in NYC – Srinivasan explained that they will announce a “surprise” in the coming months, maybe all of the aforementioned products and chips were the alpha phase of a much larger operation? Maybe they were, but we probably won’t find out.
Either way, it is worth keeping in mind that between 2013-2016, cryptocurrency-specific startups collectively received a little more than $1 billion in external funding, with nearly 15% of that funneled into just one startup. One who has had to pivot multiple times to find the right product-market fit and tech-market fit. Keep in mind too that other companies such as Bitfury and Bitmain were able to make superior chips and do so initially without major venture backing.22
If the most funded, best connected startup continually struggled to see consumer traction, what are the prospects for less funded and less connected cryptocurrency startups? This is worth revisiting in another long-read, especially in seeing what the $125 million was actually spent on (salaries? chips? toasters?).
One of the investors in 21.co responded to Nathaniel Popper above claiming that Srinivasan wasn’t actively involved in the first two years.
Does it matter? Sure, when you are claiming successes and denying failures that should or shouldn’t be attributed to you.
Below is a quick series of interrelated anecdotes.
In December 2014, Srinivasan and I both attended and presented at what would become the second of three round table events organized by R3 (a family office then called R3 CEV). This was prior to the formal creation of the DLG consortium.23 Unfortunately I do not have his presentation, but the layout and design were nearly identical to the leaked slides that have circulated for years — just with different content. For instance, the design of his slides at a public talk in the spring of 2015 is pretty close to the other two decks.
In January 2015, I was unexpectedly shown a long set of slides for a company called 21e6, most of which look similar to what has been leaked in the past and linked to above.24
Later that same month – due to a variety of circumstances – I met up with Srinivasan in Palo Alto and he quickly paged through the leaked presentation and stated it was an older deck from October / November 2014.
While there is a little more to our subsequent interactions, I think the key part here and the only reason I brought up this personal anecdote is the fact that Srinivasan was able to dismiss the deck of having any relevance on the current fundraising 21e6 was doing (remember, this was less than two months before the round was publicly announced).25
So while he may not have been “day to day” as he disclaims in his post, he clearly was involved in the fundraising process if not more (deck creation?). He said as much in a post published in March 2015.
So what to make of all of this news?
An exit is an exit, right?
What ultimately appears to have happened is that Andreessen Horowitz took one of its floundering portfolio companies and merged it with another portfolio company… and declared it a great success.2627
There also appear to be a few parallels with Juicero.28 For those unfamiliar, Juicero is a now-defunct Silicon Valley-based startup that built and sold a custom $400 machine that would squeeze juice packets. It raised $120 million and unceremoniously shut down last year after reporters showed that the hands from mere humans were capable of squeezing the same juice packets.
In much the same way, during the second pivot of 21.co, no one really bothered to buy the “Pitato” because users could easily do the math: that it was far more effective to either buy bitcoins outright or buy and use more capable mining hardware.
Why hasn’t anyone written about this before?
Most of the knowledge above is public, or at least, pretty well known if you have spent much time in Bitcoinland. Other reasons involve some tinfoil hat theories around retaliation.29
Funnily enough, back in March 2015 I had a long email exchange with Michael Casey and Paul Vigna over at The Wall Street Journal regarding 21.co and other several other topics.
Tim Swanson, a consistently skeptical digital-currency consultant who makes a habit of challenging bitcoiners’ unbridled optimism, is unequivocal. 21′s plan is “a dumb idea,” he says, adding that “the investors deserve to get what’s coming to them.”
And while a few of those investors probably did, it is Coinbase share holders that likely got it on the chin this week.30 If you’re looking for more memorable gems, be sure to read this older WSJ article. It is chocked-full of hubris, kind of like Juicero.31
In closing, raise your hand if you’d like to get paid every time you respond to an email and moreso to a cold email? I know I would.
So maybe with all of the kinks, toasters, pivot denialism, and chest thumping there is still a future for a pay-to-respond model to thrive. Maybe Coinbase can turn the ICO sanctuary of Earn.com into a legitimate mainstream product that is integrated with various webmail providers and social media platforms. Or maybe this ends up like ChangeTip, whose platform was basically used to spam coin dust on Twitter… to ultimately shutting down after an acquihire from Airbnb.
Either way, there was a bit more to this story than what was let on in Srinivasan’s original Medium post on Monday.
Would that be a Bitcoin-powered bus that the management team was thrown under? [↩]
Over the past several years, multiple venture funds have had their by-laws amended or re-written to allow them to purchase cryptocurrencies and directly invest into ICOs. [↩]
In March 2015, 21inc announced that it had raised a total of $116 million, however according to Nathaniel Popper’s account of their history, they had raised about $125 million. For one reason or another, historically many cryptocurrency companies do not typically reveal their active user numbers or revenue figures. Instead they prefer talking about how much outside funding they have raised. And 21.co was not an exception to this. [↩]
There are several reasons why this was the case. With the right guanxi: a combination of electricity, land, and taxes could be cheaper in certain parts of China versus the US. In addition, 21e6 and other US-firms were consistently unable to manufacture mining machines and operate farms at a similar scale as their peers. Part of this was logistics as well: large portions of the supply chain were based overseas (primarily in Guangdong and Taiwan). I have written about this in multiple different posts over the past several years, such as this piece. [↩]
One of the interesting things that Srinivasan’s article confirms was a rumor I first heard two years ago from one of their mining competitors: that 21e6 had signed leases with data centers whose energy rates were so abysmal that you might as well just bought coins instead as it would basically be impossible to recoup those costs. Another unconfirmed rumor was around immersion cooling: that between 2014-2015 21e6 had experimented and burnt through a large quantity of chip inventory in a radical attempt to reduce the cooling needs and costs of mining chips. [↩]
Form D – note that the domains 21e6.com and .net and .org all registered around March/April 2013. [↩]
Why Are Computer Chips So Expensive? from Forbes. In addition to non-recoverable engineering, there are also component costs and testing thereof: PCB, SMT, power supply, fans, integration. Testing and trouble-shooting cannot be ignored. For instance, Hashfast was an example of a company who built a relatively fast chip but had problems with managing the power source and consequently went bankrupt. [↩]
At the time it was frequently reported that 21.co had raised $116 million but that was the sum total of all funding rounds. The Series C was ~$56 million. [↩]
Srinivasan did talk about micropayments as early as March 2014. [↩]
In May 2015 it was reported that Cisco may invest or may have invested in 21inc. Padma Warrior, former Chief Technology and Strategy Officer at Cisco, was rumored to be a key individual involved in that deal. Note: as of August 2016, a site redesign on 21.co removed investors and corporate information from the homepage. [↩]
This is mainly because an earlier 21e6 pitch stated that the company would integrate mining chips in always-on consumer electronics and appliances. [↩]
One reviewer commented: “I’d say one more thing worth adding is that it’s worth critiquing not just the feasibility of the Pitato but also the ethics. Because Pitatoes are inherently less efficient than regular mining farms due to economies of scale, the only way that they could be competitive relative to just buying bitcoin is if they were using free electricity; that is, basically all profitable usage of Pitatoes would be people using other people’s electricity in workplaces, universities, Starbucks, hotels, homes if the landlord pays for it, etc. I predict that if it actually became popular, then we’d see all the places that provide free electricity today become much more cautious about it, which could greatly reduce convenience for everyone but bitcoin miners.” [↩]
In one of its incarnations, Bitsplit was basically a euphemism for socializing CPU labor and privatizing some of the gains… now commonly called cryptojacking. [↩]
One reviewer said: “That earn.com pivot was done through the 21.co meetups that would host with Bitcoin engineers trying to buils apps on the 21 computer, which was eventually bricked. The idea for paying engineers for github pull requests led to earn.com’s business model.” [↩]
Note: in between the second and third pivot, during January 2017, Srinivasan deleted his tweets and interviewed for the top job at the FDA in Washington DC. [↩]
One reviewer commented that: “My personal view is that the current Earn.com concept is fundamentally legitimate and probably will see some usage (I can totally imagine consultants charging $50 for replying to emails, as that’s a very low-transaction-cost way to get one-time advice from people), but it deserves to exist as one of the 173 configurable settings in an email provider or social media service, not an independent multi-hundred-million dollar company. Perhaps the Coinbase acquisition actually will be utility-improving, in that gives the Earn.com team an ability to try to be useful by making gadgets for an existing company that has a userbase and services, rather than trying to build their own ecosystem which never made any sense (though it’s still a pretty disappointing end relative to Balaji’s original hype and aspirations).” [↩]
The line of reasoning is as follows: some startups attempted to randomly give away shares to strangers via various gimmicks but ultimately had to either take it back and/or were sued. If certain ICOs are deemed securities, you might not be able to just give them away to anonymous people. Reminder: I am not a lawyer, talk to a securities lawyer. [↩]
One competitor noted that: “21e6’s decision to go the Intel fabrication route was a non-starter. [↩]
Someone should remind me to talk about the dinner conversation that evening as well. [↩]
Coincidentally a few days prior to receiving those slides, I spoke with a NYC-based investor who was asking about the pros and cons of embedded ASICs for mining cryptocurrencies. Specifically: should the fund invest in a startup designing embedded ASICs for bitcoin mining. I provided my view point (the answer was no, still is a no). During this same time frame there was a big meme being pushed by many Bitcoin boosters: that mining would somehow become re-decentralized via some unknown magic bullet. Some of these promoters believed that 21.co would be the one to do it, without much evidence that the company could (or that anyone could). Note: there have been multiple other attempts at building and shipping embedded ASIC mining chips including from Midea and Bitfury. None have been successful by any measure. [↩]
Remind me to mention the coincidence at Chipotle. [↩]
Another reviewer said: “This is acquisition theater, everyone is just trying to save face because this wasn’t a great idea, had wasteful execution, and the hype and hoopla reflects poorly on all involved. The players fundamentally misunderstood the tech, the economics and use cases. I get that a VCs job is to make unsubstantiated bets on tech entrepreneurs they like. But here, an outright $116m investment in Bitcoin would have yielded X billions. And the “we returned all capital” probably because of BTC dividends and its price hike than cash returns.” [↩]
What are the repercussions for publicly asking critical questions regarding bold claims such as those from a fireside chat with both Srinivasan and Andreessen? Being blocked on ol’ Twitter. [↩]
Since we are going into the anecdote highway: in March 2015, at the Stanford Blockchain Workshop event, I approached Adam Ludwin after his panel discussion. On the panel he had mentioned that there could be a “redecentralization” of mining through an upcoming “Silicon Valley moment.” I assume he was talking about 21e6’s plan for mining chips being integrated into always-on devices because he was affiliated with one of its investors. When I told him I had seen a 21e6 deck and that it was making some very wild, likely incorrect assumptions, he basically said: we will see about that. Well, we have seen that once again: the difficulty rating rises with prices thereby diluting existing hash generating devices making them obsolete. [↩]
Some of the comments from the 21.co spokesperson are enjoyable. These hashing devices still wouldn’t be profitable at the current prices today because the difficulty rating has increased in proportion to the price yet all of the hashing units inside phone chargers and toasters had a fixed unit of labor. It’s a no-win situation for device owners as they would still have to pay for both the depreciating capital good (the device) as well as the electricity. [↩]
A couple hours after publishing this, a reader reached out and mentioned that: “I’m a proud owner of a Pitato. You forgot to mention that Balaji taught a course at Stanford about cryptocurrency and basically used it to promote 21co and Pitato to students. He gave it for free to students but all the labs were on this hardware. IMO it’s a conflict of interests for him as a professor <-> manager. The instructional material and repo is still online: (1) (2) (3) (4)”. Note: I don’t think this is a conflict of interest, professors and lecturers sometimes have their students purchase a book they may have authored/co-authored as they are the subject-matter expert. [↩]