Digital Gold (18 page)

But both of these key Silicon Valley figures were also getting more comfortable with Bitcoin itself. The investment firm that Thiel had helped create with some of his PayPal riches, the Founders Fund, began talking with an engineer at Facebook who had founded an e-mail list for Silicon Valley insiders, dedicated to Bitcoin, about joining the firm to look for virtual currency investments.

The growing openness to Bitcoin was helped along by Silicon Valley's ballooning sense of self-importance in early 2013. With the Nasdaq composite stock index soaring, shares of Google at an all-time high, and startups selling for mind-boggling sums, many in the tech industry believed that they were going to be able to revolutionize and improve every element of modern life. Investors and entrepreneurs were cooking up ever more ambitious schemes involving virtual reality, drones, and artificial intelligence, alongside more quotidian projects, like remaking public transportation and the hotel industry. The PayPal founders were among the most ambitious, with
Thiel advocating for floating structures where people could live outside the jurisdiction of any national government. Elon Musk, an early PayPal employee and founder of SpaceX, was
aiming for the colonization of Mars. If there was ever a time that Silicon Valley believed it could revive the long-deferred dream of reinventing money, this was it. A virtual currency that rose above national borders fitted right in with an industry that saw itself destined to change the face of everyday life.

March 2013

A
t the same time that Bitcoin's reputation was getting a makeover in Silicon Valley, the physical infrastructure of the Bitcoin network was also undergoing an extensive transformation.

For much of the previous year and a half, the computing power underpinning the network had grown steadily, but slowly. Over the course of 2012 the amount of computing power on the Bitcoin network barely doubled. What's more, everyone was still relying on basically the same technology—graphic processing units, or GPUs—that had been introduced back in 2010 by Laszlo Hanecz, the buyer of the Bitcoin pizzas. By the end of 2012 there was the equivalent of about 11,000 GPUs working away on the network.

But even back in 2010 it had been clear that if Bitcoin became more popular there was a logical next step that would eclipse GPUs. An application-specific integrated unit, or ASIC, is a chip that is built to specifically accomplish just one task—an even more specialized computing unit than a GPU. If someone could build an ASIC designed specifically to solve the Bitcoin hash
function, it would probably be able to crunch the numbers hundreds of time faster than a GPU and thus likely to win hundreds of times more Bitcoins.

But designing and fabricating a new ASIC chip could cost millions of dollars, and take several months, requiring contracts with one of the five specialized chip foundries that produced virtually all the chips in the world. For most of 2011 and 2012 Bitcoins simply were not worth enough to justify this investment.

But as Bitcoin's price had continued to rise in the second half of 2012, a couple of enterprising engineers had thrown caution to the wind and begun racing to create the first ASIC chip dedicated to mining Bitcoins. The first entrant in the race was a company in Kansas City that went by the name Butterfly Labs.
In June 2012 the founders announced that they would deliver specialized mining computers installed with custom chips in October 2012 and quickly sold $5 million of the machines on preorder.

A few months later, when Butterfly announced that the release of its machines would be delayed,
a young Chinese immigrant in New York, Yifu Guo, announced that he had created a company, Avalon, with a group of engineers in China, which was building its own Bitcoin-dedicated ASIC chips.

Yifu, a shaggy-looking twenty-three-year-old, promised that each device would be able to do 66 billion hashes per second, compared with the 2 billion that a GPU card could do. What's more, his chips required a lot less energy—and thus lower electricity costs—to do the work. The price for each machine? A cool $1,299.

The process of putting the machines together, first in Beijing and then in Shanghai, and then shipping them to customers in the United States, proved to be more complicated than Yifu and his team anticipated. But on January 30, 2013, Jeff Garzik, the Bitcoin developer in North Carolina, posted on the forum pictures of the bulging boxes that DHL had just delivered and the gleaming
silver box inside, built to do nothing but mint new Bitcoins. Within hours, new Bitcoins were showing up in Jeff's wallet, and within nine days the machine had earned back what Jeff had paid for it. The machine was eating up so much energy that it was heating up the room that it occupied.

Over the next month and a half, as the rest of Avalon's first batch of three hundred mining computers reached customers, the effect was evident on the charts that tracked the power of the entire Bitcoin network. It had taken all of 2012 for the power on the network to double, but
that power doubled again in just one month after Yifu's machines were shipped. At the same time, the network automatically adjusted the difficulty of the problem the miners needed to solve, to ensure the ten-minute gap between new blocks of Bitcoins. For people who had built up fleets of GPUs making a profit quickly became a lot harder.
^*

A few other companies were making big promises about their own, specialized mining chips that they were working on. But the most aggressive project—and the one that revealed the most about the untapped potential that many saw in Bitcoin mining—was top secret and open to only a small elite. The company 21e6—shorthand for 21 million, the number of Bitcoins to be released—was created by Balaji Srinivasan, a Silicon Valley prodigy who had founded a successful genetics testing company from his Stanford dorm room. In the spring of 2013, Balaji was quietly assembling a team of top engineers to build a Bitcoin mining chip that would go beyond anything that had been contemplated before—rolled out in data centers built exclusively for the 21e6 machines. If the chips worked as promised they would mint money for investors. This was a simple enough proposition, and the price of Bitcoin was rising fast enough that it attracted interest from venture capitalists
who were still queasy about tying their firms to Bitcoin. Both of the founders of Andreessen Horowitz, Marc Andreessen and Ben Horowitz, signed up to put some of their own personal money into Balaji's project, as did several of the original founders of PayPal, including Peter Thiel and David Sacks. Soon enough, Balaji was closing in on a $5 million fund-raising round.

The Bitcoin arms race had begun.

T
HE TYPE OF
chip was not the only thing about Bitcoin mining that had changed since late 2010. Over the course of 2011 and 2012, more and more users were joining collectives that pooled their mining power. These mining pools allowed lots of people to combine their resources, with each person getting a proportional fraction of the total winnings, thus increasing the chances that everyone would get something every day.

The pools, though, generated concern about the creeping centralization of control in the network. It took the agreement of 5 percent of the computing power on the network to make changes to the blockchain and the Bitcoin protocol, making it hard for one person to dictate what happened. But with mining pools, the person running the pool generally had voting power for the entire pool—all the other computers were just worker bees. As a couple of pools harnessed significant computing power, some people worried that the operators of those pools could conspire to change or undermine the rules of Bitcoin.

But an incident in March 2013—the network's most significant technological failure to date—was a reminder of how the incentives built into the Bitcoin network could still work as Satoshi had hoped. Gavin Andresen, now the chief scientist of the Bitcoin Foundation, was in his den in Massachusetts after dinner, when he saw some online chatter about disagreement between computers
or nodes on the network over what block the nodes were trying to mine—was it the 225,430th block since the network began back in 2009, or the 225,431st?

Gavin quickly realized that this was what had long been known as the biggest potential danger to the Bitcoin network: a “hard fork,” a term coined to describe a situation where one group of computers on the network went off in one direction, agreeing about which node had mined each block, while another group of computers on the network moved in another direction, agreeing on a different set of winners for each block. This was disastrous because it meant that there was disagreement about who owned which Bitcoins. So far, there had been a split only on the last few blocks—not the whole blockchain history—but if it wasn't fixed, there would essentially be two conflicting Bitcoin networks, which would be likely to result in no one trusting either of them, or Bitcoin itself.

“this seems bad,” a user on the chat channel wrote a few minutes after the problem first appeared.

“‘seems' is putting it lightly,” another shot back.

“We have a full fork,” one of the most respected developers, a Belgian programmer named Pieter Wuille, pronounced a few beats later.

The price of Bitcoin dropped from $49 back to $45 in a half hour, erasing all the previous week's gains.

Mark Karpeles joined the discussion a half hour later, and quickly stopped processing all transactions at Mt. Gox; a few minutes after that, Erik Voorhees said his gambling company, SatoshiDice, was doing the same.

By the time Gavin entered the conversation, it was clear that the problem was not the result of one node overpowering the network or of any sort of malice. Instead, computers that had downloaded a recent update to the Bitcoin software were accepting blocks—and awarding new Bitcoins to miners—that were not considered legitimate by the old software and the computers still running it.
Generally, if a block was accepted by a majority of nodes, it would be accepted by everyone, but the old software, version 0.7, had a rule that specifically did not allow a type of block that the new software, version 0.8, did allow.

The solution to this was clear: everyone on the network had to agree to move en masse to one of the two versions and adopt the blockchain accepted by that software. But there were no rules for deciding which version to pick—and once a version was chosen, no one knew how long it would take for all the nodes to get on board.

After racing through the possibilities, Gavin concluded that the most fundamental rule of Bitcoin was the democratic principle that the blockchain with the most support was the official one. In this case, the version created by the new software, 0.8, had a lot more computing power behind it. That was, in no small part, because the most sophisticated miners, especially the large pool operators, had been among the first to update their software. Gavin thought that if they had the most power, everyone else needed to update to join them. In addition to having more power, the miners on the new software had newly generated coins that they would be unlikely to want to give up.

Gavin quickly faced resistance from almost everyone else involved in the conversation; most participants believed that only the large miners would be responsive enough to change their software to fix the problem. Somewhat surprisingly, the operators of the biggest mining pools quickly agreed that they would revert to the old software, version 0.7. The operator of the prominent pool BTC Guild said that just switching his pool alone would get a majority of the computing power back on the earlier software. Doing this would mean losing the Bitcoins that had been mined since version 0.8 came out. But the losses would be much greater if the entire Bitcoin network lost the confidence of users.

“There is no way the 0.8 chain can continue in this situation,” the operator of BTC Guild, who went by the screen name Eleuthria, said.

The developers on the chat channel thanked him, recognizing that he was sacrificing for the greater good. When he finally had everything moved about an hour later, Eleuthria took stock of his own costs.

“It could've been worse if I hadn't been able to start moving back to 0.7 quickly.” But, he wrote, “this fork cost me 150–200 BTC”—over $5,000.

For the broader Bitcoin ecosystem, the price had fallen to $37, some 20 percent, within a few hours, and some online reports struck an ominous note.

“This is a dark day for Bitcoin. Implications for the exchange rate will likely be huge,” a site called The Bitcoin Trader announced.

The incident had indeed revealed the sort of unanticipated problems that frequently occur in decentralized networks, which rely on lots of different members, with all their vagaries, acting independently.

But almost as soon as Eleuthria had fully switched his servers over to version 0.7 the price began recovering, and within hours people were talking about how the event had actually demonstrated some of Bitcoin's greatest strengths. The network had not had to rely on some central authority to wake up to the problem and come up with a solution. Everyone online had been able to respond in real time, as was supposed to happen with open source software, and the users had settled on a response after a debate that tapped the knowledge of all of them—even when it meant going against the recommendation of the lead developer, Gavin. Meanwhile, the incentives that Satoshi Nakamoto had built into the network had again worked as intended, encouraging people to look out for the common good over short-term personal gain.

A
WEEK LATER
, Gavin was back at his desk in the den not long after dinner, when an unexpected announcement popped up. It came from the Financial Crimes Enforcement Network, or FinCen, the division of the Treasury Department responsible for monitoring money laundering and enforcing the Bank Secrecy Act. In opaque bureaucratic terms, the release stated its intent to
“clarify the applicability of the regulations implementing the Bank Secrecy Act (‘BSA') to persons creating, obtaining, distributing, exchanging, accepting, or transmitting virtual currencies.”

Reading behind the legalese, Gavin could see that this was the United States government's first statement on the legality of Bitcoin.

“oh wow,” Gavin Andresen wrote on the chat channel before passing along a link to the announcement for everyone else.

Everyone had feared that at some point the authorities would step in and declare virtual currencies illegal. As Gavin and others furiously scanned the lengthy document, the doomsayers were quick to give their read.

“this kills the Bitcoin,” one user on IRC responded to Gavin.

But as Gavin and others read on, they saw that it was not, in fact, all bad. Yes, the document noted that anyone selling virtual currency for “real currency or its equivalent” would now be considered a money transmitter—a category of business subject to lots of stringent federal rules. But the release also made clear that many parts of the virtual-currency universe—including miners—were not subject to these regulations. More important, Jeff Garzik, the programmer in North Carolina, noted, the basic implication of the message cleared up the biggest single cloud: “this solidifies Bitcoin status as legal to possess and use for normal people.”

Indeed, Gavin said: “More legal/regulatory certainty is definitely a good thing . . . even if we might not like the regulations.”

Over the next few days, Bitcoin companies all raced to understand the specifics of the FinCen guidance. Exchanges clearly needed to register as money transmitters, but what about companies like BitInstant that just worked with exchanges? And did exchanges also need to register as money transmitters with each state, as companies like Western Union had to do?