Addressing the many concerns related to Obelisk

in #sia7 years ago

Why make ASICs at all?

Our blog has a longer post on the subject, but the ultimate answer is that GPU mining is very insecure. For the vast majority of GPU mined coins out there (including Sia), it is the case that there are multiple, if not many, individuals who operate enough GPUs to execute a 51% attack against the coin all by themselves. There are some very large Ethereum GPU farms out there, and they are a threat to all small GPU-mined coins. (our market cap is a factor of 50 smaller than Ethereum - we are a small coin). And it's not just Ethereum farms to be afraid of, there are massive GPU farms dedicated to machine learning as well, and other big-data related use cases. All of those are potential sources for a 51% attack. Even worse, if the price of the coin tanks following such an attack, the attacker has nothing to lose, because the core purpose of their hardware is unrelated to Sia, and unaffected by a change in price.
Though it sounds terrible and unintuitive, a single centralized entity running ASICs would be a much more secure situation than this. Because with a single central ASIC entity, you get two huge advantages:
There's only 1 entity capable of performing a 51% attack. This is much better than having multiple entities that are each individually capable of performing a 51% attack.
If the price of the coin falls, the entity that has all of the hardware loses a lot of money. That hardware isn't good for anything besides Sia mining, so that entity is quite invested in propping up the siacoin price.
We chose ASICs over GPUs because even the worst case scenario is more secure and better for the coin than the situation with GPU mining.
But we also did not want a single entity owning and operating all of the ASICs. That's when we realized, if we were ASIC manufacturers ourselves, we could guarantee that at least one entity is selling chips to the larger community. The unfortunate fact is that either way, there is going to be a small number of chip manufacturers who have the power to sell chips to the community. Even so, this is a better situation than what you get with GPU mining.
We are making ASICs so that we can guarantee the first batch of ASICs will make it to the Sia community. Without that, we have no idea if the first batch of ASICs will be sold to the public or hoarded by some greedy investors who were able to pay the full price of manufacturing up-front.
Why are you doing the presale so early?

We, put simply, don't have enough cash even to do the early development of the chips. We need financing to pay for chip development.
Traditionally, we would find some private investors, have them front some millions, and in return promise them a very good deal on some hardware. The private investors would get the first stab at buying ASICs, they'd get a huge chunk, and they'd get them at an exclusive deal for taking on the risk early. We actually had private investors come forward offering this to us, with enough money to fund the full development and manufacture of the first batch of chips - this isn't a hypothetical, it's a real offer that the Sia team received.
This didn't seem fair to us. When we finally did get to the point where the miners were ready to be sold to the community, we would have to offer the community a worse deal. Less risky, but ultimately it would mean that the community was excluded from the opportunity of participating early, and the result is a huge chunk of the chips going to some private investors.
Such a situation is still better than GPU mining, but it didn't seem like the best that we could do. We felt that we could do better by opening the early presale to everyone.
Why not accept credit cards?

Payment processors are not friendly to Bitcoin products. We contacted Stripe and were told point-blank that they would not process payments for cryptocurrency miners. We appreciate everyone who pointed us towards Stripe as a bitcoin-friendly company, but they gave us a direct no.
Paypal has a long history of freezing merchant accounts with little warning, and when they do so they freeze your existing money in addition to freezing incoming payments - we would be unable to pay our bills if Paypal did this to us, and it would unquestionably cause delays. Visa and MasterCard are not much better in terms of track record.
Losing access to our accounts would unquestionably cause delays. ASIC hardware is already well known to suffer from serious delays, and we need to limit our exposure to delays.
We are in an industry that is unfortunately fraught with fraud. With revenue-generated devices such as miners, criminals are much more likely to try to target these devices as a way to cash in on stolen credit cards, stolen identities, hacked bank accounts, etc. The fraud rates are staggering, and as a result most payment processors outright refuse to deal with it. We are aware that Bitmain is partnered with Paypal, though we don't know the details behind how that came to be.
Why not accept Siacoin?

This was a harder decision. We could quite easily choose to accept siacoin, however we fear that Siacoin is not ready to handle such a massive presale. The market cap and daily volume of Bitcoin is a factor of 100 times as large as the Siacoin market cap and volume. Moving millions or tens of millions of dollars through Bitcoin is not likely to make much of a dent. Siacoin on the other hand, a sudden sell order for millions of dollars would likely tank the price. That not only means the ecosystem is unhappy with us, it also means that we might only be able to sell $2499 of siacoin for $2200.
A lot of people have accused us of not having confidence in our own coin. Unfortunately, this is true. Even at a $500 million market cap, Sia is not ready to handle a presale of this size. It's a pragmatic decision based on the fact that we don't want to dump our own coin. We know that people will be selling siacoin to buy the miners anyway, but we still feel that this situation is much better than us accepting siacoin directly.
This decision was a disappointment for us as well. We would love to accept siacoin, and if we weren't talking about processing millions of dollars in a single day, we absolutely would be accepting siacoin. And, as Sia continues growing up, the concerns above will become less and less.
What about this 5% gains/losses stuff?

Our intention was never to play fishy financial games with our users, and honestly this isn't even something that crossed our minds as a potential problem point. I think a big part of the issue was that people did not realize we will be converting to US dollars as fast as possible - we will be doing the conversion in minutes or hours as long as we can keep up with the order volume.
The rationale is very simple. If the price plummets before we are able to convert the Bitcoin, we won't have enough money to create the hardware. We really don't expect this to matter, because we don't expect the price to swing by more than $100 (which is what would be required) in the few hours that we're going to be sitting on the BTC. If it does, we'll need more coins or we can't produce the hardware - our costs are in dollars, which means we need to end up with the right amount of dollars in our account at the end of the day.
The original stance on not returning gains was also very simple. There's no transparency into when we sell the coins. If we sell the coins within 60 minutes of receiving them, and then 4 hours later there's a huge surge in the price, we will almost certainly have users emailing us and posting about how we owe them a refund. We won't have that refund, because we'll have sold the coins before the price rise.
There's not much we can do to provide transparency into this either. And we're likely to get requests for refunds even if it takes 3 months for Bitcoin to rise by 5%. This promise of returning gains that we've put forward is going to be a massive headache, because we're not expecting to have any gains, even if the price goes up by that much we'll have likely converted to USD faster than that. Our whole goal is to convert to USD as fast as possible.
We're sorry that we have to go through this headache at all. If we could get set up with a processor like Stripe, we could accept both Bitcoin and USD and let them deal with the conversion process, slippage risk, and all the other headache associated with using multiple currencies.
Why shipping a full 12 months away?

Before we set out to make Sia miners, we did a study of companies who had previously sold and pre-sold Bitcoin miners. This included talking to both Avalon and Butterfly Labs, and talking to professionals and advisors who have shipped hardware successfully in other industries. The core piece of advice we got was pretty consistent: expect delays. Expect lots of delays, and expect them to come from the most absurd setbacks. (Example: one of the people we talked to had to delay their product because there was a global shortage of power supplies, and they had to wait in line behind billion dollar companies to get some).
Our projections indicate that if all goes well, we should be able to ship the miners in 6-8 months. Nothing we are doing is new. Plenty of companies have gone through the process of developing a chip, manufacturing it, putting it in a box, and then shipping it to users. There is almost no innovation risk here. Sia's PoW algorithm is deliberately very ASIC friendly, even more than Bitcoin. We have advisors who have gone through this process before, and the types of challenges facing us are well known.
6-8 months is reasonable, except that every single person we've talked to has told us that unexpected delays is a guarantee, and that by nature of being unexpected, there's not really any way to prevent them by planning around them. Delays are just inherent to shipping hardware. So we chose to set our target at 12 months.
We will ship the miners as soon as they are ready. If we are a few months ahead of schedule, and have somehow managed to avoid the foretold delays, we will ship them months ahead of schedule. But we want our users to have a realistic understanding of the expected delays. We've baked a generous amount of time for setbacks into our shipping date. We'll almost certainly need at least some of it.
Why $2499?

Making chips is very expensive. We have to sell thousands of units to cover the cost of the chips. A nontrivial percentage of the price is going to go towards chassis, shipping, power supply, control board, fans, etc. Those costs are relatively the same even if we put in fewer chips, which means the total percentage of our budget going towards chips drops significantly. If we cut the price in half, we'll have to sell roughly three times as many units to break even on the cost of the chips. If we cut the price in half again, we'd need to sell a completely unreasonable number of units to break even on the cost of the chips. It's unfortunate, but the fixed costs of chip manufacture means that we really need vast majority of the price of the unit to be spent on chips, otherwise we simply won't be able to sell enough units.
There is a second reason as well. As stated in the section above, the industry is plagued by delays an unexpected expenses. We need a healthy budget to plan around potential setbacks, because we've been guaranteed that there will be multiple significant setbacks by those who have gone through this process before. If we bring down the price of the unit, we will also be reducing the amount of wiggle room we have for disaster if suddenly we have to replace parts, re-do designs, or otherwise perform expensive adjustments to our plans.
Are you guys qualified to be working on hardware?

Zach is a mechanical engineer, I've been in the Bitcoin space since before ASICs started shipping, and we have advisors who have successfully shipped hardware before. The team that is designing the chips for the miner has designed chips and shipped chips for Bitcoin miners previously - they are familiar with the whole process, and have done it before. The people in charge of designing the PCB board and other aspects of the miner are also all experienced with their respective tasks. We will be facilitating frequent and strong communications between everyone working on the various components of the miner.
The ultimate answer is that the Sia development team is not qualified to be making this type of hardware. However, the Sia development team is not the team working on the hardware. Most of the heavy lifting is being performed by teams with lots of experience in this industry, including experience that is directly related to cryptocurrency miners.
What we are doing is not new. Dozens of cryptocurrency miners have been created and shipped in the past, and we are not starting from day zero. We have many advantages over the previous rounds of pre-sale cryptocurrency miners, but the biggest is that it's no longer the wild west of hardware design. There is a standard, and there are tried-and-true methods for making reliable cryptocurrency miners. We get to fall back on the mistakes and successes of the many miners that have been built previously, and we will be leaning heavily on teams and people that have direct experience in this field as opposed to doing everything ourselves.
Does this mean that Sia is getting less attention from the developers?

Sia right now has four full time employees. Myself, Zach, Luke, and Johnathan. Zach was hired in June 2017, less than one month ago. He is not a programmer.
Luke and Johnathan will continue with the same responsibilities that they've always had. They helped out a little bit in setting up the website, and in setting up a secure database to process orders + payment information, however the majority of their time has been focused on Sia even as we set up this presale. Going forward, they will be almost entirely uninvolved in Obelisk.
I have had to allocate about 25% of my time to Obelisk. Slightly more this week, due to the PR meltdown we had from the initial announcement. But most of my time is still going towards Sia. Most people know I work over 100 hours per week (some weeks will eclipse 120), and that a quarter of my time is not a small amount.
Zach is closer to 50% Sia, 50% Obelisk at this point. We're expecting that to tone down once the presale is over - much of this time has been spent with banks, with lawyers, with payment processors, and we won't have to do that beyond the initial setup phase. Zach and myself will still be having weekly conversations with every part of the Obelisk supply chain, including the chip designers, chip manufacturers, control board designers, the miner assembly teams, and the fulfillment centers, so even after the presale there will be effort going towards Obelisk.
But nobody on the Sia team is doing chip design, nobody is doing control board design, most of the really heavy work is being done by experienced teams and suppliers that we've found and already spent weeks vetting and verifying. We incorporated Obelisk as a separate company precisely so that Obelisk would eventually have a completely separate team.
And finally, as Obelisk is wholly owned by Nebulous, a successful hardware company does mean revenue and income for the Sia team. Cryptocurrency mining tends to be low margin, so tens of millions in revenue for Obelisk does not necessarily millions in funding for the Sia team. But it is something, and it will give us more time to get the storage platform to the next levels of maturity.
Conclusion

I know that a lot of you are concerned about the miner presale that we are conducting. I hope that this post has helped to alleviate those concerns. I hope it makes sense why we are doing a public presale, instead of seeking private investment until we have a full prototype. I hope this post has clarified our decisions around payment methods, and around our price point. I hope you feel more confident that this is something we will be able to pull off. And finally, I hope I've reassured you guys that Sia is still our primary focus, and that we haven't suddenly pivoted into being a hardware company.
We are ultimately doing this to provide better security to the Sia network. GPU mined coins are frighteningly insecure, and Sia is now large enough where there is serious money on the line. We are doing this to gain security, and also to ensure as much decentralization as possible when it comes to chip manufacture.
We are typically viewed as one of the most reputable teams in cryptocurrency, and I know it's why a lot of you are here. We hope that the Sia ASIC that we are going to be manufacturing and selling strengthens this reputation, but ultimately we will not find out until the miners are actually being shipped.
We continue to be excited about this new product. We truly do feel that ASICs are the right direction for Sia, and we also feel that we are doing the right thing by bringing the opportunity to own a Sia ASIC to the broader Sia community. We are sorry for the fallout from our sloppy original announcement, and we hope that we have since made up for it.
Finally, we hope that you are interested in buying a miner. Even if we only sell a small batch, ASICs are going to utterly dominate the hashrate of Sia going forward. This is an egalitarian sale where everyone has equal opportunity to buy a miner - there's no cap, and we will ensure that small buyers are not shut out by larger buyers in any way.