Last Friday (seven days ago) I sat down with Alex Ao, one of the founders, and currently the VP of Business Development at Innosilicon.
Here’s his business card that he gave me (phone numbers blacked out).
A few days earlier Innosilicon had announced a Zcash hardware mining unit that does 50 KSol/s at 650 Watts, named “ZMaster”. That’s 5 times as much hashes per second as the Bitmain Antminer “Z9-mini”, at about 2 times the power draw. The current list price for Innosilicon ZMaster is $9999 dollars. The Bitmain Z9-mini was already selling at $1520 dollars, and a few days ago Bitmain announced a halving of the price to $760 dollars per unit.
The meeting was brokered by Dovey Wan of Danhua Capital.
Here are my recollections from the meeting, informed by some handwritten notes I took during. I might forget or misremember some bits. By writing this, I’m not asserting that I believe everything Alex told me, nor am I asserting that I disbelieve anything. I’m just reporting what was said.
Innosilicon is an IP (chip) design firm. In 2014–2015 they made a 28 nm BTC miner, and in 2015 they made a 28 nm LTC miner. They started design work on a ZEC miner about four months ago, in February 2018. (I was surprised that it took them only four months from beginning design work to having a large number of live units in production!)
He said they currently have 1000 ZMaster units in production, mining on the Zcash mainnet. They’ve been live for a couple of weeks. These are the units that are for sale—they are intended to be shipped to customers. They are mining on them right now for testing and aging. A first batch, they would mine it for a week and then sell. Later on, subsequent batches, they would just run a couple of days.
I asked what mining pool they use. He said their own private pool and also that they test on different pools.
I asked if I can publish this stuff, he said no problem.
He said that their whole idea an “open ASIC”, that they want to support the ecosystem. He said that they would be happy to sell others the chips and they can build the miners, or that they would be happy to licence the IP and let others fabricate chips. I said that this sounds related to Josh Cincinnati’s recent idea about an open ASIC and they should talk.
He said that they also made a Monero miner. I asked if they ran a Monero mining farm and he said yes. I asked when they started and he said he wasn’t sure because he wasn’t the manager of that project but he thought it was last December. I asked how many units and he said fewer than 1000.
!!! I was excited, because this means we’ve found some of the mysterious Monero stealth ASICs! When Monero switched their proof of work in April, the hash rate fell by about 380 MH/s, and subsequently that much hashrate started mining on Monero forks like XMO and XMC. If Innosilicon was running up to 1000 units of their “A8+ Cryptomaster”, which does 240 KH/s of Cryptonight hashing each, then that accounts for about half to two-thirds of the mystery hash-power!
I asked if they work with Bitmain. He immediately said “No!” and laughed. “We are competitors!”. I didn’t think he was lying.
I asked if Innosilicon worked with ASICMiner. He hadn’t heard of it. I showed him ASICminer Equihash 40K profitability | ASIC Miner Value, and he said he hadn’t seen it before and they don’t work with them.
I said that I thought they were missing out on a lot of sales. I said that a lot of potential customers have been contacting me, asking if the Zcash community was going to change the Proof-of-Work to be incompatible with the current ASICs, and that I always tell them I can’t say — that I don’t know what the community will decide. I said that these customers are reluctant to buy, because if the community changes the Proof-of-Work then the miners will produce a lot less value, like the “XMO/XMC” miners now. He nodded.
I said that I thought if he could get in touch with The Zcash Foundation and get them to make some public commitment about not changing the Proof-of-Work, or about including both Equihash-200,9 and other things in the future Proof-of-Work, then he could get a lot more sales.
I said that what would increase trust, for me personally, as well as in my opinion for many other members of the community, is transparency. I said that what I wanted was a commitment from Innosilicon that they would disclose to the public all shipments of Zcash miners, including what size batches they are selling in, as well as to disclose all deployments that Innosilicon does themselves of running Zcash miners, as well as what policy they have about which blockchains, transactions, pools, and precisely what software they run (i.e. which specific verson of zcashd, or if they run their own patches to zcashd). I explained that this latter part would help the dev team diagnose and respond to issues on the network if we knew more about precisely what software the miners were running.
I showed him the announcement and the tweets from Bitmain disclosing their shipments of the Z9 mini, which shows that they’ve shipped the first batch of 77 units, total. He said if you ship 77 units it is okay to disclose but for 1000 it would be a different story. I said that I understood that, but that what I want is a commitment to disclose all shipments and all deployments and not to omit any shipments or deployments. (Which, by the way, I haven’t seen a clear statement of a commitment to that policy from Bitmain, and I think that would be a great improvement to their transparency initiative.)
Also, I said, come to Zcon0 and meet Josh Cincinnati and everyone. He happily agreed.
[Note: after I wrote this, Josh Cincinnati reminded me that he won’t be physically present at Zcon0 because he’s about to have a baby! But Josh will be cyber-present, and there are plenty of other members of the community that Alex can meet there.]
[Note: after I wrote this but before I posted it, The Zcash Foundation issued an open request to Bitmain, Innosilicon, et al. to commit to transparency with a similar list of requests. This is no coincidence! The Foundation and me have been talking about this.]
He said again that Innosilicon’s background is in IP (chip) design, and their strategy is to promote the whole ecosystem. He said that he had helped negotiate the deal by which Litecoin added Segwit. He said something like:
We made contributions to the Litecoin Foundation and funded developer to make more features. Then a certain party blocked the addition of Segwit on Litecoin [note: he meant Jihan Wu]. Then I convinced everyone to support Charlie to introduce Segwit, which made the price go from [some number that I didn’t write down] to [some other number that I didn’t write down].
He told me more about the process of breaking that deadlock that I didn’t write down, but one detail stuck in my memory: he said that at one point the certain party that was blocking Segwit threatened to use its substantial Litecoin hash power, and he revealed that he had even more Litecoin hash power. This detail stuck with me after the end of the conversation…
The conversation was winding down. I asked again if they intend to run a Zcash mining farm. He answer: that’s not our intention. If the market is not well, if I cannot sell them, I will run them. Secondly, for testing and aging. I cannot play as both athlete and referee.
I asked, why were you running the Monero mining farm? He answered, we had a big customer who wanted to buy the miners, but they don’t have electricity, so we run the farm for our customer.
[Note: after I started writing these notes, Zcash engineer str4d made these graphs of mining pool hashrates.
Maybe Innosilicon’s 1000-ZMaster-unit testing farm is the one labelled t1Q2Q and t1XE1? I’m not sure 1000 ZMaster’s would produce that quite that much hashrate, so perhaps Innosilicon has more than 1000 running, or perhaps someone else is also pointing miners at the same pool? Or maybe that is the right hashrate for 1000-ZMaster’s?]
P.S. After I write this whole thing and just before I posted it, Dovey Wan from Danhua texted me to say that there is another new pool which quickly gained 40% of the current hashrate! I haven’t looked for the relevant stats from str4d’s new visualizer yet. Never a dull moment!