Due to a lack of interest and funding, Obelisk has had to cancel the GRN1 project. All who purchased the GRN1 directly from Obelisk will be receiving a full refund for their purchase. Refunds will be provided in the form of BTC equal to the USD value of the amount paid at the time of purchase. If credit was used, the credit will be added to the amount refunded. If coupons were used, the coupons will be returned as coupons.
Several factors proved to be significant barriers to the GRN1 project. The biggest challenges were the split block reward and algorithm phase-out. When the phase-out of Cuckatoo31 was announced, Obelisk had to decide whether to abandon the current design and switch to building a Cuckatoo32 capable miner, which would result in shipping delays, increased manufacturing cost, and reduced competitiveness, or to continue pursuing Cuckatoo31 and gamble that the Grin price would be high enough to justify such a short lived miner.
Obelisk chose to believe in the price of Grin and move forward with a more aggressive and more competitive piece of hardware. This ended up being the wrong choice, and the Grin price fell to below $2 right at the critical point for Obelisk (May 2019).
Though that was ultimately a strategic error from Obelisk, the error was the result of a highly complex set of tradeoffs presented by the Grin team in their choice of PoW algorithm:
Option 1: GPUs. Lowest financial risk, because the GPUs can always be pointed at another cryptocurrency later in time.
Option 2: Flexible ASICs. Assuming no algorithm-specific ASICs come out, this option has the most profit. The Grin devs had explicitly stated that they did not want ASICs targeting the Cuckaroo algorithm, so choosing this option is a political risk.
Option 3: ASICs that can hit both Cuckatoo31 and Cuckatoo32. This option made the most sense if no other manufacturer made a Cuckatoo31 exclusive ASIC.
Option 4: ASICs that can hit exclusively Cuckatoo31. This option had the lowest total revenue, but ensured that the hardware would be competitive regardless of the decisions of other manufacturers.
If the goal of the Grin devs was a competitive landscape that could support multiple manufacturers, there should have been only one option. Since multiple manufacturers chose different options, it was unclear which products would be successful, and as a result none of the manufacturers could easily move forward.
Another significant barrier was the overall complexity of building a Cuckatoo31 solver. Though Cuckatoo31 was often advertised as a very simple algorithm, an optimal solver is anything but. The GRN1 chip design had thousands of hashing cores and memory banks, with a routing network that included two different types of sorting and multiple novel techniques for maintaining backpressure on our hashing cores and memory banks. In addition to this, the chip contains two different hashing functions (siphash and blake2b) as well as an entire on-die custom cycle finder.
Even with all of this complexity, substantial optimizations were left on the table. For example, as you perform edge trimming, you can begin to apply compression to the increasingly sparse graph, which improves memory usage and reduces the total memory requirements of the chip. As another example, the GRN1 had no analog optimizations in it at all. Simply by switching to an analog construction we could have significantly reduced the size, cost, and energy consumption of the chip while also improving the speed. We believe that successive generations of the GRN1 would have likely increased substantially in sophistication, and also greatly reduced the capability of competition to effectively enter the market.
We had hoped to be significant contributors to the Grin ecosystem, both in terms of providing hardware and network security, and also in terms of providing funding to the development team, however a number of factors both inside and outside of our control have ultimately led to the cancellation of the project.
We will be sending an email to GRN1 customers detailing how to receive your full refunds. While we are saddened to cancel the project, we are excited to bring future products to market. Please email us at hello@obelisk.tech with any questions or concerns.
