Original Article: http://onchain.caijing.com.cn/20180828/4506873.shtml
Disclaimer: The original news is written by journalist Ji Qian from Caijing. I am not responsible for publishing the original document. I’m not employed by Bitmain or any Bitmain related companies. I translated the original news on a volunteer personal interest basis in my spare time. I did this translation in a personal capacity. I did not get paid for doing the translation by any third party. This translation is not related to my current employer. This is simply a translation of the original article, it should not be regarded as a reference to Bitmain’s IPO. I have no comment regarding Bitmain’s IPO or anything related to Bitmain.
Translation takes time, if you want to read more amazing pieces, please tip here:
Bitmain and Its Patent Dispute With Whatsminer
Antminer S7, S9 chips have been flagship products for Bitmain. However, its design is not originally from the development team of Bitmain, but from Dr. Yang Zuoxing who graduated from Tsinghua University’s Department of Engineering Physics. Yang’s departure is one of the most important reasons for the technical bottleneck at Bitmain in recent years.
On August 26, according to a number of media reports such as Reuters, etc, Bitmain will submit an IPO application to the Hong Kong Stock Exchange this week to start the listing process. The IPO sponsor underwriter is CICC, and the listing target is valued at $300m.
Earlier, Bitmain’s second-quarter financial statement was leaked, with losses ranging from $600m to $700m. At the same time, Bitmain’s Pre-IPO investors have changed their minds again, and both Softbank and Tencent have denied investing in Bitmain.
Since 2015, Bitmain has been supporting its market share with S7 miner (1385 chips) and S9 miner (1387 chips). However, the core developer who designed these two chips is not from Bitmain’s development team, but was Dr. Yang Zuoxing who graduated from Tsinghua University Department of Engineering Physics. Yang’s departure is exactly one of the important reasons for the technical bottleneck in Bitmain in recent years.
Yang is the chairman of Shenzhen BitWei Technology Co., Ltd., and Whatsminer. This company was sued by Bitmain for infringement of patents.
Both parties have attended court many times. In the end, the Patent Reexamination Board of the State Intellectual Property Office gave a ruling, saying Bitmain’s patent was invalid.
- Previous Relationship: Timely Help
Dr. Yang was an employee of “Fried Cat” — Mr. Jiang Xinyu (Mr. Jiang was a Chinese bitcoin legend back to 2012–2015, but he disappeared during 2014–2015 bear market. No one has ever seen him nor located him since then). Yang designed a version of the BE300 chip with a full custom method, but because of Jiang’s disappearance, the chip designed was not able to be released to the public. Yang didn’t want this solution to be kept from the public, so he told Bitmain’s Micree Zhan the secret of the chip design — the implementation of the “full-customization methodology” which can reduce costs and energy consumption. This method can achieve the ultimate in power consumption at the same process node, and later Yang worked at Bitmain part time to help Bitmain survive the “Mining Winter”.
“In the bear market in early 2015, many people left the industry, but before leaving the industry, I was a bit unwilling. I think my full-customization methodology is especially suitable for bitcoin chips. I told Micree how to apply manual custom design methodology. After my introduction, Micree hoped that I could stay, and join Bitmain. Eventually I used the full-custom design method to help Bitmain design the S7 miner (1385 chip) and the S9 miner (1387 chip), these two chips.” Yang told Caijing.
Full custom methodology is one of the ways to design integrated circuits. Similar methods are custom, semi-custom, and programmable logic devices and logic cell array design methods.
“The benefits of the semiconductor process improvement are close to the benefits of adopting the manual custom design itself. Another reason is that the design is more and more complex and the scale is getting larger and larger. This makes the manual custom design cycle very long, which is not conducive to the rapid development of new products. In the past 20–30 years, almost no one has used custom methodology to design mass-produced chips.” Yang explained to Caijing.
After adopting FINFET from 16nm, the benefits brought by the new and old two-generation semiconductor processes are greatly reduced, by about 2 times. In the case of the reduced benefits of the process, it is necessary to improve the performance of the product through a manual custom design methodology. Yang’s main contribution is that in current ultra-large scale, the first one adopts a full-customization methodology, and within three to six months, the performance of the designed chip is better than the traditional APR (automatic layout) methodology. The designed performance is about an order of magnitude better. In the cryptocurrency industry, Yang doubled the performance by design when other competitors began to adopt the manual custom design methodology.
In June 2015, there was already a set of 1385 chip design schemes at Bitmain. Yang used a full-customized methodology to design a new scheme. The two schemes were finished at the same time. Eventually, Bitmain adopted the scheme proposed by Yang to produce S7 miner. Compared to Bitmain’s S5 miner chip BM1384, S7 chip’s power consumption and cost reduction is nearly 50%.
At that time, Yang wanted to apply the full customization method to a new chip. The project was about a passive wireless camera. Yang once again sought out Bitmain, hoping to get some financing, and at the same time, he continued to design the S9 for Bitmain. (1387 chip). Caijing tried to verify this with Bitmain but hasn’t received a response.
In the “mining winter” of 2015–2016, with the help of the S7 miner and S9 miner, which had obvious power consumption and cost advantages, ranked first in the industry and occupied more than 70% market share in the bitcoin mining machine market.
2. Gracefulness: Root of the Problem
After five rounds of negotiations, Yang failed to receive funds from Bitmain. On June 8, 2016, Yang’s negotiations with Bitmain was officially ended. He felt he didn’t get a return even though he helped a lot during Bitmain’s hard times, so he decided to set up his own company, design his own chips, and no longer provide development support for Bitmain.
“I left the specific implementation of the full-customization methodology to Bitmain. They probably knew what I was working on in general, but they don’t understand the details. It’s hard to continue to evolve based on my design. Technology is not static. It is constantly evolving. This is why there is no new product after I stopped working with them.” Yang told Caijing.
In July 2016, Yang established Shenzhen BitWei Technology Co., Ltd., whose main product is the Whatsminer, used for Bitcoin mining.
Since the termination of this cooperation, the industry №1 leader Bitmain has stopped or even retrograded on the BTC product line.
In the past two years, Bitmain has not introduced any new chips or new miners with subversive advantages like S7 and S9. Bitmain’s 16nm (nano) BM1X89 chip, 12nm BM1X90 chip, and 10nm BM1X93 chip design all failed. The 7nm process has also been used by many domestic and foreign manufacturers, and there is no longer a competitive advantage.
In addition, Bitmain’s Litecoin L3, Ethereum D3, Zcash Z9 miners and other technical indicators have also been overtaken by competitors such as Innosilicon. The AI chip Sophon BM1682, built by Bitmain, has more than three times the power consumption as Cambrian, and its power consumption is 10 (W/Ths), while the Cambrian dadiannao consumes only 2.8 (W/Ths).
Yang explained that the full-customization methodology has very high requirements for designers and requires first-class design engineers. After the project has been tempered, it can be slowly grasped. The optimization of the methodology is mainly reflected in four aspects: cell optimization, placement optimization, clock optimization and gate-level netlist optimization. From the Antminer S9 to the Whatsminer M10, under the same process conditions, performance has doubled. This is because some parts of the specific implementation of the methodology are constantly surpassing.
3. Sniper War: Patent War
Soon, Bitmain sued Shenzhen BitWei Technology for infringement of its patents and asked the supplier to stop cooperation with Shenzhen BitWei Technology in the indictment. On April 8, 2018, the Patent Reexamination Board of the State Intellectual Property Office declared that Bitmain’s patent was invalid.
“If a technical solution of a claim has a distinguishing technical feature from the prior art, but the distinguishing technical feature belongs to common knowledge in the art, it is obvious that the common general knowledge is incorporated into the prior art, and the claim is the technical solution is not creative.” This is the decision key point given by the Patent Reexamination Board of the State Intellectual Property Office in the instrument for reviewing the invalidation request (№35537).
The patent filed by Bitmain is called “serial power supply”, and is a basic technology frequently used by mining chip manufacturers. This patent was successfully applied by foreign technicians many years ago.
In 2013, Vasily, an engineer at Bitfury in Russia, read some patent documents on series power supply and realised that chips can also be connected in a series like LED lights. Vasily was inspired to use the serial power supply concept for the first time on a version of the mining chip.
When Vasily visited “Fried Cat” in Shenzhen, he brought this idea to China and used the series power supply concept on the “Prisma Miner”. Prisma Miner was the first product to use the technology of series power supply for publicly sold products in China.
In the same year, Bitmain’s S5 mining machine also adopted this technology and submitted a series power supply patent. Later, Yang’s Whatsminer used this technology, and was sued by Bitmain.
Yang hired two legal teams and requested the State Intellectual Property Office invalidate Bitmain’s patent. Yang also took out the existing patent evidence in China and abroad. Eventually, the Patent Reexamination Board of the State Intellectual Property Office decided to review the invalidation request (№35537). A ruling was given in an instrument that Bitmain did not comply with the “creative” requirement shown in Article 32 of Article 23 of the Patent Law of the People’s Republic of China. Bitmain’s “series power supply ideas, virtual currency mining machines and computer servers” patent was invalidated.
Earlier, Bitcoin Gold (BTG) founder and Lightning Intelligence CEO Jack Liao had also challenged Bitmain’s application for “ASIC BOOST” patents in China, saying the patent had been applied abroad. Several developers accused Bitmain of being a patent troll.
Why is it that many people in China and abroad have applied to patent this technology, yet Bitmain can continue to apply? Lawyer Zhou Han once responded on Baidu’s Knowledge that foreign patents were not enforceable in China and were not protected by Chinese patent law. “Beyond foreign countries, there is actually this patent in China. Because the big concept can cover small concepts, when applying for the original patent, it does not say which field, that is, all fields are acceptable, so it is actually impossible to apply for this patent in the field of innovation.” Yang said to Caijing.
4. Battle Drums: Competition is Heating Up
“In the early days, we made a bold investment in Bitmain, and it soon developed into a blind gamble that ignored the laws of technological development, and even evolved into a self-inflicted suppression of rivals. It’s not a good thing both for the company’s cash flow and for the industry-wide competition and cooperation ecology in the early stages of development.” 2048 Capital founder Allen Guo once analyzed the technical dilemma of Bitmain on 2048Capital WeChat public group.
Allen believes that the key to the success of Bitmain is to seize the dividend of the chip process, dare to invest in the most advanced manufacturing processes, and once again lead chip design innovation. Over time, all competitors are aware of the importance of improving the manufacturing process. On the other hand, Bitmain’s core chip design capability has stopped or even regressed to become the company’s Achilles heel.
Caijing.com has compiled some data on the mainstream Bitcoin miners. The power consumption of the AntMiner 1387 chip is 94 (W/Ths), and the Avalon A851 consumes 100 (W/Ths). Innosilicon’s 10nm chip T2T consumes 89 (W/Ths), and of course, its cost will also increase. For the same 16nm chip, Whatsminer’s M10 power consumption is only 65 (W/Ths) .
“The price of cryptos will keep dropping. The average profit of the market mining will drop to less than 30%. Miners only look at the price and energy consumption ratio, the lower the better.” Mining machine dealer Wu Wenjie shared with Caijing his own opinion on the mining machine market in the second half of the year.
Currently, three major mining machine manufacturers are “in the midst of a cold winter”, and all of them are in the midst of an IPO in Hong Kong. Bitmain will submit an IPO application to the Hong Kong Stock Exchange this week to start the listing process, while the latecomers also took out the “nuclear weapon”, or “waiting to eat the rest of the “cake”, in this mining machine market.
It is easy to start a war, but hard to defend your achievement. At any time, the core competitiveness of technology companies is technology. Competition in the second half of the 2018 mining machine market may be even more intense.