Bitmain: IPO and Technical Challenges

This blog post was originally written by Allen Guo from 2048 Capital and translated with his permission.

Abstract: During the market’s best two years, Bitmain did not roll out any new chips or new mining hardware to update its core Bitcoin mining product line, which can be fatal for a chip company in this fast-growing industry. This article takes you through the twists and turns of the path the king of hashrate took to climb – from a startup, to the top of the industry, before its product line fell behind. In the face of other uprising competitors, can Bitmain be reborn by branching out into the AI field and launching an IPO?

This is the third time that this article has been published. 2048 Capital is an offshore cryptocurrency investment fund. Our official WeChat account had previously been run by our business partner. After this article was published, the public relations team of Bitmain had requested that WeChat remove this article. Rejected by WeChat, Bitmain continued to exert pressure upon our business partner and deleted this article twice. Finally, after some effort, we took back the account, re-published this article, and began to share our findings and thoughts with the industry. The mining chip industry is the only area where the Chinese chip industry can lead the world. We hope that our efforts can provide more facts and insights for Chinese chip companies, miners, and investors; as well as reducimg vicious competition in the industry and promote the healthy development of China’s blockchain market in the long run. If you find that any part of this article is factually incorrect, please feel free to contact us and we will make corrections after verifying the content.

In the second half of 2018, there will be three Chinese mining chip companies preparing to go public, which is particularly interesting considering the recent China-US tariffs. Among these companies, Bitmain deserves a closer look. As a monopolist in the industry, the company has turned to AI and attempts to raise money before an IPO.

The Purpose of this Article

In the past year, we have mainly invested in blockchain infrastructure, public chains, tools, and transactions. To this day, we still have not bought any miners nor mined any bitcoins. We only recently started to pay attention to mining chips and Bitmain for the following reasons:

  • I had done some research on low-power-consumption digital chips and systems back in the days when I was studying at Berkeley. After many years of struggling in the Internet industry, I suddenly realized that the cornerstone of the blockchain industry is my old trade, computer chips.
  • Although I have been away from the chip industry for many years, recently there has been a plethora of new and old investors forwarding Bitmain’s fundraising slide deck materials to me and asking for my opinion. I can only try my best to provide them with my opinions.
  • Now that I have a preliminary understanding, what puzzles me is the fact that the flagship product for Bitcoin mining was launched back in 2016. During the market’s best two years, Bitmain did not release any new chips or miners to update its core Bitcoin mining product line. This is rare for a chip giant in this fast-growing industry.

Recently, Bitmain happens to be preparing for fundraising, and there have been piles of articles and comments out on the Internet. However, if you take a closer look you will find that most of these articles have misinterpreted the underlying logic of the chip industry. Focused only on a part of the company such as its financial data, technological advantages, products, and competencies have rarely been analyzed. Here I will write down my thoughts and results from research I’ve conducted these past few days which may not be complete, but will be as objective as possible.

Core Competencies of a Mining Chip Company

Any industry has its core competency. For a mining chip, there is only one core feature aside from its cost: the power consumption ratio. That is the power required to mine a bitcoin or the power consumption to support a hashrate of one terahash (W/THs). Mining chip companies are doing everything they can to reduce the W/THs value, and thus the power consumption of their miners.

There are two basic approaches to reducing W/THs. One is to optimize the design of chips to reduce power consumption in all aspects. However, possible power optimization of the chips is limited. The second approach will be to rely on the foundry to improve their processes from 28nm to 16nm, and from 16nm to 10nm or 7nm. The improved process will reduce power consumption, but will also increase the cost of chips. The “28nm” and “16nm” here refers to the width (line width) of the wires between the triodes on the chip. The smaller the line width is, the lower the power consumption and the better the performance; but the reduction in the line width is eventually limited by the laws of physics. At present, it is agreed that 0.5nm is the physical limit. TSMC has just kicked off its mass production of 7nm chips, and is scheduled to move to 5nm mass production in 2019.

The core competency of a mining chip company is the continuous update of its chip design, and improvement of foundry processes to reduce the power consumption ratio. Once a company stops updating and improving these, it will fall behind. The improvement in chip design depends on the strength of the chip development team, and the process improvement depends on the foundries.

Analysis of Different Generations of Bitmain Mining Hardware and their Competition

1. First Generation Miners

In August 2013, Bitmain taped out its first Bitcoin mining chip, the BM1380, using TSMC’s 55nm process. In October 2013, the test results of the chip were decent. The chip provided process nodes two generations ahead of its rival’s, Fried Cat (烤猫)’s BE100, and achieved lower power consumption. Acknowledged by miners, this chip achieved several hundred million RMB in sales, which helped Bitmain amass their original war chest.

First Generation Mining Hardware Specs
Chip Process Efficiency (W/TH)
BE100 130nm 9,353.8
Bitmain S1 (BM1380) 55nm 2,000

Data Sources:

https://www.cybtc.com/portal.php?aid=58&mod=view
https://www.cybtc.com/thread-11251-1-1.html

2. Another Big Win

After its first success, Bitmain was fully aware of the benefits of advanced technologies, and immediately started the development of their second Bitcoin mining chip, the BM1382, using the TSMC 28nm process.

The second chip, the BM1382, was taped in April 2014 and finished tape-out in June with a slightly better power consumption performance than that of Fried Cat (烤猫)’s BE200 with 40nm process node.

  Second Generation Mining Hardware Specs
Chip Process Efficiency (W/TH)
BE200 40nm 1,125
Bitmain S3 (BM1382) 28nm 773.8

Data Sources:

https://www.cybtc.com/article-1008-1.html
https://www.cybtc.com/article-1327-1.html
https://www.cybtc.com/article-1073-1.html

From a technical point of view, compared to Fried Cat (烤猫)’s BE200, the Bitmain chip did not have obvious advantages. However, as it happens, Fried Cat (烤猫)’s team decided to be a solutions provider instead. Thanks to this decision, Bitmain continued to lead the mining chip industry, with the second chip generating hundreds of millions in sales revenue for Bitmain.

3. Mysterious Disappearance of Rivals in the Face of Gloomy Business Prospects

The power consumption of the BM1382 did not meet expectations, so Bitmain continued to optimize the chip using the 28nm process. In October 2014, Bitmain taped out its third chip, the BM1384. Although the power consumption of the BM1384 was greatly optimized compared to that of the BM1382, it lagged behind Fried Cat (烤猫)’s 28nm BE300.

  Third Generation Mining Hardware Specs
Chip Process Efficiency (W/TH)
BE300 28nm 300
Bitmain S5 (BM1384) 28nm 448

Data Sources:

http://digi.163.com/15/0211/13/AI64KBVD001618JV.html
https://www.cybtc.com/article-1598-1.html
https://www.cybtc.com/article-1561-1.html

Despite this, Bitmain still ordered a large number of BM1384 wafers and invested in the production of a large number of S5 miners. The production of these S5 miners supposedly consumed all the profits accumulated before.

The Bitcoin market went down through 2014. The price of coins dropped to 2,000 RMB or so in August 2014, but the trend did not stop there. In January 2015, the price dropped further to a record low of less than 1,000 RMB. Bitmain, without exception, also experienced a rough time.

However, this trough marked a significant and dramatic turn in Bitmain’s fortunes, leading to dominance in the industry. During this bear market, Bitmain’s biggest rival, Fried Cat (烤猫), mysteriously disappeared, and their company, ASICMiner (深圳比特泉), was disbanded.

4. A Solo Show

In June 2015, Bitmain continued to use the 28nm process to tape out its BM1385 chip. Compared to the BM1384, the power consumption and cost of this chip were reduced by nearly 50%. The biggest difference between this chip and all other chips previously made by Bitmain was that this chip used a new design methodology, which had not come from its original chip design team:

  Fourth Generation Mining Hardware Specs
Chip Process Efficiency (W/TH)
Avalon A6 28nm 300
Bitmain S7 (BM1385) 28nm 257.3

Data Sources:

https://www.cybtc.com/article-2045-1.html
http://8btc.com/thread-25325-1-1.html

Compared with Avalon’s A6 which came three months after the BM1385, the BM1385 had obvious advantages in terms of both power consumption and cost. With the Bitcoin industry going up in 2015, the sales performance of Bitmain’s S7 mining machine was extraordinary. According to Internet data, Bitmain sold 200,000 models which generated two billion RMB in sales and a profit margin of about 80%; achieving a market share of 80%.

With its S7 miner, Bitmain outshone the entire mining machine industry.

5. Dominance and Prosperity

In December 2015, Bitmain used the TSMC 16nm process to tape out its BM1387 chip. This chip continued to use the design methodology mentioned above, resulting in a power consumption that was reduced by more than half from that of the BM1385. This offered Bitmain an advantage over Avalon, its rival at that time, as well as the rest of the industry.

  Fifth Generation Mining Hardware Specs
Chip Process Efficiency (W/TH)
Avalon A7 16nm 150
Bitmain S9 (BM1387) 16nm 110

Data Sources:

https://www.cybtc.com/article-2450-1.html
https://www.cybtc.com/article-2338-1.html

6. Hidden Dangers Among Expansion

Over the following two years, the Bitcoin market saw an upward trend, and Bitmain continued to dominate the Bitcoin mining community with its BM1387.

lionsden_blog_bitmain_ipo_price_chart

Above is a price chart of Bitcoin: which shows the price at 5,000 RMB at the beginning of 2017 to the highest point of more than 120,000 RMB at the end of that year.

From the beginning of 2017 to the first quarter of 2018, Bitmain sold more than three million miners (including more than two million S9 miners), and generated roughly 30 billion RMB in sales with a gross profit ratio of approximately 60%. For some months, the production capacity of Bitmain would account for more than 10% of TSMC’s production capacity, making Bitmain its second largest customer in mainland China after HiSilicon.

However, as sales continued to rise, we noticed two worrying realities.

First: Technological Stagnation

In the past two years, Bitmain’s products have fallen behind. The power consumption of most of Bitmain’s products have more than doubled compared to that of their competitors.

1. BTC Chip: Consumes 40% more power

Today, the Bitcoin mining chip is still the foundation of Bitmain’s business, but since 2016 its new designs of Bitcoin mining chips have met nothing but sheer failure:

  • In February 2017, the optimized BTC 16nm version of BM1X89 failed.
  • In December 2017, the BTC 12nm version of BM1X90 failed.
  • In March 2018, the BTC 10nm version of BM1X93 failed.

Therefore, Bitmain still relies on its BM1387 (the S9 miner) from early 2016 to support themselves, while rivals have outpaced its technological performance. For example, the ShenMa M10 could have a power consumption ratio 40% better than that of Bitmain’s.

Chip Process Efficiency (W/TH)
Bitmain S9 (BM1387) 16nm 110
Innosilicon T2 10nm 80
ShenMa M10 16nm 65

2. LTC Chip: Consumes 25% more power

Chip Efficiency (W/MH)
Bitmain L3 1.6
Innosilicon A6 1.2

Data Sources:
https://shop.bitmain.com/product/detail?pid=00020180723151802237tPzmvI8O06E5
http://www.88xindong.com/kuangji/33.html

3. DASH Chip: Consumes 2x More Power

Chip Efficiency (mW/MH)
Bitmain D3 57
Innosilicon A5+ 23

Data Sources:
https://shop.bitmain.com/product/detail?pid=00020180723152123156MoNf1JH9065
http://www.88xindong.com/kuangji/55.html

4. ZEC Chip: Consumes 2x more power

Chip Efficiency (W/KH)
Bitmain Z9 28
Innosilicon A9 12

Data Sources:
https://shop.bitmain.com/product/detail?pid=00020180710164901333EGb4bPcY0681
http://www.88xindong.com/kuangji/61.html

5. ETH Chip: Consumes 2x more power

Chip Efficiency (W/MH)
Bitmain E3 4
Innosilicon A10 2

Data Sources:
https://shop.bitmain.com/product/detail?pid=000201807231357009361z1imQTc06B3
http://www.it-bound.com/25808.html

6. SC Chip: Consumes 5x more power

Chip Efficiency (W/TH)
Bitmain A3 1,560
Innosilicon S11 320

Data Sources:
https://shop.bitmain.com/product/detail?pid=00020180723152506643yD7K4TLH0652芯动S11
http://www.88xindong.com/kuangji/58.html

7. AI Chip: Consumes 3x more power

Chip Efficiency (W/TH)
Bitmain BM1682 10
dadiannao 2.8

Data Sources:
http://news.hqew.com/info-320546
http://www.360doc.com/showarticle.aspx?id=689101971

Second: Gambling Away Money

If this rumor turns out to be true, Bitmain has recklessly gambled away 6 billion RMB in failed tapeouts.

1. BM1385 28nm Wafers for a Million Miners

In the fourth quarter of 2017, Bitmain sold out all of its miners thanks to the prosperous Bitcoin market. It is reported that Bitmain suddenly purchased BM1385 28nm wafers from TSMC for the production of one million miners to produce the new V9 model. At that time, the entire industry (including TSMC) was shocked by Bitmain’s move. After all, the BM1385 had been out of production for more than a year. Its historical mission had been completed, and it was no longer an advanced technology. The only reasonable explanation for this huge purchase of outdated chips seemed to be that Bitmain wanted to outrace competitors who still used the 28nm process, but had a better product design.

Miners soon understood that the V9 mining machine was actually the S7 mining machine, and the V9 models ended up overstocked. Eventually, more than one million V9 miners, worth 2 billion RMB, were gathering dust in Bitmain’s warehouse.

2. Risky BM1X90 12nm Wafers

It is known that at the end of 2017, Bitmain paid one billion RMB to purchase risky untested BM1X90 12nm wafers in hopes of catching the prosperous Bitcoin market trend, and expected a successful tapeout before mass production. In the end, it was nothing but a one billion RMB failure.

3. Risky BM1X93 10nm Wafers for a Million Miners

It is rumored that in early 2018, Bitmain continued to gamble away three billion RMB in risky BM1X93 10nm wafers for about one million miners, a three billion RMB failure.

Summary

The development of blockchain technology and cryptocurrency has given us amazing accumulations of wealth in a short period of time, and in a concentrated area. An important cornerstone of the confidence underpinning this accumulation is the cryptocurrency chip technology that drives the future of computing power. In the beginning, a technological disruption is often ignored or not recognized by the “mainstream.” Bitmain did make its share of contributions to the dominance of the Chinese mining industry. Where international chip giants have neglected or been unwilling to invest in cryptocurrency chips that they think are too trivial, Chinese entrepreneurs have keenly seized the historic opportunity and brought China’s global supply chain potential to a peak.

The key to Bitmain’s success is that the company took advantage of the chip process dividend, dared to invest in the most advanced manufacturing process, and once led in chip design innovation. The correct decision-making and seizing of opportunity in these two dimensions laid the foundation for its success over the past four years.

Over time, the landscape in these two dimensions has undergone major changes, and is quite different today. On the one hand, all of Bitmain’s competitors are aware of the importance of upgrading the manufacturing process, and the cognitive advantage in this area no longer exists. On the other hand, Bitmain faces stagnation or even regression in its ability to design core chips, which has become its Achilles heel. To make matters worse, the bold investment which brought Bitmain initial achievements, soon developed into blind gambling and even self-destructive suppression of rivals, which harms both the company’s cash flow and the competition and cooperative ecosystem of this young industry.

The core of a chip company will always be its chip design technology, and this core technology often comes from the founding team. The advantage enjoyed by Bitmain’s BM1385 and BM1387 came from its custom design methodology, which made full use of power consumption using the same process. However, as far as we know, this methodology did not come from the founding team of Bitmain. Ignorant of the importance of talent and technologies, Bitmain relinquished ownership of the core technology for competitive chip design after the design team left.

It is an inevitable trend for mining chip companies to expand into the new realm of AI, but this field has similar requirements for a technical team. Low power consumption is still the key, otherwise customers are not buying efficient computing power, but rather an electric oven. Bitmain also faces the same technical challenges in the AI ​​space.

Finally, a suggestion for those of you who have read this far; be you an insider or perhaps the chairman of Bitmain. Today, you can read in the media that “Zhan Ketuan (Micree) graduated from Tsinghua University,” which is not true. It could be a typo, or a mistake made by careless reporters, but now that he has put a lot of effort in fundraising and public relations, he’d better make a statement to the media to correct this as soon as possible; lest it become a source of ridicule someday. We believe that the background of a person is not important.

I hope that after its successful fundraising, the focus of Bitmain will one day return to technological innovation, being an industry ecosystem builder, and becoming a driving force in the new digital economy.

search previous next tag category expand menu location phone mail time cart zoom edit close