Exascale Power Co., a Portland, Ore.-based company, plans to build a supercomputer in New Mexico that will boost the power of the block chain processing infrastructure. The “One ExaFlop” system will bring parity with the computational power of machine systems and the human brain, allowing…
Exascale Power Co., a Portland, Ore.-based company, plans to build a supercomputer in New Mexico that will boost the power of the block chain processing infrastructure. The “One ExaFlop” system will bring parity with the computational power of machine systems and the human brain, allowing for the dawn of a new era of true artificial intelligence, the company claims.
Phase two will be a “One Zettaflop” supercomputer, which is one thousand times larger than the Exaflop system.
“Our mission is to build large data centers with wholesale power contracts for economies of scale, allowing us to provide competitively-priced commodity cloud computing services to improve the human condition,” John Fitzpatrick, CEO of Exascale Power Co, told CCN.
The primary purpose of each system is to offer competitively-priced commodity cloud computing services to industry. Economies of scale will allow Exascale to price lower than Amazon, Google, Microsoft and all others per CPU hour, Fitzpatrick said.
The secondary purpose of each system is to accelerate computational research in medicine to cure diseases, and energy research to develop clean, cheap fusion power to improve the human condition.
The tertiary purpose of each system is to scale Bitcoin block chain processing for wider adoption in the global financial system, Fitzpatrick said.
“We will have ready access to nearly the entire electrical grid in the United States, and we will get the very best price at any moment in time as a volume buyer of power,” said Fitzpatrick noting that New Mexico is at the intersection of the three major power grids in the U.S.
The Exascale power system architecture includes Mellanox Infiniband fiber, Nvidia chipsets, IBM open source power PC chips and custom ASICS chips for block chain processing.
With ASICS processors, the bitcoin community has already reached an aggregate global capacity of One Exaflop in block chain processing power, he said.
“The barrier to reaching One Exaflop on Linpack with the technology of today is a 1,000-MW power supply,” Fitzpatrick said, “and with the New Mexico location, we will have the ability to scale to Zettaflop capacity in phase two.”
Scaling of infrastructure is critical to realizing the promise of block chain processing for the global financial system, he noted.
“The enabling part of building out this bitcoin infrastructure is the mining fees,” Fitzpatrick said, “Any bitcoin miner has the seignorage power of a sovereign nation to create money.”
Not only does an exascale system require enormous electrical power, but also enormous financial resources for the projected cost of $50 billion, Fitzpatrick said. The company has filed a debt offering with the Securities and Exchange Commission. “Our rights offering is for debt, convertible to equity at buyer discretion,” Fitzpatrick said. “Our lenders own us.”
The company has a core group of 50 engineers, with staffing of 2,500 required at full system build-out,” Fitzpatrick said, noting the company is currently hiring.
The project marks Exascale’s first foray into bitcoin and block chain processing.
The company brings a 15-year history in financial services and supercomputing.
According to the MIT Technology Review, Fitzpatrick in 2002 was president of JJX Capital, a startup hedge fund company. According to the article, JJX Capital was developing a computer to predict the future price movements of every stock, bond, and commodity traded in the U.S.
According to the Institute of Electrical and Electronics Engineers (IEEE), the U.S. probably won’t build its first practical exascale supercomputer until 2023. To hit that target, engineers will need to do three things. First they’ll need new computer architectures capable of combining tens of thousands of CPUs and graphics-processor-based accelerators. Engineers will also need to deal with the growing energy costs required to move data from a supercomputer’s memory to the processors. Finally, software developers will have to learn how to build programs that can make use of the new architecture.
Last modified: January 25, 2020 11:05 PM UTC