The Compute Report: Recent Developments & Future Forecasts
David Bellman
Key Takeaway #1
For some time, BitOoda has assessed that a convergence would occur among Bitcoin mining, Ethereum and other blockchain technologies, high-performance computing (HPC) (which is fueling the growth of artificial intelligence and large language models, or LLMs) and the power markets.
Key Takeaway #2
This vision appears to be materializing as Bitcoin miners enter the HPC space, data centers proliferate, AI demand continues to explode, and access to power becomes increasingly constrained for all participants.
Key Takeaway #3
In this report, we will provide an overview of recent developments across the digital asset and compute landscape and some forecasts for its future.
Key Takeaway #4
The report is divided into sections on Bitcoin, Ethereum and other blockchains, and HPC.
For some time, BitOoda has assessed that a convergence would occur among Bitcoin mining, Ethereum and other blockchain technologies, high-performance computing (HPC) (which is fueling the growth of artificial intelligence and large language models, or LLMs) and the power markets. This vision appears to be materializing as Bitcoin miners enter the HPC space, data centers proliferate, AI demand continues to explode, and access to power becomes increasingly constrained for all participants.
In this report, we will provide an overview of recent developments across the digital asset and compute landscape and some forecasts for its future.
The report is divided into sections on Bitcoin, Ethereum and other blockchains, and HPC. We begin with Bitcoin.
Premium Content
Unlock exclusive insights with our cutting-edge digital finance platform. Gain access to next-gen data analytics and digital asset products crafted with applied science. Subscribe now to stay ahead of the curve.
For some time, BitOoda has assessed that a convergence would occur among Bitcoin mining, Ethereum and other blockchain technologies, high-performance computing (HPC) (which is fueling the growth of artificial intelligence and large language models, or LLMs) and the power markets. This vision appears to be materializing as Bitcoin miners enter the HPC space, data centers proliferate, AI demand continues to explode, and access to power becomes increasingly constrained for all participants.
In this report, we will provide an overview of recent developments across the digital asset and compute landscape and some forecasts for its future.
The report is divided into sections on Bitcoin, Ethereum and other blockchains, and HPC. We begin with Bitcoin.
Bitcoin Market Update:
• Surpassing $66,000: It took 846 days after Bitcoin’s previous all-tie high price on November 9, 2021 to once again surpass the $66,000 mark, where it remained for 10 days.
• Pre-Halving Mining Activity: With the halving event approaching (currently estimated for April 18, 2024), the block issuance rate is slightly ahead of schedule. This acceleration likely stems from two factors:
(1) Increased margins:Bitcoin prices have climbed faster than the hash rate, improving margins for all types of miners.
(2) Increased mining activity:Miners are pushing to complete projects before the halving to maximize their earnings.
• ETF Inflows and Market Growth:The launch of the first Bitcoin ETFs has witnessed unprecedented inflows, indicating quickly-growing investor interest. This trend is expected to continue as awareness and understanding of these investment vehicles expands. So far, the ETFs have expanded to around $60 billion in market cap.
Since the pre-launch phase, the market cap of Bitcoin has grown from $900 billion to $1.4 trillion (see next slide), reflecting the positive impact of ETFs, although ETFs represent only about $60 billion of the $500 billion increase in total BTC market cap. We conservatively estimate that a 2% allocation by North American fiduciaries (which manage around $53 trillion in assets) could push the market capitalization to $1.9 trillion, even with just ETF flow.
• Halving Update:We updated our halving analysis to compare our forecasted S21 specifications from last year with the actual specifications. Notably, our initial expectations closely aligned with reality (see slide 3).
• The first tranche of BTC ETFs launched when BTC market cap was $900 billion – it is now over $1.2 trillion.
• ETF inflows have soared to around $60 billion – still a far cry from the potential 2% penetration of North American fiduciaries that manage $53 trillion in assets.
Sources: Glassnode, Bloomberg
ASICS Over Time: Moore’s Law Applies to ASICS
• Hashrate per device is a design choice; we urge miners to focus on efficiency and hashrate per MW.
• Continued improvements expected for ASICS.
Sources: Glassnode, Bloomberg
Ethereum and Other Blockchains: Hard Fork Reduces L2 Fees in a Major Step toward Mass Adoption
The Ethereum ecosystem last week took its largest step toward mass adoption Ethereum’s most ambitious and most complex hard fork since the “Merge” (the transition from Proof of Work to Proof of Stake) was successfully activated last week. This hard fork, called “Dencun,” introduced a new structure called a “blob,” which drastically reduces the settlement cost for Ethereum L2s. Why are blobs significant?
Ethereum’s final form is for the base L1 layer to be the most secure, the most decentralized, and the most accessible settlement layer upon which a global digital economy can function. However, security and decentralization come at a cost: slower speeds and higher cost.
As a result, the Ethereum roadmap has been to modularize, or harden its L1, ossifying its security and decentralization while outsourcing “execution,” or the user layer, to an ecosystem of “Layer 2” blockchains, or L2s. These L2s can be fully customized (meaning they can integrate privacy, faster speeds, regulatory requirements like KYC, and more), while inheriting the security of Ethereum L1.
The only way for Ethereum (or any L1 blockchain) to sustainably scale to billions of users is to modularize via adding layers (the L2 roadmap). In the long run, all L1s will have L2 execution environments – we are already seeing this on Bitcoin and other competing smart contract blockchains.
The most prominent example of an L2 ecosystem is Base, created by Coinbase, which shows the potential of building a regulated L2 ecosystem that is intricately linked to a company.
However, L2 fees have still been too high ($1+ per transaction). This is due to congestion in storage capacity at Ethereum L1 (called data availability). The Dencun Hard Fork enabled a fundamental change – allowing L2s to use blobs for data availability.
The bottom line is that the recent Ethereum hard fork has reduced fees for L2s by an order of magnitude, with transaction costs on many prominent L2s reaching sub-10 cent (and sometimes sub-cent) levels. This is the first major step toward mass adoption and true accessibility of the L2 ecosystem.
Why does Ethereum focus on scaling via L2s versus increasing capacity and speed and lowering fees on L1? Earlier this month, we released a BitOoda white paper outlining the Ethereum institutional thesis.
The key assumption of the white paper is that the core function of a blockchain is to offer more robust security than a database, which necessitates decentralization. This results in a tradeoff of slower speeds and higher fees. The white paper then delves into 8 institutional theses about Ethereum as an ecosystem and ETH as the base asset securing the ecosystem. We welcome client feedback on the white paper, especially as the institutional focus could shift to ETH as the ETF approval deadline approaches in May.
In other L1 news, Solana’s native token, SOL, saw a major price move upward over the past week, reaching nearly $200 from its previous support level around ~$140. This has been largely driven by a surge in “meme coin” activity in the Solana ecosystem, with several meme coins launching and driving new users to the ecosystem. Solana, which has been designed to keep all users at the L1 level, will be a key ecosystem to watch as a major competitor to Ethereum’s L2 ecosystem.
Ethereum and Other Blockchains: Hard Fork Reduces L2 Fees in a Major Step toward Mass Adoption
• ETH has the potential to appreciate faster than BTC, given the history of the ratio.
• A surge in meme coins resulted in a surge in Solana prices.
Sources: Glassnode, Yahoo Finance
HPC: Optimization in AI: Leveraging Software and Hardware Innovations for Enhanced Performance
In the rapidly evolving field of AI, particularly in the realm of LLMs, the demand for high-performing chips has skyrocketed. NVIDIA, a leading provider of these essential components, faces supply constraints due to its reliance on TSMC and a complex geopolitical landscape. Despite these challenges, significant strides in both software and hardware innovations are driving efficiencies, maximizing the potential of limited GPU resources. This section addresses these advancements and highlights their pivotal role in the current landscape of AI optimization.
Software Optimizations: Empowering Smaller Entities and Individuals
Recent developments such as Low-Rank Adaptation (LORA) and its quantized counterpart (qLORA) have been game-changers for LLMs. These innovations enable smaller companies and individuals to fine-tune existing open-source models on commercial hardware. This approach sidesteps the need for resource-intensive training from scratch, fueling the progress of LLMs over the past 18 months.
A noteworthy breakthrough in this domain is Answer.AI's latest open-source project, which merges Fully Sharded Data Parallel (FDSP) with qLORA. This combination facilitates the training of substantial 70-billion-parameter models on commercial gaming GPUs, such as two NVIDIA 3090s or 4090s. Previously, such tasks demanded specialized hardware like the A100. This development democratizes access to LLM experimentation, eliminating the need for exorbitant investments in top-tier GPUs.
Additionally, researchers like Josh Alma, Virginia Williams, Yinzhan Xu, and R Zhou are making headway in optimizing matrix multiplications—a cornerstone operation in machine learning. These advancements promise to further revolutionize the efficiency of LLMs.
Hardware Optimizations: Paving New Pathways
On the hardware front, Jonathan Ross, who was pivotal in the development of Google's TPU, has founded Groq. His company introduces a novel 'language processing unit,' a chip expressly designed for language model inference. Remarkably, despite being a 14 nm chip (in contrast to NVIDIA's 4 nm offerings), it offers superior speed, scalability, and reduced operational costs. Groq's innovation presents a compelling alternative to NVIDIA's leading GPUs.
Consumer hardware is also witnessing exciting developments, with devices like Truffle 1 promising significantly faster inference times. Such advancements are not just incremental improvements; they represent significant leaps forward in AI hardware technology.
Conclusion: Navigating the Chip Shortage with Innovation
The ongoing chip shortage underscores the importance of these software and hardware optimizations in AI. By enhancing the capabilities of existing resources and introducing novel, efficient alternatives, the AI community continues to make remarkable strides in a constrained environment. These developments not only address immediate challenges, but also lay the groundwork for future innovations in AI optimization and accessibility.
HPC: Hardware & Software Acceleration
Hardware Optimization: The image on the right shows how inference on Groq's LPUs is much quicker than other solutions, highlighting how specialized architecture can be useful. The unit is Tokens per second.
Hardware Optimization: The image on the right shows how inference on Groq's LPUs is much quicker than other solutions, highlighting how specialized architecture can be useful. The unit is Tokens per second.
Disclosures
Purpose
This research is only for the clients of BitOoda. This research is not intended to constitute an offer, solicitation, or invitation for any securities and may not be distributed into jurisdictions where it is unlawful to do so. For additional disclosures and information, please contact a BitOoda representative at info@bitooda.io.
Analyst Certification
David Bellman, the research analyst denoted by an “AC” on the cover of this report, hereby certifies that all of the views expressed in this report accurately reflect his personal views, which have not been influenced by considerations of the firm’s business or client relationships.
Conflicts of Interest
This research contains the views, opinions, and recommendations of BitOoda. This report is intended for research and educational purposes only. We are not compensated in any way based upon any specific view or recommendation.
General Disclosures
Any information (“Information”) provided by BitOoda Holdings, Inc., BitOoda Digital, LLC, BitOoda Technologies, LLC or Ooda Commodities, LLC and its affiliated or related companies (collectively, “BitOoda”), either in this publication or document, in any other communication, or on or throughhttp://www.bitooda.io/, including any information regarding proposed transactions or trading strategies, is for informational purposes only and is provided without charge. BitOoda is not and does not act as a fiduciary or adviser, or in any similar capacity, in providing the Information, and the Information may not be relied upon as investment, financial, legal, tax, regulatory, or any other type of advice. The Information is being distributed as part of BitOoda’s sales and marketing efforts as an introducing broker and is incidental to its business as such.BitOoda seeks to earn execution fees when its clients execute transactions using its brokerage services. BitOoda makes no representations or warranties (express or implied) regarding, nor shall it have any responsibility or liability for the accuracy, adequacy, timeliness or completeness of, the Information, and no representation is made or is to be implied that the Information will remain unchanged. BitOoda undertakes no duty to amend, correct, update, or otherwise supplement the Information.
The Information has not been prepared or tailored to address, and may not be suitable or appropriate for the particular financial needs, circumstances or requirements of any person, and it should not be the basis for making any investment or transaction decision. The Information is not a recommendation to engage in any transaction. The digital asset industry is subject to a range of inherent risks, including but not limited to: price volatility, limited liquidity, limited and incomplete information regarding certain instruments, products, or digital assets, and a still emerging and evolving regulatory environment. The past performance of any instruments, products or digital assets addressed in the Information is not a guide to future performance, nor is it a reliable indicator of future results or performance.
All derivatives brokerage is conducted byOoda Commodities, LLC a member of NFA and subject to NFA’s regulatory oversight and examinations. However, you should be aware that NFA does not have regulatory oversight authority over underlying or spot virtual currency products or transactions or virtual currency exchanges, custodians or markets.
BitOoda Technologies, LLC is a member of FINRA.
“BitOoda”, “BitOoda Difficulty”, “BitOoda Hash”, “BitOoda Compute”, and the BitOoda logo are trademarks of BitOoda Holdings, Inc.
Copyright 2024 BitOoda Holdings, Inc. All rights reserved. No part of this material may be reprinted, redistributed, or sold without prior written consent of BitOoda.
For some time, BitOoda has assessed that a convergence would occur among Bitcoin mining, Ethereum and other blockchain technologies, high-performance computing (HPC) (which is fueling the growth of artificial intelligence and large language models, or LLMs) and the power markets. This vision appears to be materializing as Bitcoin miners enter the HPC space, data centers proliferate, AI demand continues to explode, and access to power becomes increasingly constrained for all participants.
In this report, we will provide an overview of recent developments across the digital asset and compute landscape and some forecasts for its future.
The report is divided into sections on Bitcoin, Ethereum and other blockchains, and HPC. We begin with Bitcoin.
Bitcoin Market Update:
• Surpassing $66,000: It took 846 days after Bitcoin’s previous all-tie high price on November 9, 2021 to once again surpass the $66,000 mark, where it remained for 10 days.
• Pre-Halving Mining Activity: With the halving event approaching (currently estimated for April 18, 2024), the block issuance rate is slightly ahead of schedule. This acceleration likely stems from two factors:
(1) Increased margins:Bitcoin prices have climbed faster than the hash rate, improving margins for all types of miners.
(2) Increased mining activity:Miners are pushing to complete projects before the halving to maximize their earnings.
• ETF Inflows and Market Growth:The launch of the first Bitcoin ETFs has witnessed unprecedented inflows, indicating quickly-growing investor interest. This trend is expected to continue as awareness and understanding of these investment vehicles expands. So far, the ETFs have expanded to around $60 billion in market cap.
Since the pre-launch phase, the market cap of Bitcoin has grown from $900 billion to $1.4 trillion (see next slide), reflecting the positive impact of ETFs, although ETFs represent only about $60 billion of the $500 billion increase in total BTC market cap. We conservatively estimate that a 2% allocation by North American fiduciaries (which manage around $53 trillion in assets) could push the market capitalization to $1.9 trillion, even with just ETF flow.
• Halving Update:We updated our halving analysis to compare our forecasted S21 specifications from last year with the actual specifications. Notably, our initial expectations closely aligned with reality (see slide 3).
• The first tranche of BTC ETFs launched when BTC market cap was $900 billion – it is now over $1.2 trillion.
• ETF inflows have soared to around $60 billion – still a far cry from the potential 2% penetration of North American fiduciaries that manage $53 trillion in assets.
Sources: Glassnode, Bloomberg
ASICS Over Time: Moore’s Law Applies to ASICS
• Hashrate per device is a design choice; we urge miners to focus on efficiency and hashrate per MW.
• Continued improvements expected for ASICS.
Sources: Glassnode, Bloomberg
Ethereum and Other Blockchains: Hard Fork Reduces L2 Fees in a Major Step toward Mass Adoption
The Ethereum ecosystem last week took its largest step toward mass adoption Ethereum’s most ambitious and most complex hard fork since the “Merge” (the transition from Proof of Work to Proof of Stake) was successfully activated last week. This hard fork, called “Dencun,” introduced a new structure called a “blob,” which drastically reduces the settlement cost for Ethereum L2s. Why are blobs significant?
Ethereum’s final form is for the base L1 layer to be the most secure, the most decentralized, and the most accessible settlement layer upon which a global digital economy can function. However, security and decentralization come at a cost: slower speeds and higher cost.
As a result, the Ethereum roadmap has been to modularize, or harden its L1, ossifying its security and decentralization while outsourcing “execution,” or the user layer, to an ecosystem of “Layer 2” blockchains, or L2s. These L2s can be fully customized (meaning they can integrate privacy, faster speeds, regulatory requirements like KYC, and more), while inheriting the security of Ethereum L1.
The only way for Ethereum (or any L1 blockchain) to sustainably scale to billions of users is to modularize via adding layers (the L2 roadmap). In the long run, all L1s will have L2 execution environments – we are already seeing this on Bitcoin and other competing smart contract blockchains.
The most prominent example of an L2 ecosystem is Base, created by Coinbase, which shows the potential of building a regulated L2 ecosystem that is intricately linked to a company.
However, L2 fees have still been too high ($1+ per transaction). This is due to congestion in storage capacity at Ethereum L1 (called data availability). The Dencun Hard Fork enabled a fundamental change – allowing L2s to use blobs for data availability.
The bottom line is that the recent Ethereum hard fork has reduced fees for L2s by an order of magnitude, with transaction costs on many prominent L2s reaching sub-10 cent (and sometimes sub-cent) levels. This is the first major step toward mass adoption and true accessibility of the L2 ecosystem.
Why does Ethereum focus on scaling via L2s versus increasing capacity and speed and lowering fees on L1? Earlier this month, we released a BitOoda white paper outlining the Ethereum institutional thesis.
The key assumption of the white paper is that the core function of a blockchain is to offer more robust security than a database, which necessitates decentralization. This results in a tradeoff of slower speeds and higher fees. The white paper then delves into 8 institutional theses about Ethereum as an ecosystem and ETH as the base asset securing the ecosystem. We welcome client feedback on the white paper, especially as the institutional focus could shift to ETH as the ETF approval deadline approaches in May.
In other L1 news, Solana’s native token, SOL, saw a major price move upward over the past week, reaching nearly $200 from its previous support level around ~$140. This has been largely driven by a surge in “meme coin” activity in the Solana ecosystem, with several meme coins launching and driving new users to the ecosystem. Solana, which has been designed to keep all users at the L1 level, will be a key ecosystem to watch as a major competitor to Ethereum’s L2 ecosystem.
Ethereum and Other Blockchains: Hard Fork Reduces L2 Fees in a Major Step toward Mass Adoption
• ETH has the potential to appreciate faster than BTC, given the history of the ratio.
• A surge in meme coins resulted in a surge in Solana prices.
Sources: Glassnode, Yahoo Finance
HPC: Optimization in AI: Leveraging Software and Hardware Innovations for Enhanced Performance
In the rapidly evolving field of AI, particularly in the realm of LLMs, the demand for high-performing chips has skyrocketed. NVIDIA, a leading provider of these essential components, faces supply constraints due to its reliance on TSMC and a complex geopolitical landscape. Despite these challenges, significant strides in both software and hardware innovations are driving efficiencies, maximizing the potential of limited GPU resources. This section addresses these advancements and highlights their pivotal role in the current landscape of AI optimization.
Software Optimizations: Empowering Smaller Entities and Individuals
Recent developments such as Low-Rank Adaptation (LORA) and its quantized counterpart (qLORA) have been game-changers for LLMs. These innovations enable smaller companies and individuals to fine-tune existing open-source models on commercial hardware. This approach sidesteps the need for resource-intensive training from scratch, fueling the progress of LLMs over the past 18 months.
A noteworthy breakthrough in this domain is Answer.AI's latest open-source project, which merges Fully Sharded Data Parallel (FDSP) with qLORA. This combination facilitates the training of substantial 70-billion-parameter models on commercial gaming GPUs, such as two NVIDIA 3090s or 4090s. Previously, such tasks demanded specialized hardware like the A100. This development democratizes access to LLM experimentation, eliminating the need for exorbitant investments in top-tier GPUs.
Additionally, researchers like Josh Alma, Virginia Williams, Yinzhan Xu, and R Zhou are making headway in optimizing matrix multiplications—a cornerstone operation in machine learning. These advancements promise to further revolutionize the efficiency of LLMs.
Hardware Optimizations: Paving New Pathways
On the hardware front, Jonathan Ross, who was pivotal in the development of Google's TPU, has founded Groq. His company introduces a novel 'language processing unit,' a chip expressly designed for language model inference. Remarkably, despite being a 14 nm chip (in contrast to NVIDIA's 4 nm offerings), it offers superior speed, scalability, and reduced operational costs. Groq's innovation presents a compelling alternative to NVIDIA's leading GPUs.
Consumer hardware is also witnessing exciting developments, with devices like Truffle 1 promising significantly faster inference times. Such advancements are not just incremental improvements; they represent significant leaps forward in AI hardware technology.
Conclusion: Navigating the Chip Shortage with Innovation
The ongoing chip shortage underscores the importance of these software and hardware optimizations in AI. By enhancing the capabilities of existing resources and introducing novel, efficient alternatives, the AI community continues to make remarkable strides in a constrained environment. These developments not only address immediate challenges, but also lay the groundwork for future innovations in AI optimization and accessibility.
HPC: Hardware & Software Acceleration
Hardware Optimization: The image on the right shows how inference on Groq's LPUs is much quicker than other solutions, highlighting how specialized architecture can be useful. The unit is Tokens per second.
Hardware Optimization: The image on the right shows how inference on Groq's LPUs is much quicker than other solutions, highlighting how specialized architecture can be useful. The unit is Tokens per second.
Disclosures
Purpose
This research is only for the clients of BitOoda. This research is not intended to constitute an offer, solicitation, or invitation for any securities and may not be distributed into jurisdictions where it is unlawful to do so. For additional disclosures and information, please contact a BitOoda representative at info@bitooda.io.
Analyst Certification
David Bellman, the research analyst denoted by an “AC” on the cover of this report, hereby certifies that all of the views expressed in this report accurately reflect his personal views, which have not been influenced by considerations of the firm’s business or client relationships.
Conflicts of Interest
This research contains the views, opinions, and recommendations of BitOoda. This report is intended for research and educational purposes only. We are not compensated in any way based upon any specific view or recommendation.
General Disclosures
Any information (“Information”) provided by BitOoda Holdings, Inc., BitOoda Digital, LLC, BitOoda Technologies, LLC or Ooda Commodities, LLC and its affiliated or related companies (collectively, “BitOoda”), either in this publication or document, in any other communication, or on or throughhttp://www.bitooda.io/, including any information regarding proposed transactions or trading strategies, is for informational purposes only and is provided without charge. BitOoda is not and does not act as a fiduciary or adviser, or in any similar capacity, in providing the Information, and the Information may not be relied upon as investment, financial, legal, tax, regulatory, or any other type of advice. The Information is being distributed as part of BitOoda’s sales and marketing efforts as an introducing broker and is incidental to its business as such.BitOoda seeks to earn execution fees when its clients execute transactions using its brokerage services. BitOoda makes no representations or warranties (express or implied) regarding, nor shall it have any responsibility or liability for the accuracy, adequacy, timeliness or completeness of, the Information, and no representation is made or is to be implied that the Information will remain unchanged. BitOoda undertakes no duty to amend, correct, update, or otherwise supplement the Information.
The Information has not been prepared or tailored to address, and may not be suitable or appropriate for the particular financial needs, circumstances or requirements of any person, and it should not be the basis for making any investment or transaction decision. The Information is not a recommendation to engage in any transaction. The digital asset industry is subject to a range of inherent risks, including but not limited to: price volatility, limited liquidity, limited and incomplete information regarding certain instruments, products, or digital assets, and a still emerging and evolving regulatory environment. The past performance of any instruments, products or digital assets addressed in the Information is not a guide to future performance, nor is it a reliable indicator of future results or performance.
All derivatives brokerage is conducted byOoda Commodities, LLC a member of NFA and subject to NFA’s regulatory oversight and examinations. However, you should be aware that NFA does not have regulatory oversight authority over underlying or spot virtual currency products or transactions or virtual currency exchanges, custodians or markets.
BitOoda Technologies, LLC is a member of FINRA.
“BitOoda”, “BitOoda Difficulty”, “BitOoda Hash”, “BitOoda Compute”, and the BitOoda logo are trademarks of BitOoda Holdings, Inc.
Copyright 2024 BitOoda Holdings, Inc. All rights reserved. No part of this material may be reprinted, redistributed, or sold without prior written consent of BitOoda.
