- The launch of DeepSeek’s AI model has been called a "Sputnik moment,” highlighting an international race for AI supremacy as well as the power of open-source technology.
- At the same time, the news highlights risks associated with centralized AI development, like data security, biases, and censorship. Risks associated with centralized AI companies can potentially be addressed by blockchain-based AI platforms like Bittensor.
- Bittensor helps facilitate the development of open and global AI systems through the use of decentralized networks and economic incentives. By leveraging blockchain technology and a global network of participants, platforms like Bittensor can potentially increase transparency, democratize access, and distribute ownership of AI systems.
- Grayscale Research believes DeepSeek's emergence could potentially reduce costs and lower barriers to entry for open-source decentralized AI projects.
What happened
Recently, China-based startup DeepSeek launched an open-source Artificial Intelligence (AI) model that matches or surpasses leading models like OpenAI’s o1 in performance.[1] Impressively, DeepSeek achieved this with significantly less computing resources and reportedly spending just around $5 million to train the model — a fraction of the hundreds of millions spent on training by OpenAI.[2] By January 27, DeepSeek overtook OpenAI’s ChatGPT on the Apple App Store rankings.[3]
Tech leaders are calling this AI's "Sputnik moment"— we are potentially witnessing a modern Space Race between China and the U.S. in artificial intelligence. [4] The emergence of DeepSeek caused a historic sell-off in tech stocks, with Nvidia, Microsoft, and others losing billions in value, as it forced investors to possibly reconsider previous assumptions about this powerful emerging technology. [5]
However, while the DeepSeek breakthroughs show the power of open-source AI, they also have brought greater attention to the risks of centralized control and development of AI technology. Immediately following news on the company’s new model performance, DeepSeek fell victim to a large-scale cyberattack, prompting the company to temporarily restrict user sign-ups.[6] This incident highlights vulnerabilities inherent in centralized systems — including the risk that cyberattacks can interrupt service. Distributed systems can potentially enhance network resilience by spreading responsibility across multiple entities. Decentralized development of AI models may also help reduce bias and improve transparency for this crucial technology.
In this piece, we’ll explore these risks related to the development of centralized AI and detail how decentralized AI platforms like Bittensor can aim to address them. We’ll also explore Bittensor’s progress to date and the potential implications of DeepSeek on broader development within decentralized AI.
The Risks of Centralized AI
Network effects and intensive capital requirements have led to many AI developers outside of large tech companies, such as small companies or academic researchers, either having difficulty gaining access to needed resources for AI development or being unable to monetize their work. This may limit overall AI competition and innovation.
Influence over this critical technology is therefore largely concentrated in the hands of a few tech giants, leading to serious questions about censorship and bias. For example, in February 2024, Google’s AI image generator Gemini revealed racial biases and historical inaccuracies, illustrating how companies can manipulate their models.[7] Notably, these concerns extend to DeepSeek. Some have claimed that DeepSeek’s model has “clear pro-Chinese Communist Party bias” as it censors certain content critical of the government — for example, it refuses to respond to questions regarding the 1989 Tiananmen Square incident or Chinese President Xi Jinping.[8]
This begs broader questions about AI governance. A small handful of people wield control over the companies developing the few models that may increasingly shape and influence society. As AI grows in influence and importance, many worry that one company or government could hold decision-making power over the AI models that have an outsize influence on society, potentially imposing guardrails, operating behind closed doors, or manipulating models to their benefit — but at the expense of the rest of society.
How do we ensure that we can trust the models we use with our data? Lacking true transparency — and with the stakes so high — how can we trust that these innovative technologies are being built in our best interests and not at our expense?
Decentralized AI and Bittensor
Enter decentralized AI: a potential solution to these challenges. By leveraging blockchain technology and a global network of participants, platforms like Bittensor can increase transparency, democratize access, and distribute ownership of AI systems.
Grayscale Research believes that decentralized AI holds the potential to bring important decisions regarding AI development out from walled gardens and into public ownership. We believe that Bittensor offers a compelling solution as a key decentralized AI platform poised to help address these risks and provide a viable alternative to centralized AI incumbents.
What is Bittensor
Bittensor is a platform that helps facilitate the development of open and global artificial intelligence systems using decentralized networks and economic incentives. It aims to create an "Internet of AI" with interconnected ecosystems called “subnets,” each focusing on a different specific use case. Currently, Bittensor has over 50 subnets, spanning a wide range of applications and use cases, including video generation, AI agents, and deepfake detection. Here is how Bittensor attempts to address the concerns related to centralized AI:
- Aligning Economic Incentives: Centralized AI companies prioritize shareholder value and profit, often resulting in value extraction from users. In contrast, through use of its TAO token, Bittensor aligns incentives between ecosystem participants, including its users and tokenholders.
- Permissionless Access to Build and Use AI: Many centralized AI platforms often have high barrier to entry for developers. In addition, as AI becomes more powerful, there could be increasing restrictions around who can build or access these applications. Bittensor offers an alternative where permission is not needed to access resources to develop and use AI.
- Open-Source Monetization: While open-source AI models like DeepSeek’s R1 and Meta’s Llama offer benefits, open-source AI still largely struggles with monetization and coordination. Bittensor helps address this issue through TAO token emissions, allowing AI developers monetize and fund their work.
We believe Bittensor’s token (TAO) provides a particularly compelling investment right now because of the following:
- Potential to address concerns related to centralized AI outlined above
- Progress so far onboarding ecosystem investors like Yuma and subnet builders like Masa (data scraping and AI agent arena), Dippy (AI roleplay app), and Kaito (decentralized search)
- The dynamic TAO (“dTAO”) upgrade slated for February[9], which would enable investment in individual subnets; we believe this could inject a wave of new liquidity into the Bittensor ecosystem
The Broader Decentralized AI Landscape
Just recently, some may have thought that open-source AI would always lag behind the best-performing closed-source models offered by tech giants. DeepSeek shows that this is not necessarily the case going forward; key AI innovation does not need to be done in silos and trickle top-down.
Grayscale Research expects that a wide variety of decentralized AI assets may benefit. The DeepSeek development could spur widespread improvements across decentralized AI as efficiency gains are learned and applied. Access to DeepSeek’s highly performant, open-source model reduces costs and lowers the barrier to entry for many open-source decentralized AI projects, particularly at the application layer.[10]
We are already seeing this take place. For example, decentralized AI agent launchpad ai16z is already allowing for agents building with its ELIZA framework to access DeepSeek’s model.[11] On January 27, Venice.ai, a decentralized application that offers access to the DeepSeek model while retaining users’ data privacy on their local devices, launched its token, which reached a valuation of over $1bn within two hours of its launch.[12]
Conclusion
As developments like the emergence of DeepSeek continue to shape the landscape of AI, the race for international tech supremacy, and society, Grayscale Research believes it's crucial that we embrace decentralized solutions that address the risks of centralization. By leveraging platforms such as these, we can potentially safeguard against monopolistic control, nurturing a future for AI that is more secure.
[1] “How a top Chinese AI model overcame US sanctions.” MIT Technology Review. Jan 24, 2025
[2] “How DeepSeek’s AI Stacks Up Against OpenAI’s Model.” The Wall Street Journal. Jan 28, 2025
[3] “China’s DeepSeek AI dethrones ChatGPT on App Store: Here’s what you should know.” CNBC. Jan 27.
[4] “China's DeepSeek AI shakes industry and dents America's swagger.” BBC. Jan 28, 2025.
[5] “China's DeepSeek AI shakes industry and dents America's swagger.” BBC. Jan 28, 2025.
[6] “DeepSeek hit with ‘large-scale’ cyber-attack after AI chatbot tops app stores.” The Guardian. Jan 27, 2025.
[7] “Google apologizes for ‘missing the mark’ after Gemini generated racially diverse Nazis.” The Verge. Feb 21, 2024.
[8] “China’s new DeepSeek AI refuses to answer these questions, experts warn.” The Independent. Jan 28, 2025.
[9] X.com
[10] “How is DeepSeek Better Than ChatGPT: Cost Comparison.” Creole Studios. Jan 28, 2025.
[11] “Does DeepSeek trigger a major reshuffle in the AI Agent sector? Is it time to buy the dip or retreat?” AICoin.com. Jan 27, 2025.
[12] “Venice AI token that gives private access to DeepSeek hits $1.6B total value.” TradingView. Jan 27, 2025.
[13] Holdings are subject to change without notice.
Important Information
Grayscale Operating, LLC (“GSO”) is the parent holding company of Grayscale Advisors, LLC (“GSA”), an SEC-registered investment adviser, as well Grayscale Securities, LLC (“GSS”), an SEC-registered broker/dealer and member of FINRA, and Grayscale Investments Sponsors, LLC ("Grayscale"). Grayscale is not registered as an investment adviser under the Investment Advisers Act of 1940 and none of the investment products (“Products”) sponsored or managed by Grayscale are registered under the Investment Company Act of 1940.
Private placement securities are speculative, illiquid, and entail a high level of risk, including the risk that an investor could lose their entire investment. The Products are not suitable for any investor that cannot afford loss of the entire investment.
Carefully consider investment objectives, risk factors, fees and expenses before investing. This and other information can be found in each Product’s private placement memorandum, which may be obtained from Grayscale and, for each Product that is an SEC reporting company, the SEC’s website, or for each Product that reports under the OTC Markets Alternative Reporting Standards, the OTC Markets website. Reports prepared in accordance with the OTC Markets Alternative Reporting Standards are not prepared in accordance with SEC requirements and may not contain all information that is useful for an informed investment decision. Read these documents carefully before investing.
The shares of each Product are intended to reflect the price of the digital asset(s) held by such Product (based on digital asset(s) per share), less such Product’s expenses and other liabilities. Because each Product does not currently operate a redemption program, there can be no assurance that the value of such Product’s shares will reflect the value of the assets held by such Product, less such Product’s expenses and other liabilities, and the shares of such Product, if traded on any secondary market, may trade at a substantial premium over, or a substantial discount to, the value of the assets held by such Product, less such Product’s expenses and other liabilities, and such Product may be unable to meet its investment objective.
If the shares trade at a premium, investors who purchase shares on the secondary market will pay more for their shares than investors who purchase shares directly from authorized participants. In contrast, if the shares trade on the secondary market at a discount, investors who purchase shares directly from authorized participants will pay more for their shares than investors who purchase shares on the secondary market.
Grayscale intends to attempt to have shares of its Products quoted on a secondary market. However, there is no guarantee that we will be successful. Although the shares of certain products have been approved for trading on a secondary market, investors in the new products should not assume that the shares will ever obtain such an approval due to a variety of factors, including questions regulators, such as the SEC, FINRA, or other regulatory bodies may have regarding such products. As a result, shareholders of such products should be prepared to bear the risk of investment in the shares indefinitely.
The shares of each Product are not registered under the Securities Act of 1933, the Securities Exchange Act of 1934 (except for Products that are SEC reporting companies), the Investment Company Act of 1940, or any state securities laws. The Products are offered in private placements pursuant to the exemption from registration provided by Rule 506(c) under Regulation D of the Securities Act of 1933 and are only available to accredited investors. As a result, the shares of each Product are restricted and subject to significant limitations on resales and transfers. Potential investors in any Product should carefully consider the long-term nature of an investment in that Product prior to making an investment decision. The shares of certain Products are also publicly quoted on OTC Markets and shares that have become unrestricted in accordance with the rules and regulations of the SEC may be bought and sold throughout the day through any brokerage account.
Except as noted below, Grayscale does not store, hold, or maintain custody or control of any Product’s digital assets, but instead has entered into a Custodian Agreement on behalf of each Product with Coinbase Custody Trust Company, LLC (the “Custodian”), a third party to facilitate the security of each Product’s digital assets. The Custodian controls and secures each Product’s digital asset account, a segregated custody account to store private keys, which allow for the transfer of ownership or control of the digital asset, on each Product’s behalf. If the Custodian resigns or is removed by Grayscale or otherwise, without replacement, it could trigger early termination of such Product. Grayscale currently serves as a key maintainer for any Product’s TAO holdings.
The Products are distributed by Grayscale Securities, LLC (Member FINRA/SIPC). SIPC coverage does not apply to the crypto asset products or services mentioned.
