Artificial intelligence news
Chinese startup DeepSeek unveiled its latest model, DeepSeek-V3, which rivals leading U.S. models at a fraction of the cost (reportedly only $5mm), underscoring China's progress despite export restrictions. OpenAI announced its $500 billion Stargate project, aiming to build advanced AI infrastructure and reduce dependence on Microsoft. Meanwhile, Meta committed up to $65 billion to expand its AI capabilities, including massive data center investments and advancements in its Llama models. Microsoft has announced plans to invest approximately $80 billion in fiscal year 2025 to develop AI-enabled data centers. This investment aims to enhance the company's infrastructure for training AI models and deploying AI and cloud-based applications. Notably, more than half of this investment is allocated to projects within the United States, reflecting Microsoft's commitment to domestic technological advancement.
- The most significant of the above is the release of the Chinese model, which rivaled Open AI's latest model, but was reportedly created for only $5mm
- The model is also open source and small enough that it could be run locally (i.e., on your device rather than in the cloud)
- The United States has long approached AI development as an arms race and has tried to limit China’s ability to use our technology (Nvidia chips) to catch up
- If the cost and performance of China’s DeepSeek model are indeed accurate (there is speculation that they had much more access than they are telling the world), this could mean that we could have been somewhat misguided in focusing on massive hardware spending to advance our AI initiatives
- Regardless, as per META, Microsoft, and the new Stargate project, hardware spending in the US will continue for the time being
- We expect more fallout as the new DeepSeek model is utilized in the days and weeks ahead
- At the very least, it could mean that generative AI can be implemented more broadly using smaller models for specific use cases, something we have written about before
Crypto developments
This week, President Trump signed an executive order establishing a Presidential Working Group on Digital Asset Markets, led by David Sacks, aiming to promote the growth and integration of cryptocurrencies into the U.S. economy. The executive order emphasizes protecting access to blockchain networks and fostering the development of stablecoins backed by the U.S. dollar. It also prohibits the creation and use of central bank digital currencies (CBDCs) in the United States, reflecting a commitment to decentralized digital assets.
- The change of tone towards the digital asset space by the new administration can have wide-ranging ramifications
- There are certain obvious use cases which could be implemented with a proper regulatory framework:
- Stablecoins as payment rails, particularly in cross-border use cases, can be faster and cheaper than existing options
- Layer 1 and Layer 2 blockchains can be used to facilitate capital formation while also being cheaper and faster than existing capital markets solutions
- Legal titles, such as for your home or automobile, can also be implemented much faster and cheaper via a blockchain solution if properly regulated
- While there are many other use cases, applying a common-sense regulatory approach can significantly increase efficiency for many parts of the financial system and real economy
- Separately, Senator Cynthia Lummis has been appointed to lead the Senate Banking Subcommittee on Digital Assets
- In 2024, Senator Lummis introduced a bill to create a Strategic Bitcoin Reserve, which would entail the US Government acquiring up to 5% of outstanding Bitcoin supply
- While still very debatable, this kind of policy would be positive for Bitcoin, which many consider a “digital gold”