AI has a PR problem. First it was power, next it was jobs and now it is regulation. This is somewhat self-induced after so much discussion around the need for AI regulation; however, this has come to a head and Mythos class models are essentially off the market. On Friday evening, Anthropic pulled all access to the Mythos class models due to the U.S. Government imposing model export restrictions on national security grounds. The complication noted by many experts is that verifying the individual accessing a model is near impossible, requiring Anthropic to restrict access for all users.
This leads to important questions investors should ask: How will we ever reach AGI in models if we just put a ceiling on Mythos access? Can this type of regulatory issue come for OpenAI or any other model builder and what are the implications for the international scale of technology development teams and customers building workflows that do not see clarity around regulation?
The second derivative in AI growth has been a consistent topic in investor conversations, and the multitude of recent challenging AI macro data points could bring this concept to the forefront. The Mythos regulatory hurdle comes along with recent headlines over AI token spending at enterprises and a potential AI token price war. Add to this the significant equity supply from capital raises for SPCX (completed on Friday), GOOGL as well as planned IPOs for OpenAI and Anthropic.
A pullback could be a chance to sharpen analysis on top themes as the more recent frenetic bottleneck chase may take a breather.
A quick update on themes:
Liquid cooling and thermal management:
ECL: We continue to believe the potential Iran deal will remove the overhang of raw material pressures on the shares. Combined with what we view as an imminent closing of the CoolIT acquisition, the stock is in a position to be a relative and absolute outperformer.
ETN: Announced the details around their Mobility division spin — an RMT with Dana (DAN). ETN continues to concentrate its business around electrification, thermal management, and aerospace.
Space ecosystem:
SPCX: The IPO was a success by all accounts for its historic size and closed up 19% on the first day of trading. Near-term catalysts will be street estimate releases from analysts, Nasdaq early inclusion (early July), as well as updates around the next Starship 3 test. We expect dips related to the above-mentioned AI macro concerns to be bought by investors. We do not see the Anthropic deal at risk due to Friday’s announcement given the significant amount of compute needed for the current base of business and the same goes for the Google deal.
AI price discovery:
AI Price Discovery: We put out a note earlier this week (note) framing our view that price discovery of AI tokens and compute is important as the AI trade evolves. It will take some time to align on the significance of moves in GPU rental prices or AI token prices and we will continue to provide our views on both.
The Silicon Data LLM Token Expenditure Index (SDLLMTK) has gained meaningful visibility over the past week and has continued to trend lower, signaling a continued shift toward lower-priced AI token models.
Source: Bloomberg
Please reach out with any topics you would like to discuss further.
