Slower frontier AI model development should not slow AI adoption

Highlights: The Anthropic CEO’s three point plan for a better paced LLM development rate was well received by its two biggest competitors. It helps address some of the public’s concerns, while allowing rapid AI adoption to continue. Any headline-related dip in chip stocks should be mild and temporary. We continue to like chips, the cloud and AI-enabled applications. The plan could have a small positive effect on bond markets but this week’s Fed meeting is more important.

• Mr Amodei’s plan is to 1) slow the development of AI frontier models to allow assessment by independent evaluators, 2) target industry-wide coordination, and 3) global coordination
• The context helps explains the timing of the plan. An Anthropic ex-employee recently created headlines talking about existential Ai-related risks. US public opinion has been increasingly cautious about AI - a potential mid-term election topic - but the government is reluctant to intervene because if the US doesn’t “win in AI, this would be a very bad position”. China’s AI models have been catching up. Markets have been questioning the return on investment. And of course, ahead of the planned major AI IPOs, it is helpful to try to address all these points
• The stated intention of slowing down LLM development between fierce competitors (and not too much, to avoid Chinese models catching up) should not lead to a major decline in investor’s demand assessment for chips and datacentres and key components. The shortage of supply and bottle neck in these areas will not disappear. As a result, any dip related to the announcement should be shallow and temporary. Moreover, the companies may now have some more time to monetise their capex before developing a new generation. Beyond chips and the cloud, the opportunities remain ample in AI-enabled applications (including physical AI), while broad-based productivity gains support global earnings growth across sectors and geographies. Chinese competitors may get a lift, but the risk is that if they advance too quickly, the US could now use increased ‘safety’ of US models to restrict Chinese model use in the US
• If hyperscalers slow the capex race, this could be a small positive for bonds, but Treasury supply remains ample. The Fed meeting on Wednesday seems a more important factor, and we expect a rate hike. Markets want more clarity on Fed policy, but the Fed is not eager to give forward guidance. We maintain a medium duration stance and like active management to react to market volatility. We also find income elsewhere, in infrastructure and dividend stocks