1 October we become SureStone CapitalSame team, same ABN. Our website and email addresses move to surestone.com.au.
Preparing video
Key points:
Chinese open source AI models seen rapidly closing the gap with US frontier modelsCheaper AI tokens viewed as a long-term positive for infrastructure and usageFocus on AI “picks and shovels” such as $NVDA, $TSM, $AVGO, CATL and power infrastructure
Alex Pikoulas from Munjarra Capital outlines a bullish long-term stance on the artificial intelligence ecosystem despite recent semiconductor weakness and concerns over AI model pricing. Pikoulas points to China’s Moonshot and its new Kimi K3 model as evidence that Chinese large language models are closer in capability to US frontier models than widely assumed, while being significantly cheaper and open source. In his view, falling AI token costs may not damage the ecosystem due to a Jevons paradox-style effect, where cheaper access drives far greater usage and infrastructure demand.
Pikoulas sees potential valuation risks for unlisted model developers such as OpenAI and Anthropic, but remains positive on the broader value chain. He expects hyperscaler capital expenditure to stay elevated, with infrastructure remaining essential regardless of whether open source or proprietary models dominate. Upcoming US tech earnings, particularly Alphabet’s $GOOGL Google Cloud numbers and AI monetisation metrics, are seen as critical, with any moderation in CapEx guidance likely welcomed over time, even if share price reactions remain volatile on the day.
Positioning remains focused on “picks and shovels” beneficiaries of AI demand, including $NVDA, $TSM, $AVGO, battery giant CATL, power-equipment producer Evonik, copper exposure, and data centre plays, alongside broader power and infrastructure suppliers.
