Subscribe
Share
Share
Embed
Podcasts from Confluence Investment Management LLC, featuring the periodic Confluence of Ideas series, two bi-weekly series: the Asset Allocation Bi-Weekly and the Bi-Weekly Geopolitical Report (new episodes posted on alternating Mondays), and a new monthly Q&A format called the Confluence Mailbag.
Welcome to the Confluence Investment Management Asset Allocation Biweekly Report for 08/17/2026. I'm Phil Adler. Investors worried about a possible AI bubble can point to a number of reasons, including rising valuations and debt. Confluence chief market strategist Patrick Fearon Hernandez joins us today to discuss another concern, China's threat to the AI investment boom. Patrick, let's look at where the market is right now.
Phil Adler:To what degree does AI dominate The US stock market's performance these days?
Patrick Fearon-Hernandez:Well, Phil, AI has become one of the most important drivers of both economic growth and stock market performance. The largest technology companies are investing hundreds of billions of dollars in AI infrastructure, including data centers, semiconductors, networking equipment, software, and power generation. AI related investment has become a major contributor to US GDP growth and is now one of the strongest sources of private fixed investment in the economy. From an equity market perspective, AI enthusiasm has become highly concentrated in a relatively small group of large cap tech companies. Goldman Sachs estimates that the leading tech firms accounted for more than half of the S and P five hundred's return in 2025 largely due to investor expectations regarding AI driven earnings growth and productivity gains.
Patrick Fearon-Hernandez:Importantly, AI's influence extends beyond the hyperscalers. The boom has lifted firms involved in semiconductors, servers, electrical equipment, construction materials, power infrastructure, cooling systems, and industrial machinery. In many ways, AI has become an economy wide investment theme rather than simply a tech story.
Phil Adler:So any serious loss of confidence could bring down many sectors, not just the broad technology bundle.
Patrick Fearon-Hernandez:That's exactly right. What makes the AI cycle unusual is how deeply it's penetrated the broader economy. The AI build out has supported growth in manufacturing and construction, utilities, cloud computing, and transport. Data center construction alone has become one of the fastest growing categories of commercial investment. As a result, if investors begin to doubt that AI related spending will generate adequate returns, the consequences could spread well beyond hyperscalers and cloud providers.
Patrick Fearon-Hernandez:Firms supplying chips, electrical equipment, building materials, and even energy infrastructure could face slower growth expectations. That doesn't mean a collapse is inevitable. However, when a market theme becomes as dominant as AI has become, the risks become systemic. A major disappointment would likely affect a wide range of sectors and could potentially weigh on overall economic growth.
Phil Adler:Patrick, your thesis this week is that China poses a very serious threat to The US on the AI front. First, how do the two nations compare technologically in the race to develop dominant AI?
Patrick Fearon-Hernandez:Well, the conventional wisdom has been that The US maintains a sizable technological lead in AI. That leadership remains significant in areas such as semiconductor design, advanced computing hardware, and frontier model development. However, the gap appears to be narrowing quickly. Recent Chinese models from firms such as DeepSeq and Moonshot AI have demonstrated performance levels that are approaching those of leading US systems. According to benchmark testing cited in our report, DeepSeq's v four flash and Moonshot's Kimi k three are performing near the level of Anthropic's OPUS 4.8 and OpenAI's latest offerings in several important tasks, including coding and reasoning applications.
Patrick Fearon-Hernandez:At the same time, China benefits from a massive domestic market, substantial government support, abundant engineering talent, and a strategic commitment to achieving technological leadership. Stanford's twenty twenty six AI index still shows The US leading in private sector AI investment, but it also notes that official investment figures likely understate China's total commitment because of big government funding programs. The key point is that China no longer appears to be merely catching up. In certain segments of the market, it may already be competing on nearly an equal footing.
Phil Adler:Patrick, are Chinese open source models simply more appealing to many customers?
Patrick Fearon-Hernandez:For many organizations, the answer may be yes. Open source or open weight models offer important They provide greater transparency, they allow firms to run models on their own infrastructure, they permit customization, they reduce dependence on a single vendor, and often lower costs substantially. These characteristics are particularly attractive to enterprises and governments concerned about data security and technological sovereignty. Research from MIT notes that open models often achieve performance levels close to proprietary models while costing dramatically less to use. In many cases, organizations don't need the absolute best model available.
Patrick Fearon-Hernandez:They need a model that is very good, affordable, and adaptable. That creates a challenge for US companies whose business models depend on charging premium prices for proprietary services. If open source alternatives continue improving, customers may increasingly question whether premium pricing is justified.
Phil Adler:And in what appears to be a major development, China is offering its AI models to the world for much cheaper rates, it appears, compared to American companies. How much cheaper exactly?
Patrick Fearon-Hernandez:Well, the pricing differences are extraordinary. As highlighted in our report, DeepSeq's v four flash charges about 28¢ for a quantity of output that costs approximately $25 using Anthropix OPUS 4.8. That represents a price difference of nearly 99%. Even after OpenAI cut the price of its GPT 5.6 Luna model by 80%, its cost remained roughly a dollar 20 for the same amount of output, still more than four times DeepSeq's price. These kinds of price differentials are difficult to ignore.
Patrick Fearon-Hernandez:Historically, when comparable products become available at a fraction of the cost, pricing pressure tends to spread throughout the industry.
Phil Adler:Patrick, how can Chinese firms offer their AI so cheaply?
Patrick Fearon-Hernandez:Well, several factors are likely at work. First, Chinese firms may be willing to accept lower profit margins in exchange for market share. Second, they benefit from national policies that often support strategic industries. Third, open source development models can reduce costs and accelerate innovation through broader collaboration. More importantly, China also has a long history of pursuing industrial strategies designed to establish dominance in key sectors by prioritizing market share over profitability.
Patrick Fearon-Hernandez:We've seen similar dynamics in industries such as steel, solar panels, rare earth processing, consumer electronics, and electric vehicles. Whether you call it industrial policy, strategic competition, or predatory pricing, the effect is the same. Competitors are forced into aggressive price reductions that compress industry profits.
Phil Adler:Patrick, is there any way US firms can counter this strategy by China?
Patrick Fearon-Hernandez:Well, we see several possible responses. One approach, for example, is technological differentiation. If US firms can continue producing models with superior performance, reliability, security, or specialized capabilities, they may be able to justify premium pricing. Anthropic appears to be pursuing exactly that strategy. Another possibility is leveraging ecosystem advantages.
Patrick Fearon-Hernandez:Companies like Microsoft, Amazon, and Google offer integrated platforms that combine cloud computing, software, enterprise relationships, and AI services. Those ecosystems create switching costs and may help preserve profitability. A third response is policy. Governments could employ export controls, procurement preferences, research funding, or other tools to support domestic champions. The challenge is that price competition on the scale we're seeing from the Chinese firms is extremely difficult to counter indefinitely if product quality remains comparable.
Phil Adler:Is there any good news in all this that US firms can hang on to?
Patrick Fearon-Hernandez:Sure. The US still possesses substantial advantages. US firms remain global leaders in advanced semiconductors, cloud infrastructure, software ecosystems, and venture capital financing. The US also retains the world's deepest capital markets in some of the strongest innovation networks. Moreover, lower AI prices are not necessarily bad for the global economy.
Patrick Fearon-Hernandez:As AI becomes cheaper, businesses can adopt it more broadly, potentially accelerating productivity growth and creating new applications. History shows that consumers often benefit enormously when powerful technologies become more affordable. So while pricing pressure may threaten AI providers, it could simultaneously increase overall AI adoption and support broader economic growth.
Phil Adler:Patrick, what appears to be the most likely outcome of of this high stakes competition?
Patrick Fearon-Hernandez:The most likely outcome is probably not total victory by either side. Instead, we may see an increasingly commoditized AI model market where pricing continues to fall and profits migrate elsewhere in the value chain. Much as we saw with personal computers and smartphones, value may ultimately accrue less to the underlying technology itself and more to the surrounding ecosystem, applications, services, and customer relationships. Under that scenario, Chinese firms could capture substantial market share through low cost models while US companies maintain leadership and higher value applications, enterprise solutions, infrastructure, and specialized services. The biggest risk is that investors have priced many AI related companies as though extraordinary profits will continue indefinitely.
Patrick Fearon-Hernandez:If margins compress significantly due to competition, equity valuations may have to adjust accordingly.
Phil Adler:And finally, Patrick, how are Confluence model portfolios positioned for a variety of outcomes, including this one?
Patrick Fearon-Hernandez:Well, at Confluence, we continue to emphasize diversification diversification and risk management rather than making concentrated bets on any single technological outcome. While we recognize AI's transformative potential, we also recognize that periods of technological enthusiasm often bring elevated valuations and increased uncertainty. Therefore, our approach focuses on maintaining exposure to long term growth opportunities while avoiding excessive dependence on a single theme. For example, in our latest quarterly adjustments to the Confluence asset allocation strategies, we deliberately broadened our exposure to a wide range of asset classes so that if and when The US's AI frenzy starts to go into reverse, we should benefit no matter where investors shift their capital. For instance, we brought back our exposure to US mid cap and small cap equities.
Patrick Fearon-Hernandez:We think those moves have helped prepare us for the eventual end to the AI boom whenever that may come.
Phil Adler:Thank you, Patrick. The title of this week's report is China's threat to the AI investment boom, and you can find a link to the written report, dated August 17, on the Confluence webpage, confluenceinvestment.com. Our discussion today is based upon sources and data believed to be accurate and reliable. Opinions and forward looking statements expressed are subject to change without notice. This information does not constitute a solicitation or an offer to buy or sell any security.
Phil Adler:Our audio engineer is Dane Stole. I'm Phil Adler.