Ideas
Chinese open-source AI will likely win.
China's AI approach is open-source/open-weight and largely free, treating AI as a utility. This makes Chinese models cheaper, more flexible, and more adoptable outside the US; he compares it to Android/Linux and expects it to win globally over time, giving China a longer and cheaper runway than the US closed-source model.
Alibaba's Qwen and ecosystem are winners.
Alibaba owns Qwen, the most powerful Chinese LLM, and can fund it through fees/data across Taobao, AliLogistics, Alipay and the broader Alibaba ecosystem. It has a large commercial network and is a likely survivor; he prefers it among Chinese AI plays.
Nvidia's moat and margins eroding.
Nvidia's hold and margins are under attack from custom ASICs such as Amazon Trainium and Google TPU, Chinese de-Nvidiaization, Moore's-law limits, advanced packaging/cognitive towers, CUDA circumvention, and DeepSeek training/inference breakthroughs. Its high-margin monolithic-chip model faces lower-margin alternatives, threatening profits and valuations.
Chinese semiconductors are moving upmarket.
China dominates low-margin commodity chips and is moving upmarket to 14-18nm medium-value chips used in vehicles, phones, towers, smart factories, drones and power systems. By 2028 it is forecast to produce more chips than it needs, reducing dependence on foreign high-end chips.
Chinese robotics deployment is structurally bullish.
Smart factories are spreading across China at extraordinary rates; China installed more robots than the rest of the world combined last year. The deployment is enabled by good-enough mid-tier chips and cheap energy, making Chinese robotics/automation a structural beneficiary of China's AI-industrial path.
Google is best positioned in AI.
Google has the strongest AI position: Gemini is powerful, distribution is huge via Android/Chrome/Samsung and potentially Apple, and TPUs are purpose-built, avoiding the worst Moore's-law/cost problems. He likes it most and says it is assembling the pieces better than anyone else.
Oracle and CoreWeave are bubble risks.
Debt is creeping into the AI system on and off balance sheets, especially at Oracle and CoreWeave. Oracle's huge lease obligations make it the focus if you are on bubble watch and heading toward a crisis point; he would prefer Meta over CoreWeave/Oracle if forced to choose.
Microsoft's OpenAI tie is deadweight.
Microsoft is attached to OpenAI, which he sees as a deadweight because OpenAI's capital costs exceed likely revenues. Microsoft may diversify, but its OpenAI exposure is a short-to-medium-term drag.
Meta's AI/LLM position is lost.
Meta has lost its way on AI: Llama has stalled, the best scientist left, no replacement model is apparent, and Zuckerberg is casting around. In 3-5 years it may be just a souped-up search/retail engine, and its LLM offering is nowhere.
Meta drawdowns are buying opportunities.
Even if Meta's AI spending leads to large losses and the stock draws down 80%, its core advertising business is hugely profitable and can eventually stop funding money-losing AI projects. Similar to 2022, such a drawdown could be a buying opportunity.
Tencent is a Chinese AI survivor.
Tencent is a Chinese AI survivor because it has large non-AI businesses and a broad commercial ecosystem, so it is not a standalone LLM dependent on monetizing model access. He prefers names like Tencent and Alibaba where there is something else involved.
Chinese AI IPO bubble is forming.
He is concerned about new Chinese AI issues and second-tier listings, especially in Hong Kong/Shanghai/Shenzhen, after huge first-day pops like MetaX's 700%. Software-only AI companies face the same commoditization pressures as OpenAI, while chip-related names are less worrying.
Baidu is China's Google-like AI play.
Baidu is the main Chinese search engine and is developing its own AI capabilities, doing something similar to what Google is doing. He is probably okay with it at the moment and sees it as a potential survivor, though less emphatically than Tencent/Alibaba.
China EUV threatens ASML monopoly.
ASML has only a two-to-three-year calm period before China's domestically developed EUV equipment, tied to Shenzhen's Manhattan project and company SME/SMEE, becomes commercially available around 2028. That would end ASML's monopoly and pressure it.
US AI capex race ends badly.
US AI hyperscalers/data-center companies are trapped in a prisoner's dilemma/Red Queen race: none can stop capex because rivals would surge ahead. Collective overinvestment guarantees a systemic bust; weaker players fail and survivors are negatively impacted, though some survive like Amazon after 2000.
This Monetary Matters video, published January 12, 2026,
features Michael Power, Jack Farley
discussing AI-SECTOR, BABA, NVDA, Chinese semiconductor industry, Chinese robotics/automation, GOOG, ORCL, CoreWeave, MSFT, META, TCEHY, Chinese AI software stocks, BAIDU, ASML, US AI hyperscalers.
15 trade ideas extracted by AI with direction and confidence scoring.
Speakers:
Michael Power,
Jack Farley
· Tickers:
AI-SECTOR,
BABA,
NVDA,
Chinese semiconductor industry,
Chinese robotics/automation,
GOOG,
ORCL,
CoreWeave,
MSFT,
META,
TCEHY,
Chinese AI software stocks,
BAIDU,
ASML,
US AI hyperscalers
