Key Concepts
- LPUs (Local Processing Units): AI chips designed for inference, focusing on low-cost and efficient model execution. Differentiated from GPUs which excel in training.
- Inference: The process of using a trained AI model to make predictions or decisions.
- Gigawatt: A unit of power (1 billion watts), used to quantify the scale of AI compute capacity.
- MI300, MI350, MI450: AMD’s series of AI accelerator chips, with MI450 utilizing 2nm technology.
- TAM (Total Addressable Market): The total market demand for a product or service.
- K-shaped Economy: A situation where economic recovery is uneven, benefiting some segments while leaving others behind.
- Antitrust Issues: Concerns regarding monopolies or unfair competition, influencing deal structures (like licensing agreements).
Nvidia’s Acquisition of Grock: A Deep Dive
The broadcast opens with discussion of Nvidia’s $20 billion acquisition of Grock, a nine-year-old AI chip startup. This represents Nvidia’s largest acquisition to date, a significant jump from Grock’s $6.9 billion valuation just three months prior during a $750 million capital raise. Jonathan Ross, Grock’s founder and CEO, will join Nvidia. Ross previously designed custom chips for Google’s AI model training.
A key insight from an interview with Ross (excerpted in the broadcast) highlights the distinction between GPUs and LPUs. Ross stated, “GPUs are really good for training models… But inference is when you actually run that model… We accelerate inference and we make it really low cost.” This underscores Grock’s specialization in inference, the practical application of AI models, as opposed to the computationally intensive training phase. The hefty price tag is attributed to the value of this specialized inference technology.
The deal is structured as a licensing agreement, potentially to navigate antitrust concerns. The acquisition is viewed as Nvidia proactively addressing potential competition in the specialized chip space, acknowledging that future revenue may increasingly come from inference-focused hardware.
The Evolving AI Landscape & Investment Strategies
Keith Buchanan, Global Investments senior portfolio manager, frames the acquisition not as a bid for a competitor, but as an investment in capacity expansion beyond Nvidia’s current GPU platform. He believes this signals continued growth in the AI infrastructure buildout, countering concerns about an overinflated “AI bubble.” Buchanan stated, “This is a positive affirmation of the capacity growth… we haven’t seen the end of this investment phase.”
The discussion highlights a shift in investment focus. The broadcast suggests a move beyond the “picks and shovels” phase of AI investment (companies providing basic infrastructure) towards more specialized areas. Buchanan anticipates the AI story will evolve, requiring justification of current valuations through innovation in previously unappreciated spaces. He notes increasing activity in the private credit market related to AI funding, raising questions about sustainability.
Buchanan advocates for expanding investment beyond the “Magnificent Seven” (large-cap tech stocks) to include companies supporting the AI buildout, such as those in industrials and utilities. He emphasizes the potential for cheaper entry points and reduced risk in these ancillary plays.
Precious Metals & Energy Market Dynamics
The broadcast pivots to discuss the surge in precious metal prices, particularly silver (up 40% in the past month). Strategists point to a market deficit and industrial demand (silver as the best conductor of electricity, copper as the second best) as driving factors. However, caution is advised, referencing historical crashes following similar price spikes in 1979-1980.
Regarding gold, the narrative centers on central bank acquisitions and its role as a hedge against global macroeconomic risks. Buchanan noted, “The charts tell a simple story… straight up and to the right for gold and silver.” He favors gold over silver due to its perceived lower risk profile.
The energy market is identified as a key area to watch, with a disconnect between increasing AI/data center demand and lower oil/gasoline prices. The potential impact on the affordability crisis and the K-shaped economy is highlighted.
AMD & OpenAI: A Strategic Partnership
The segment on AMD and OpenAI centers around their $6 billion deal for AMD to provide AI compute capacity. Lisa Su, AMD’s chair and CEO, describes the agreement as a multi-year partnership to build six gigawatts of compute power. This will translate to “tens of billions of dollars of revenue” for AMD.
Su emphasizes the importance of collaboration on technology, particularly AMD’s MI450 chip, which OpenAI considers “best-in-class” AI computing. The deal evolved over several years, beginning with OpenAI’s testing of AMD’s MI300 series.
Su explained the impact of six gigawatts of compute: “Millions of GPUs will go towards every user having… intelligence at their fingertips.” She acknowledges the immense difficulty of building this infrastructure, requiring collaboration across the entire ecosystem, including power providers and cloud service partners.
Addressing concerns about potential risks associated with industry consolidation, Su argues that bold moves and industry collaboration will unlock previously unimaginable capabilities. She stated, “I think Brian you know I used to say that I thought the AI TAM… was going to be $500 billion and some people thought boy that was a big number and and now maybe I've underestimated that number.”
Investing Resolutions & Closing Thoughts
The broadcast concludes with a preview of the host’s ten investing resolutions for the coming year, including a focus on scrutinizing management claims on earnings calls and monitoring the interplay between Bitcoin/crypto and precious metals. The host notes the “smack talking” between investors in these asset classes.
The final discussion highlights the surprising resilience of corporate earnings despite tariffs and inflationary pressures. The broadcast ends with a sense of optimism about the long-term potential of AI, framing it as the beginning of a ten-year “super cycle.”
AI summaries can miss context or contain errors. Check important details against the original video.
