Is Now the Best Time to Invest in AI Stocks?

TL;DR

The speaker sees major opportunities in AI stocks, especially AMD and Intel, but Berkshire Hathaway’s nearly $400 billion cash position signals that investors should remain cautious. Agentic AI requires far more CPU capacity than the speaker previously realized, while announced AI infrastructure spending exceeds $700 billion this year, up 77% from last year. Read on for the CPU-to-GPU math, AMD’s growth, and the risks behind its major deals.

Transcript

One month ago, the market was panicking over the Iran war and broken supply chains. Today, it's verging on extreme greed. But below all the headlines, big institutions are quietly building cash because they see a massive opportunity ahead. My name is Alex and I spent 8 years as an electrical engineer and AI researcher at MIT, which helped me find s... Read More

Key Insights

• Warren Buffett's Berkshire Hathaway is holding a record $400 billion in cash, indicating a cautious approach amidst market greed.
• AI infrastructure spending has surged, with major tech companies investing over $700 billion this year, a 77% increase from last year.
• AMD's data center CPU sales are growing faster than its GPU sales, highlighting the increasing importance of CPUs in AI data centers.
• Intel has secured major customers like Tesla and Google, marking a significant turnaround for its foundry business.
• ARM's new AGI CPU offers 54% better performance than Nvidia's, potentially disrupting the AI CPU market.
• Meta Platforms has committed to using AMD's Instinct GPUs, but ARM's AGI CPU could have offered significant cost savings.
• The AI CPU market is becoming more competitive with ARM entering the fray, challenging established players like Nvidia and AMD.
• A potential breakthrough by Sub Quadratic in AI efficiency could drastically reduce costs, impacting demand and infrastructure needs.

Questions & Answers

Q: Is now the best time to invest in AI stocks?

The speaker believes major opportunities are emerging in AI stocks, particularly AMD and Intel, but does not present the market as risk-free. The fear and greed index has been near extreme greed for three weeks, while Berkshire Hathaway is holding almost $400 billion in cash.

Q: Why is Berkshire Hathaway holding so much cash?

The speaker points to war, broken supply chains, rising costs, and a market approaching extreme greed. Berkshire Hathaway holds almost $400 billion in cash, equal to 32% of its portfolio and described as an all-time record for the firm.

Q: How much is being spent on AI infrastructure this year?

The biggest technology companies announced more than $700 billion in AI infrastructure spending for this year. According to the transcript, that is 77% more than last year despite stalled supply chains.

Q: Why are CPUs important in agentic AI data centers?

CPUs handle orchestration work such as tool calls, sub-agent spawning, and context management, while GPUs perform the heavy mathematical processing. A coding agent can make many separate tool calls, spawn hundreds of sub-agents, and increase its memory usage tenfold during a session.

Q: What CPU-to-GPU ratio do AI data centers require?

Traditional AI data centers use roughly one CPU for every eight GPUs. For the agentic AI example, the speaker calculates about 4,600 Vera CPUs for every 12,000 Rubin GPUs, or one CPU for every 2.6 GPUs, while the rack ratio is nine GPU racks for each CPU rack.

Q: What mistake did the speaker make when analyzing AI data centers?

The speaker focused on rack counts instead of chip counts. Although agentic AI requires nine times more GPU racks than CPU racks, counting chips shows only about three times more GPUs than CPUs, meaning there were roughly three times more CPUs than the speaker had realized.

Q: How strong were AMD’s latest reported results?

AMD reported quarterly revenue of $10.3 billion, up 38% year over year. Data center revenue reached $5.8 billion and grew 57%, with most of that revenue coming from data center CPUs rather than GPUs.

Q: What risk comes with AMD’s deals with Meta and OpenAI?

AMD gave Meta the right to buy roughly 10% of the company for one penny per share if AMD stock reaches $600. The speaker says AMD has the same arrangement with OpenAI, creating potential dilution of 20% for AMD shareholders.

Summary & Key Takeaways

• Warren Buffett's Berkshire Hathaway is amassing cash amid market greed, suggesting a cautious investment approach. The video discusses the growing importance of CPUs in AI data centers, highlighting AMD's and Intel's recent successes and strategic moves. ARM's new AGI CPU could disrupt the market with superior performance. (50 words)

• The speaker acknowledges a past oversight in evaluating AI data center needs, emphasizing the need to consider chip counts over racks. Recent developments by AMD, Intel, and ARM are reshaping the AI infrastructure landscape, with significant implications for investors. The video also mentions a potential breakthrough in AI efficiency by Sub Quadratic. (50 words)

• The video explores the evolving AI infrastructure market, highlighting Warren Buffett's strategic cash reserves and the increasing role of CPUs. AMD, Intel, and ARM are making significant strides, with ARM's AGI CPU offering potential cost savings. The speaker stresses the importance of understanding market dynamics and acknowledges past analytical mistakes. (50 words)