Everyone Hates AI Right Now. Four Stocks That Are Bulletproof

BWB - Business With Brian•Sep 25, 2026

Summary

Brian states that the current panic surrounding AI slowdowns is unfounded, as the massive buildout of AI infrastructure is a fundamental and unstoppable trend, comparable in scale to historical American economic booms. He emphasizes that physical constraints, such as the limited capacity of existing data centers for new, power-intensive chips, and long wait times for new grid connections (up to 4 years), ensure that demand for older chips and basic infrastructure components remains high. Brian believes that money will continue to flow into four key areas: processors, memory, optics (for data transfer), and inference (companies renting out finished AI compute power). He is not trying to guess which AI company will win, but rather aims to own the essential "parts that none of them can win without." Brian uses a personal investment system that analyzes fair value, price trends, and historical data to generate "buy a lot," "buy," "buy a little," "hold," or "sell" signals, often adjusting signals based on historical valuation issues or limited public data.

Brian applies his investment system to several companies:

**Micron:** Brian's system identifies Micron as one of only three manufacturers of a crucial type of memory, leading to a significant shortage and a surge in operating profit from 11 cents to 80 cents per dollar sold. While the price trend initially signals "buy a lot," the system flags that Micron is historically expensive, having reached this valuation only about 2% of the time in its 10-year history. Consequently, Brian's system adjusts the signal to "buy a little," and he intends to await the September earnings report to potentially buy more.

**SanDisk:** As a flash storage provider spun out from Western Digital, SanDisk has seen its operating profit jump to 78 cents per dollar sold. Brian notes its attractive valuation, trading under eight times next year's earnings with a PEG ratio of 0.3, indicating cheap growth. Both price and trend suggest "buy a lot." However, due to SanDisk's short public history (19 months), Brian's system lacks 10 years of data for historical comparison, which leads it to downgrade the signal to "buy."

**Lumentum:** Brian examines Lumentum, which produces lasers vital for inter-chip communication in data centers. The company has shifted from losing money to retaining 27 cents of operating profit per dollar. Despite a reported earnings "miss," the stock surged, indicating strong underlying business. However, Brian's system signals caution, as the stock is currently valued at 195 times last year's profit, significantly higher than its usual 70 times, suggesting the price has outrun the business's fundamentals, even with a PEG ratio under one. Brian implies a "sell some" action for this reason.

**Marvell:** Brian reviews Marvell, a designer of custom AI chips for major cloud companies like Google. Google has committed to buying $120 billion worth of chips through 2033, securing a 6.7% stake in Marvell. While this represents 7 years of locked-in business, Brian's system indicates that the stock price has already factored in this future growth. Marvell is trading at 80 times last year's profit, far exceeding its historical average of 33 times. Brian's system generates one of the "strongest sell signals" for Marvell in the video, asserting that its price has moved too far ahead of its business value.

**Cerebras:** Brian discusses Cerebras, a company that builds massive AI chips and rents out its compute power, with 70% of its revenue from long-term contracts. Despite a $450 million loss in its latest quarter (primarily non-cash stock compensation), Cerebras holds $8.6 billion in cash and has an impressive $25 billion backlog against less than $1 billion in current sales. Although the stock is sliding, Brian's system's price analysis suggests "buy a lot." However, its very limited public history (4 months) prevents a full historical valuation check, leading Brian to rely on Wall Street analyst estimates for fair value and notch down the signal. Brian states he will monitor the backlog conversion for future buying decisions.

**CoreWeave:** Brian presents CoreWeave, which rents Nvidia chips by the hour, catering to labs and cloud companies. Its quarterly sales have dramatically increased from $400 million to $2.6 billion, with customers signing for three times its current power capacity. Brian highlights a significant risk: CoreWeave burned $13.7 billion in cash over the past 12 months to build ahead of demand, against $7.6 billion in sales. Brian updated his information to reflect this substantial cash burn and aggressive expansion.

**Nvidia:** Brian positions Nvidia as the anchor of the AI revolution, noting its quarterly sales have soared from $30 billion to $96 billion in two years, with high operating profits. Despite a "cringey" detail about Nvidia guaranteeing up to $18 billion in customer data center leases, Brian's system is highly bullish. Nvidia is priced at 28 times last year's profit, significantly below its 5-year average of 52 times, and boasts the cheapest growth on Brian's list with a PEG ratio of 0.35. Brian concludes that Nvidia's valuation doesn't reflect the immense demand and growth, and its history check comes back with no problems. Brian recommends dollar-cost averaging for this company.

Mentioned Stocks

Sentiment: BUYAction: BOUGHT

Reasoning: Brian explains that Micron is one of only three companies globally manufacturing a crucial type of memory, leading to a significant shortage. Its operating profit surged from 11 cents to 80 cents per dollar sold. Brian's system, while seeing a 'buy a lot' signal based on current price trend, notes that the stock is historically expensive, having been at this valuation only 2% of the time in its 10-year history. This historical check leads Brian's system to adjust the signal down to 'buy a little,' and Brian states he will buy a little, awaiting the September earnings report for further decisions.

Loading chart...

NVDA

Sentiment: BUYAction: RECOMMENDED

Reasoning: Brian positions Nvidia as the anchor of the AI industry, with quarterly sales soaring from $30 billion to $96 billion in two years and maintaining high operating profits. Despite a potential risk of co-signing up to $18 billion in customer data center leases, Brian's system generates a very strong 'buy a lot' signal. He highlights its attractive valuation at 28 times last year's profit, significantly below its 5-year average of 52 times, and notes it has the cheapest growth on his list with a PEG ratio of 0.35. Brian concludes that Nvidia's valuation doesn't reflect the immense demand and growth, and its history check comes back with no problems. Brian recommends dollar-cost averaging for this company.

Loading chart...

MRVL

Sentiment: SELLAction: SOLD

Reasoning: Brian discusses Marvell, which designs custom AI chips for major cloud companies like Google. He notes that Google has committed to buying approximately $120 billion worth of chips through 2033, securing a 6.7% stake in Marvell. However, Brian's system indicates that the stock's price has already incorporated this significant future growth. It is trading at 80 times last year's profit, which is substantially above its historical average of 33 times. For these reasons, Brian's system generates one of the 'strongest sell signals' in the video, concluding that the price has significantly outpaced the underlying business value.

Loading chart...

LITE

Sentiment: SELLAction: SOLD

Reasoning: Brian analyzes Lumentum, which produces lasers essential for communication within data centers, noting its shift from losing money to retaining 27 cents of operating profit per dollar sold. Despite strong underlying business that saw the stock rise 13% after an earnings 'miss,' Brian's system signals a potential sell. He points out that the stock is currently trading at approximately 195 times last year's profit, significantly higher than its historical average of around 70 times. Even with a PEG ratio under one, Brian believes the price has outrun what the business fundamentally deserves, leading to a 'possibly sell some' signal.

Loading chart...