This ALWAYS Happens in a Chip Crash - Buy THIS Instead

BWB - Business With Brian•Jul 19, 2026

Summary

Brian analyzes the current volatility in the semiconductor sector, noting that the SOXX index recently fell into a bear market. Brian explains that such sell-offs often drag down strong companies alongside overvalued ones, creating a 'hideout' for real investment opportunities. Brian's thesis focuses on firms that have already booked future revenue through massive backlogs, ensuring growth even if the broader market remains shaky.

Brian highlights several key themes in the AI boom, including data center construction, liquid cooling, optical networking, and custom chip design. Brian emphasizes the importance of the PEG ratio (Price/Earnings-to-Growth) as a metric to prove these stocks are 'cheap' relative to their expansion. Brian also points out that many of these companies are transitioning from losses to significant profitability, which often precedes a major stock price surge.

Sterling Infrastructure (STRL): Brian describes this company as a critical contractor for hyperscale data centers that self-performs site work to avoid delays. Brian notes that their data center revenue grew 174% year-over-year and they have a $5.15 billion backlog against $2.49 billion in annual revenue. Brian considers the stock undervalued with a PEG ratio of 0.88.

Oracle (ORCL): Brian highlights Oracle's role in renting GPU capacity to AI labs like OpenAI through its neutral OCI cloud. Brian mentions that the stock was cut in half from its $279 high, yet the business is thriving with a signed contract backlog of $638 billion. Brian expects the stock to climb significantly as free cash flow turns positive over the next two years.

Modine (MOD): Brian points out that AI server racks require intense liquid cooling, which Modine provides through its Airedale brand. Brian notes that one customer committed to over $4 billion in cooling gear through 2029, showing massive forward demand. Brian views the stock as a fair price for growth with a PEG ratio near 0.6.

Broadcom (AVGO): Brian explains that Broadcom designs custom AI chips for giants like Google and Meta while dominating network silicon. Brian mentions the company has $73 billion in booked AI orders and generates 42 cents of free cash flow for every dollar of revenue. Brian argues the stock is cheap on a forward earnings basis with a PEG of 0.48.

Lumentum (LITE): Brian identifies Lumentum as a 'choke point' supplier, providing 50-60% of the laser chips needed for AI optical transceivers. Brian emphasizes that Nvidia invested $2 billion into the company and designated its optics for the next-generation Reuben platform. Brian notes the company's swing to a 22% operating margin as a sign of building momentum.

Credo Technology (CRDO): Brian states that Credo owns 88% of the market for active electrical cables used in AI clusters. Brian highlights their 68% gross margins and the fact that revenue tripled last year. Brian views the stock as relatively cheap because its PEG ratio remains well under 1.0.

SanDisk (WDC): Brian discusses the potential of High Bandwidth Flash (HBF) to challenge HBM in data centers due to higher capacity at similar costs. Brian mentions the company recently wiped out its debt and authorized a $6 billion buyback. Brian notes the stock trades under 10 times next year's earnings despite explosive 251% revenue growth.

Tower Semiconductor (TSEM): Brian highlights Tower as a specialty foundry leader in Silicon Photonics, used to turn electrical signals into light. Brian reports that they have $1.3 billion in orders booked for 2027 and have taken $290 million in upfront payments to reserve capacity. Brian considers the forward PEG of 0.91 to be very reasonable for 50% annual earnings growth.

Cohu (COHU): Brian views this semiconductor testing company as a cyclical play currently at its bottom. Brian notes that while AI is only 2% of sales today, their AI test pipeline is $750 million, which is larger than their entire revenue last year. Brian argues that investors are buying at a price that reflects a downturn just as recovery and AI demand arrive.

Innodata (INOD): Brian explains that Innodata provides the labeled data necessary to train frontier AI models for major tech companies. Brian mentions that revenue has nearly tripled in two years and the company has swung from a loss to a $32 million profit. Brian believes the market is treating the stock as an afterthought despite accelerating growth.