Top Chip Analyst: Semiconductor Oversupply Is Nearly Impossible Before 2028. Why He’s Bullish on Memory

A top Wall Street chip analyst says the semiconductor selloff gripping markets has almost nothing to do with actual supply and demand, and the real reason why comes down to a physical constraint most investors are overlooking.

Published July 30, 2026, 8:22am ET · 3 min read

A magnificent golden bull statue is depicted mid-charge on a paved city street, with sparks and golden particles erupting around its hooves. In the blurred background, tall city buildings are visible, and translucent golden financial charts and data are overlaid on the left side of the image. The bull faces directly forward, conveying power and dynamism.
A golden bull, a classic symbol of a surging market, charges forward amidst digital financial data, reflecting the accelerated capital markets and bullish earnings predictions for big banks. © 24/7 Wall St.

Wolfe Research Senior Analyst Chris Caso used a CNBC appearance on July 29 to defend the semiconductor sector after a sharp pullback, arguing that the physical infrastructure needed to trigger an oversupply cycle simply does not exist yet. The SOXX index has pulled back 25% from its recent high following SK Hynix’s earnings report, but he believes the classic chip-cycle bust remains years away because fabs and data centers cannot be conjured up on demand.

Caso argued: You just don’t have the physical space to make the semiconductors right now. So it’s really hard to see a situation right now where we’re in oversupply.” He added that “the current up cycle requires the building of brand new buildings that take a long time to build, and even the potential of getting to that oversupply situation is really 2028 at the very earliest.

NVIDIA and AMD Show AI Demand Is Still Accelerating

Caso pointed to TSMC being completely sold out, with no physical space to make semiconductors. That constraint radiates through the entire supply chain, from advanced logic customers like NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) and Advanced Micro Devices (NASDAQ:AMD) down through memory and storage.

NVIDIA’s Q1 FY2027 results underscore the demand backdrop. Revenue reached $81.61 billion, up 85.2% year over year, with Data Center at $75.25 billion and networking up 199%. Management guided Q2 revenue to $91.0 billion and disclosed total supply-related commitments of $119.0 billion. CEO Jensen Huang described the moment as the largest infrastructure expansion in human history.”

AMD’s Q1 2026 earnings report reinforced the same theme. Revenue came in at $10.25 billion, up 37.9%, with Data Center revenue of $5.78 billion, up 57%. CEO Lisa Su said “leading customer forecasts exceeding our initial expectations” for the MI450 series.

Memory Supply Is Where the Shortage Looks Most Severe

Caso was blunt on DRAM and HBM: The memory suppliers are severely supply constrained. They can’t produce more, which is the very reason why we’re bullish on memory.” He noted Micron reported about a month before SK Hynix, with one of the best reports in company history.

Micron Technology (NASDAQ:MU) posted Q3 FY2026 revenue of $41.46 billion, up 345.7% year over year, with GAAP gross margin of 84.6% and Q4 guidance of $50.0 billion in revenue at roughly 86% gross margin. CEO Sanjay Mehrotra highlighted that multi-year Strategic Customer Agreements now anchor the order book, with HBM4E volume production expected in calendar 2027.

MU earnings quotes

Chip Equipment and Storage Stocks Confirm the Capacity Crunch

KLA Corporation (NASDAQ:KLAC) sits at the front of any capacity buildout. Q4 FY2026 revenue was $3.66 billion, up 15.2%, with Q1 FY2027 guided to $4.0 billion. CEO Rick Wallace said momentum is “accelerating in the second half of calendar 2026 and continuing through 2027.” Shares still fell 6.18% on July 28 as investors focused on valuation.

Storage tells the same story. Western Digital (NASDAQ:WDC) reported Q3 FY2026 revenue of $3.34 billion, up 45.5%, with non-GAAP gross margin crossing 50.5%. CEO Irving Tan noted that “virtually every AI workload, from training, inference, agentic AI to physical AI, creates data that is stored persistently and cost-efficiently on HDDs.”

NVIDIA Is Using Its Balance Sheet to Protect the AI Supply Chain

Caso also defended NVIDIA’s supply-chain investments. He cited seven tranches of bonds issued in June at a weighted-average spread of less than 0.5% and noted that NVIDIA backstopped $500 billion to SK Hynix and $250 billion to OpenAI. His view: “I’m pretty comfortable with providing equity and such to shore up the supply chain because the capital is required for your customers to expand capacity.”

If Caso’s capacity math is right, the recent semiconductor selloff reflects a shift in sentiment rather than a fundamental change in the underlying cycle. Investors should watch fab construction timelines, HBM4E qualification progress and hyperscaler capex commentary for the first concrete signals that meaningful new supply could finally arrive in 2028. Until then, the physical constraints supporting the semiconductor cycle appear to remain firmly in place.

Contact [email protected] for any questions or corrections.

Thomas Richmond

Thomas Richmond is a financial writer and content strategist with 5+ years of experience covering stocks and financial markets. He has published over 250 articles focused on individual stock analysis, helping investors better understand business fundamentals, stock valuations, and long-term opportunities.

Thomas previously served as a Content Lead at TIKR, a stock research platform, where he helped scale the company’s blog to hundreds of articles per month and contributed to a weekly newsletter reaching more than 100,000 investors.

He specializes in breaking down complex companies into clear, actionable insights for everyday investors, with a focus on fundamentals-driven research.

His work has also been featured on platforms including Seeking Alpha and Sure Dividend.

Outside of work, Thomas enjoys weight lifting and soccer.

All articles →