Hey there, bargain hunter.

Markets are messy right now. Not broken. Messy. That distinction is doing a lot of work at the moment, and most investors are not bothering to make it.

Here is what actually happened. Since hitting an all-time high in late June 2026, the Philadelphia Semiconductor Index has fallen roughly 24%. This week alone, the index is down around 11% — its steepest weekly decline since March 2025. The VanEck Semiconductor ETF has posted its third weekly loss in four weeks, dropping close to 9% in the most recent period. Global semiconductor stocks have shed more than $3.3 trillion in market value since June 22. Apple surpassed Nvidia as the most valuable U.S. company on Friday as Nvidia shares dropped more than 3.5% in a single session.

A bear market. By the numbers. In a sector that was up nearly 99% year to date just weeks ago.

What Actually Happened

The chain of events started with Broadcom. The company’s guidance in early June failed to deliver the upward revision in AI chip forecasts the market had priced in, sending Broadcom shares down between 12% and 19% across multiple sessions. That single data point was enough to trigger a wave of selling that spread far beyond Broadcom itself.

Then TSMC — a company that just reported the strongest quarterly results in its history — poured fuel on the fire by raising its full-year capital expenditure forecast to between $60 billion and $64 billion, up from prior guidance of $52 billion to $56 billion. The market read that as a warning sign about runaway costs rather than what it actually was: a company investing aggressively because demand is real and accelerating.

Add in SK Hynix suffering its largest single-day share decline on record shortly after its blockbuster Nasdaq debut, a fresh Chinese AI model from Moonshot AI narrowing the gap with Western offerings, and geopolitical pressure from the Middle East pushing energy prices higher, and you had a perfect storm of headline risk hitting a sector that had become extremely crowded.

What the market is doing is extrapolating all of these short-term friction points and pricing them as if AI infrastructure spending is about to collapse. The evidence does not support that conclusion.

The Fundamentals Have Not Moved

This is the part people skip.

TSMC just posted record second-quarter 2026 results. Revenue came in at $40.2 billion, up 36% year over year. Net income surged 77.4% to a record for the fifth consecutive quarter, beating analyst expectations of $19.7 billion by coming in at $22.4 billion. Gross margin reached 67.7%. Operating margin hit 60.3%. The company guided Q3 revenue between $44.6 billion and $45.8 billion — another step up. High Performance Computing, which includes AI chips, now constitutes 66% of TSMC’s total revenue.

These are not the numbers of a business facing structural impairment. These are the numbers of a business in the middle of a generational demand cycle.

AMD’s story is similar. Full-year 2025 revenue was a record $34.6 billion. Q4 2025 revenue was $10.3 billion, up 34% year over year. Q1 2026 continued that trajectory, with revenue climbing 38% year over year to $10.3 billion and data center revenue rising 57%. Trailing twelve-month free cash flow reached $8.57 billion, up more than 211% from the same period a year earlier. The company guided Q2 2026 revenue between $13.8 billion and $14.8 billion.

Slight tangent, but it matters: the top three chip stocks still account for 80% of the combined $9.5 trillion market capitalization of the top 10 global chip companies. Concentration like that means when sentiment shifts, selling becomes mechanical, not analytical. A lot of the names dragged down in this selloff were caught in the undertow with no company-specific reason to fall.

Separating Cheap from Broken

Not everything that fell is worth buying. That is the real work here.

Intel remains a difficult case. The company is forecasting Q2 2026 revenue of $13.8 billion to $14.8 billion with non-GAAP EPS of just $0.20. Its foundry business was burning $2.51 billion in operating losses as recently as Q4 2025. At a forward P/E that recently sat near 147 times earnings, Intel’s valuation is asking for a level of execution the business has not yet demonstrated. Only about a third of analysts covering the stock rate it a buy. Intel is not cheap. It is popular for the wrong reasons.

TSMC is a different conversation entirely. This is a business with a 60.3% operating margin, a 55.6% net profit margin, and a management team that just guided revenue higher for the fifth straight quarter. The stock sold off 4% when it reported those record results. That is the kind of disconnect that should get a value investor’s attention.

AMD lands somewhere in the middle. Free cash flow is growing at triple-digit rates. The data center business is accelerating. But the stock had surged roughly 142% year to date before this pullback, and at a price-to-free-cash-flow ratio that had stretched to over 80 times, there was genuine valuation excess baked in. The selloff has compressed that. Whether it has compressed enough depends on how fast earnings continue to grow — and right now, the trajectory still looks intact.

What the Market Is Getting Wrong

The market is treating rising capital expenditures as a warning sign. It might be the opposite.

When TSMC raises its capex forecast from $52 billion to $64 billion in a single quarter, that is not a company spending recklessly. That is a company responding to orders it has already received. The semiconductor shortage is forecast to last at least through 2028, with the broader memory market expected to grow at an 11.6% compound annual rate through 2030. TSMC’s customers — the hyperscalers, the cloud giants, the AI infrastructure builders — are not quietly pulling back. They are locking in capacity.

The concern about AI return on investment is real and worth monitoring. But it is a long-term question being used to justify short-term panic. Hyperscaler capex commitments have not reversed. Demand for leading-edge nodes has not softened. TSMC’s Q3 guidance did not suggest a deceleration.

What changed is that a crowded trade got a few pieces of bad news in quick succession, and investors decided to reduce exposure. That is a positioning story, not a fundamental one.

The Cheap Investor Scorecard

• Business Quality: TSMC and AMD remain category leaders with widening competitive moats. Score: High.
• Financial Strength: TSMC posted a 60.3% operating margin and 55.6% net margin in Q2 2026. AMD free cash flow up 211% year over year. Score: Strong.
• Valuation After Selloff: TSMC now trades at a meaningful discount to recent highs despite record fundamentals. AMD valuation compression is real but still above historical averages. Score: Improving.
• Catalyst Strength: Q3 earnings season begins shortly. Hyperscaler capex commentary will be the key data point. Score: Near-term binary.
• Competitive Position: TSMC manufactures chips for virtually every major AI player. AMD is gaining share in AI accelerators and data center CPUs. Score: Durable.
• Balance Sheet: Both companies carry manageable debt relative to cash generation. Score: Sound.
• Management Execution: TSMC has beaten estimates for five consecutive quarters. AMD posted record revenue and free cash flow in 2025. Score: Consistent.
• Margin of Safety: TSMC’s selloff following record results offers the clearest margin of safety in the sector. Score: Present and measurable.
• Long-Term Potential: AI infrastructure buildout, sovereign AI investment, and 2nm node ramp all point to sustained demand through the end of the decade. Score: Compelling.
• Broken vs. Cheap: Intel carries execution risk and stretched valuation. TSMC and AMD show no signs of fundamental impairment. Score: Intel = caution. TSMC and AMD = opportunity worth evaluating.

Bull, Base, and Bear

Bull case: Hyperscalers report strong Q2 earnings and reaffirm AI capex plans. The sector recovers as investors recognize the selloff was positioning-driven. TSMC’s 2nm ramp accelerates. AMD closes the gap with Nvidia in AI accelerators. Investors who bought during the panic are rewarded within 12 to 18 months.

Base case: Earnings season brings mixed signals. Some hyperscalers trim capex language. The sector stabilizes but does not roar back immediately. Fundamentals continue to improve, and patient holders are eventually rewarded as AI demand proves durable.

Bear case: AI capex spending peaks sooner than expected. Chinese competition from models like Kimi K3 reduces the hardware requirements for cutting-edge AI. The geopolitical environment deteriorates further. A genuine demand air pocket develops and the current selloff turns out to be the beginning of a longer correction, not a buying opportunity.

The Bottom Line

If TSMC reports record results for the fifth straight quarter and its stock falls 4% on the day, the market is not doing fundamental analysis. It is managing fear.

That is not a reason to be reckless. Valuation still matters, and some names in this sector were priced for perfection before this correction began. But there is a difference between a business whose fundamentals are deteriorating and a business whose stock is falling because the sector got crowded and then uncomfortable.

The question to ask right now is not whether chips are a good idea over the next three months. It is whether the businesses you are looking at will generate significantly more free cash flow five years from now than they do today. For TSMC, the answer seems clear. For AMD, the direction is compelling even if the entry point is still sorting itself out. For Intel, the turnaround still requires more proof.

Markets periodically hand you the chance to own great businesses at prices that would have seemed impossible a month earlier. Whether this is one of those moments depends on your conviction, your time horizon, and your tolerance for further short-term turbulence.

Worth thinking about carefully.

— The Cheap Investor