This week’s semiconductor rebound looks more like a sorting event than a green light to buy the whole complex. After the June 5 air pocket and June 18 bounce, export controls, customer concentration and valuation gaps are now separating durable AI winners from names simply riding the theme.
The June chip rebound is being read too generously. A violent selloff that wiped out more than $1 trillion in semiconductor market value on June 5, followed by a sharp rebound on June 18, does not amount to a clean fundamental reset for the whole group. It tells us the trade was crowded, oversold and quick to mean-revert — not that every semiconductor stock deserves to snap back in unison. The market is finally being forced to price semis as different businesses again, and that is healthy even if it is less comfortable for dip buyers.
The easiest mistake here is to treat semiconductors as one AI beta trade. That worked when capital was chasing anything tied to data-center demand, but it breaks down once policy risk and execution risk start showing up in earnings reactions. Broadcom’s early-June stumble was the warning shot: AI semiconductor revenue of $10.8 billion, up 143% year over year, still was not enough to satisfy investors when management reiterated rather than raised its fiscal 2027 AI revenue target. A stock can post strong AI numbers and still get punished if the market decides the future is already fully priced.
That is why valuation discipline matters more now than it did a quarter ago. The group is not trading on one common multiple, and investors pretending otherwise are missing the point of the June washout.
NVDA: 32.32x P/E, 65.5% revenue growth, 63.0% net margin
AMD: 180.85x P/E, 34.3% revenue growth, 13.4% net margin
MU: 56.09x P/E, 48.9% revenue growth, 41.5% net margin
TSM: 40.09x P/E, 33.0% revenue growth, 47.0% net margin
ASML: 63.69x P/E, 15.6% revenue growth, 29.7% net margin
Those numbers do not argue for a blanket sector rebound; they argue for discrimination. NVDA is expensive on sales at 20.21x, but its earnings growth, revenue growth and 63.0% net margin still show a business converting AI demand into extraordinary profitability. AMD, by contrast, is sitting at 180.85x earnings with a 13.4% net margin. That does not make AMD a bad company, but it does make it a much thinner-margin story priced with far less room for disappointment. ASML at 63.69x earnings on 15.6% revenue growth is another reminder that strategic importance does not exempt a stock from valuation gravity.
The bulls are not wrong that the AI capex cycle remains real. TSMC has said it is still trying to keep up with demand and expects revenue to rise again in 2026, and ASML has lifted its 2026 revenue outlook as AI drives tool demand. But that is exactly why this is a sorting event rather than an all-clear: a real cycle can still produce bad entry points, uneven earnings conversion and sharp multiple compression. Strong end demand is not the same thing as universal upside when some names are already priced for near-perfect execution.
Export-control enforcement is the other reason the rebound should not be mistaken for safety. The Commerce Department’s move to close a loophole around advanced chip access for Chinese entities outside China means policy risk is still active, not fading into the background. Reuters also reported that about 10 Chinese firms had been cleared to buy Nvidia’s H200, but no deliveries had been made. That matters because restrictions do not hit every name the same way: compute leaders, foundries, equipment makers and memory suppliers each face different exposure to timing disruptions, customer rerouting and demand substitution. A rebound in the tape does not erase that policy asymmetry.
The memory trade adds another layer of separation. MU is up 278.1% year to date, far ahead of NVDA at 12.6%, TSM at 47.3% and AVGO at 15.0%, which tells us the market is already segmenting winners by where AI spending is landing. Memory is being treated as its own pulse check ahead of Micron’s June 24 earnings, not as a generic read-through for every semiconductor name. If Micron delivers, that may validate demand in one part of the stack; it will not automatically justify paying 180.85x earnings for AMD or 63.69x for ASML. The June bounce makes more sense when viewed as investors re-underwriting subsegments — compute, memory, foundry, equipment, networking — rather than blessing the entire sector.
Even the strongest names should not be exempt from scrutiny. AVGO has a respectable 38.8% net margin and a low 0.51 PEG, but its 66.06x P/E still leaves it vulnerable to the same "good, not good enough" reaction that hit after its June report. TSM, on the other hand, looks more defensible at 40.09x earnings with 47.0% net margins and 33.0% revenue growth because it sits closer to the picks-and-shovels center of the AI buildout. That is the distinction investors need to make now: not who has AI exposure, but who has the cleanest path from AI demand to durable earnings without depending on heroic assumptions.
The right read on this month’s volatility is not that semis are broken, and it is not that the all-clear has sounded. It is that the market is becoming less forgiving. After a selloff of more than $1 trillion in value and a reflex rebound a couple of weeks later, the easy phase of buying the whole semiconductor basket looks over.
We would change our mind if export-control risk clearly eased and more companies started converting AI demand into raised guidance rather than simply reaffirming lofty targets. Until then, the chip dip should be treated as a test of quality, margins and valuation discipline. The winners can still win from here. The passengers are the ones most likely to discover that a rebound in the index was never the same thing as a sector all-clear.
Our take, not advice. This is opinion commentary — informational only, not personalized investment recommendations. Markets carry risk. Do your own research and consider your own situation before any trade.
▌The Daily Briefing · Free
A new stock idea, every evening.
One stock worth watching each weekday, plus the analysis behind it. Free, in your inbox.
