TSMC posted record profit and raised guidance, and chip stocks fell anyway

[Freya]

Taiwan Semiconductor reported a 77 percent year over year surge in Q2 net profit to a record $21.99 billion, comfortably beating Wall Street's EPS estimate with a print of $4.31 against $3.80 expected. Revenue hit $40.2 billion, at the very top of the company's own guidance range, with gross margin at 67.7 percent and operating margin at 60.3 percent, both above the high end of what TSMC had guided for

Despite the clean beat, TSMC shares fell around 2 to 4 percent, and the drag spread across the wider chip sector, Nvidia, AMD, Intel and Micron all traded lower on the news. The trigger wasn't the results themselves, it was the forward guidance, TSMC raised its 2026 capital expenditure range to $60 billion to $64 billion, up from a prior $52 billion to $56 billion, roughly a 15 percent jump, and said capex over the next three years will be even more significantly higher than the previous three year period. As part of that spending, the company announced another $100 billion in planned US investment, taking its total Arizona commitment to $265 billion

High performance computing, the category covering AI accelerators, made up 66 percent of TSMC's revenue this quarter, up from 61 percent in Q1, underlining just how tightly the world's most important chipmaker is now tied to the AI buildout specifically. TSMC also raised its full year revenue growth forecast to slightly more than 40 percent, up from a prior 30 percent plus target, and guided Q3 revenue to $44.6 billion to $45.8 billion, comfortably ahead of the $43.11 billion analysts expected

The market reaction comes down to one specific worry articulated by Vital Knowledge analyst Adam Crisafulli, that a sharply higher capex budget raises the same debate as always about whether semiconductor demand is cyclical or genuinely secular, and if it turns out to be more cyclical, accelerating capacity additions right now is usually read as a signal to fade the sector rather than chase it further. TSMC chair C.C. Wei pushed back on that framing directly, telling investors the company's conviction in the multi year AI megatrend remains very high, but with free cash flow already declining as capex ramps, the market's skepticism about whether this spending translates into durable returns clearly hasn't gone away

Tsmc And Ai Stocks Fall Despite Upbeat Earnings Heres Whats Worrying Investors 4794941www.investing.com

[Grover26]

Record profit, guidance beat, and the stock still falls purely because they're spending more to meet demand is such a strange but very telling reaction, shows how nervous the market actually is about AI capex right now

[Tracey49]

The cyclical versus secular debate is really the whole ballgame for the entire AI trade at this point, this TSMC report didn't resolve it either way, if anything it just raised the stakes on getting that call right

[Harper84]

66 percent of revenue now coming from high performance computing versus 61 percent last quarter shows just how fast TSMC's fortunes have become tied to one single growth story rather than a diversified chip business

[Harper48]

Another 100 billion into Arizona on top of the existing 165 billion commitment is a massive vote of confidence in US manufacturing, whatever the stock does in the short term

[CodyRhodes29]

Free cash flow declining while capex ramps is the number that actually matters here long term, impressive revenue growth means less if it's not converting into cash the company can actually deploy or return to shareholders

[AJStyles]

Wei's conviction in the multi year AI megatrend comment reads almost like a direct rebuttal to the skeptics, curious whether the next couple of quarters end up proving him right or validating the cyclical worry instead

[Dark Elizabeth]

Market reaction makes perfect sense once you zoom out a bit. When a stock runs hard into earnings, the bar is no longer "beat estimates," it is "obliterate and raise more than expected."

TSMC delivered a massive print, no debate there. But positioning was already crowded. Funds piled into anything AI-adjacent for months.

So what happens next is mechanical. Profit taking, rebalancing, and some rotation into laggards.

Also worth noting that guidance, while raised, did not signal a fresh acceleration curve. It confirmed strength, not a new leg higher.

In that context, a pullback is not bearish. It is digestion. Big difference.

[BackRowBob]

This is classic "sell the news" behavior dressed up as macro commentary. Happens every cycle, just with different buzzwords.

Nvidia had the same pattern around certain quarters. Blowout numbers, stock dips, then resumes trend weeks later.

Traders front-run the narrative, not the print. By the time earnings land, the easy money is already made.

What matters is whether demand actually rolls over. Nothing in TSMC's report suggests that yet.

If anything, the backlog and capacity constraints say the opposite.

[Stuart78]

People keep underestimating how much expectations matter. A 77 percent profit jump sounds insane, but the stock was pricing in something close to that trajectory already.

AI demand is not a surprise anymore. It is consensus.

So the real question becomes: can it exceed consensus again next quarter?

If the answer is "maybe" instead of "definitely," you get hesitation.

Markets are forward-looking to an almost absurd degree

[Tiger]

Another angle: margins. Everyone is cheering revenue and profit growth, but the durability of those margins is what long-term investors care about.

If customers like Nvidia, Apple, and AMD start pushing back on pricing, that changes the equation.

TSMC has enormous leverage right now, but cycles always normalize.

The selloff could simply reflect some fear that peak pricing power is near.

Not saying it is, but that is the debate underneath the surface.

[RightNutter82]

Macro is playing a role too. Rates, liquidity, and positioning across tech all matter.

Even the best company in the world can drop if the broader risk appetite cools.

Seen this movie before: strong earnings, weak price action, then a rebound once macro stabilizes.

It is frustrating, but it is not irrational.

Equities are not just about fundamentals, they are about flows.

[NightCrawler33]

The AI megatrend is real, but the timeline is where arguments happen. Wei is clearly signaling confidence in multi-year demand.

Skeptics are questioning whether we are in a front-loaded capex burst.

If hyperscalers slow spending even slightly, the ripple hits TSMC quickly.

That is why the next couple of quarters matter more than this one.

Confirmation beats conviction every time.

[DataStream Luca]

Something else: currency and geopolitics quietly sit in the background of every TSMC discussion.

Taiwan risk never fully goes away, and large funds constantly price that tail risk.

It does not need to materialize to affect valuation multiples.

So even with stellar execution, there is a ceiling on how aggressively some investors will bid.

That discount shows up during moments like this.

[WearyCoder]

Feels like people wanted a "blow the doors off" guide raise, not just a solid one.

When sentiment is stretched, good becomes disappointing.

Same thing happened with cloud companies last year. Strong growth, but not accelerating fast enough, and down they went.

Then a few months later, everyone realized nothing was actually broken

Patience tends to win in these setups.

[Jordan_68]

The supply chain angle is fascinating. TSMC is the bottleneck, but also the enabler.

If they expand capacity too aggressively, they risk future oversupply.

If they stay tight, they maximize margins but cap volume growth.

Balancing that is incredibly tricky.

Investors are trying to game that balance in real time.

[Dylan]

Retail tends to read headlines, institutions read positioning. That gap explains a lot of these reactions.

Headline: record profit. Price action: down.

Conclusion for newcomers: market is broken.

Reality: positioning was skewed, so the path of least resistance was lower.

Not exciting, but very common.

[NeonTundra]

One contrarian take: a pullback here is healthy for the broader AI trade.

Parabolic moves invite sharp corrections later.

Cooling off resets expectations and creates better entry points.

If TSMC just kept ripping nonstop, that would be more concerning.

Markets need friction to sustain trends.

[StringTheory97]

Look at how semis trade as a group. Correlation is high.

If one big name stumbles or even pauses, the whole sector wobbles.

TSMC dropping pulls on Nvidia, AMD, ASML, the lot.

It becomes less about individual performance and more about sector rotation.

That is why the reaction feels exaggerated.

[SortedMate]

The funniest part is watching sentiment flip in real time. Yesterday it was "unstoppable AI demand," today it is "peak cycle?"

Nothing fundamentally changed in 24 hours.

Price moved, narrative followed.

Give it a few weeks and the bullish case will dominate again if data holds.

Same script, different quarter

[Crossing65]

Another piece people ignore: capex cycles are lumpy.

Customers do not spend in straight lines. They surge, pause, then surge again.

TSMC sits right in the middle of that volatility.

So even with a strong long-term trend, short-term numbers can spook investors.

That tension drives moves like this