TSMC VP Tien Bor-Zen buys again while AI capex stays hot

The foundry trade still runs through one company. S&P Global puts TSMC’s share of leading foundries at roughly 67 to 72 percent, and that gap matters because advanced-node capacity is not something rivals can conjure with a better slide deck. Nvidia may command the loudest AI multiple, but TSMC is the toll booth, and Samsung Electronics and Intel remain the laggards in the race that matters here, advanced process execution.

Before the filing even enters the picture, the bull case is already visible. TSMC reported second-quarter 2026 results that beat expectations, lifted full-year revenue growth guidance to above 40 percent in U.S. dollar terms, and raised the 2026 capital-expenditure midpoint to $62 billion. Price increases of up to 10 percent on certain chips are scheduled from 2027, which is what tight capacity looks like when the supplier has leverage and the customer base has few alternatives.

Our cohort data for director-level buys at mega-cap names is not a forecast, but it is not nothing either. The bucket has a 54.7 percent 90-day win rate across 2,886 cases, with a 3.94 percent average return over 90 days and a 53.84 percent average return over 365 days. That is the kind of historical backdrop that makes a buy more interesting when the company already has operating momentum and the stock has not exactly been left for dead.

Tien Bor-Zen kept buying into strength

The latest filing, reported in a July 30, 2026 SEC Form 4, shows Tien Bor-Zen, a vice president, buying on July 28. The disclosed trades included 15 American depositary shares at $390 each and 1,000 common shares at approximately $67.97 each, for a combined value near EUR 59,739 after euro-normalisation at ingest. Earlier July purchases by the same executive added several thousand shares at prices in the $74 range, which increased his direct holdings.

That is the part that gives the filing some texture. One buy can be noise. A series of buys by the same operating executive, in the same month, at different price points, is a different animal. It does not turn a filing into a thesis on its own, but it does tell you the insider was willing to keep adding while the stock was already strong and while the company’s operating story was still being rewarded by the market.

The market value of the latest purchase was tiny relative to TSMC’s size, under 0.01 percent of market value according to InsiderTrades data. That matters. A purchase of roughly EUR 59,739 is not the kind of number that changes a capital allocation story at a company with a market cap of EUR 1.816 trillion. It is, however, enough to show that the executive was not merely filing paperwork from a stale grant or a routine vesting event. He bought stock.

The cluster detail also matters, though only so far. InsiderTrades data shows the name as a cluster, with 2 distinct insiders and 12 recent declarations, including multiple July 23 and July 30 buys by Tien Bor-Zen. That is a real pattern, and it is more useful than a lone line item. Still, the cluster is not a substitute for price, valuation, or the company’s own operating cadence. It is one more piece of evidence that the people inside the company were willing to commit fresh money while the stock was already well bid.

Why the filing matters less than the setup

The obvious problem with leaning too hard on this filing is that TSMC is not coming off a collapse. Shares closed at $403.31 on July 30 after a 7.64 percent gain that day, and the stock has traded between $223.70 and $479.00 over the last 52 weeks. That is a strong chart, not a distressed one. When a name is already near the upper end of its range, an insider buy can still matter, but it has less room to be read as a contrarian tell.

The second problem is scale. A purchase of 15 ADS at $390 and 1,000 common shares at about $67.97 is a real buy, but it is not a balance-sheet move, and it is not a signal that the executive sees a dramatic mispricing. It is a modest addition by a vice president in a company whose market cap is larger than most countries’ annual output. You can respect the action without pretending it is bigger than it is.

The third problem is that the stock already has a lot of good news in it. The company beat second-quarter expectations, raised revenue guidance above 40 percent, and lifted capex to $62 billion. Analysts have been moving with the story too, with Needham raising its price target to $530 from $480 while keeping a Buy rating. That is a crowded bullish frame. When the market has already accepted the AI foundry narrative, an insider buy has to work harder to add information.

There is also a timing wrinkle. The latest filing came after earlier July purchases at roughly $74, which means the same executive was buying before the latest leg higher and then again after the stock had already moved. That can be read as persistence, or simply as a willingness to average in. The filing does not tell you which. It only tells you the buy happened, and that the executive kept doing it.

Historical cohort context for the July buys

The cohort read is useful here because it keeps the filing in proportion. Director-level buys at mega-cap names have historically produced a 54.7 percent 90-day win rate, a 3.94 percent average 90-day return, and a 53.84 percent average 365-day return across 2,886 cases. That is a decent historical backdrop, especially for a company with real operating momentum. It is also exactly what it says on the tin, history.

You should not stretch that into a promise. TSMC is not the average case in the bucket, and the market regime is not frozen in amber. The Federal Reserve is holding its target range at 3.5 to 3.75 percent, inflation is still sticky, and the AI trade has already done a lot of work for semiconductor multiples. A cohort with a modestly positive 90-day average can coexist with a stock that stalls, chops, or keeps running. The math does not rescue you from valuation.

The strategy framework in our data is best treated the same way. The live out-of-sample tokens are 0.53, 17.1, and 51.5, and they survive only on a restricted EU venue universe with a short, single-regime window. That is a screen, not an alpha claim. It is useful for context, not for pretending the next 90 days are prewritten.

What the cohort does do is sharpen the question. If a vice president at a mega-cap foundry keeps buying in July, and the company has just raised guidance and capex, the burden shifts to the bear case. You are no longer asking whether there is any evidence of confidence. You are asking whether the confidence is already fully reflected in the stock, and whether the next leg of the AI buildout is enough to justify what the market has already priced.

Macro conditions and the foundry trade

The broader market is still being carried by AI-related optimism, and that matters because TSMC sits at the center of the hardware chain that makes that optimism real. JPMorgan’s market outlook notes that major central banks are largely on hold after 2025 easing cycles, while the Fed’s July 2026 monetary policy report shows the policy rate still anchored in restrictive territory. That combination usually does not kill growth names outright, but it does keep the market selective. Cash flow and execution matter more when rates are not falling fast enough to bail out weak stories.

TSMC has the kind of execution that survives that filter. The company’s scale, its advanced-node lead, and its pricing power are the reasons the stock has been able to absorb a lot of good news without immediately breaking. The raised capex midpoint to $62 billion is not a sign of caution. It is a sign that demand is still strong enough to justify more spending, and that the company expects the bottleneck to remain real.

That is also why the peer comparison matters. Nvidia can trade on the demand side of AI, but TSMC trades on the supply side, where the economics are usually less forgiving and the barriers to entry are higher. Samsung and Intel are still trying to close the gap in advanced process technology, and the gap has not closed. If anything, the scale advantage has widened over recent years, which is why the market keeps treating TSMC as the default beneficiary when AI capex stays hot.

The stock’s own move reinforces that point. A 7.64 percent daily gain is not the sort of move you usually see in a name the market has forgotten. It is what you get when a high-quality semiconductor leader prints numbers, raises guidance, and keeps the AI trade alive. The insider buy sits inside that strength, not against it.

The long case is real, but the valuation and the size of the filing still matter

If you want the strongest honest long case, it starts with the business, not the filing. TSMC has dominant foundry share, a raised revenue outlook, a larger capex plan, and a pricing environment that still favors the supplier. The company is not fighting for relevance. It is setting the pace. That is why the stock can trade near its highs and still attract buyers.

The insider filing adds a useful layer because it came from an operating vice president, not a passive director, and because it came in a cluster. The repeated July buys suggest the executive was willing to keep adding while the stock was already strong. That is more informative than a one-off token purchase. It says something about internal willingness to own the name through a strong tape, even if it says nothing definitive about the next quarter.

But the catch remains the same. The filing is small relative to the company, the stock has already had a strong run, and the bullish narrative is well known. You are not discovering TSMC through this Form 4. You are checking whether the people running parts of the business are still willing to buy into the same story the market is already paying for. They are.

That leaves you with a balanced read, not a clean one. The bull case is straightforward, and it is strong. The risks are just as straightforward, and they are not trivial. TSMC can keep compounding, but the stock is no longer priced like a secret. The next thing to watch is whether the company’s July buying pattern continues into August filings, because another cluster would tell you the conviction is still there even after the stock’s latest push to $403.31.

What to watch in the next filing

The next useful data point is not another headline about AI demand. It is whether Tien Bor-Zen keeps buying, whether the cluster broadens beyond the same July pattern, and whether the company’s operating momentum keeps outrunning the stock’s already strong move. If the next Form 4 shows more of the same, the market will have to decide whether to treat it as routine accumulation or as a continuing internal vote of confidence.

The other thing to watch is whether the stock can hold its gains while the market digests the raised capex and the 2027 pricing plans. Those are good signs for the business, but they also tell you the cycle is still active and the market is still paying up for it. If the shares keep climbing, the insider buys will look more like confirmation than foresight. If the shares stall, the same filings will look more interesting in hindsight.

For now, the evidence points in one direction. TSMC remains the dominant foundry play, the AI capex cycle is still feeding it, and a vice president kept buying in July after earlier purchases in the same month. That is enough to keep the name on the list, not enough to declare the job done.

Dig deeper: TAIWAN SEMICONDUCTOR MANUFACTURING CO LTD's full insider filing history.

This is not investment advice.