This week (July 13–19), the tech market can be summed up in one line. A debate over “AI bubble or supercycle” rattled chip stocks and pulled the KOSPI down more than 8% in a single week — yet the companies that actually stepped up to report earnings posted record numbers, pushing back hard against the bubble thesis. In this review, we unpack what really drove the “AI bubble” scare, look at how TSMC and Netflix answered with their earnings, cover Korea’s record first-half ICT exports, and preview next week’s Big Tech earnings rush — all from a tech-industry lens. 📊

TL;DR

  • A chip-driven “AI bubble” debate sent the KOSPI down 6.37% on July 16 alone and 8.77% for the week (close of 6,820.60), spiking volatility.
  • Earnings told the opposite story: TSMC’s Q2 net profit hit a record +77.4%, and H1 ICT exports of $253.8B topped half of all exports for the first time.
  • Netflix’s results met expectations, yet its stock fell 8–9% after hours purely because Q3 guidance came in soft — a lesson in “the weight of expectations.”
  • Next week’s key watch items are late-July Big Tech earnings (Microsoft, Alphabet, Meta, Amazon, Apple) and the FOMC. Whether AI capex keeps flowing goes on trial.

📉 Why Chips, of All Things, Shook the Market This Week

The epicenter of this week’s volatility was the “AI bubble” debate around semiconductors. The KOSPI dropped 6.37% (−463.81 points) on July 16 alone and closed the week down 8.77% at 6,820.60. The KOSDAQ also fell 5.44% for the week to 791.84.

Chips and tech stocks were at the center of the decline. Some in the market voiced so-called “AI bubble” fears — that AI infrastructure investment is past its peak — along with talk of a memory “peak-out,” triggering a flood of profit-taking. On the other side, many analysts at home and abroad pushed back, arguing AI infrastructure spending is still in its early stages and that it’s too soon to read this correction as a trend reversal. In short, this was a week in which the psychology of the “AI bubble” debate moved ahead of the fundamentals.

Adding to it, international oil prices surged about 14% over the week, layering on inflation worries and fueling risk-off sentiment. That macro thread was covered in last weekend’s economic review, so this retrospective focuses on the signals from the tech industry itself.

🏭 TSMC Answered the Bubble Thesis With Results

In the thick of the debate, TSMC pushed back with record earnings. Its Q2 (April–June) net profit, reported on July 16 local time, came in at NT$706.56 billion (about $22 billion), up 77.4% year over year — a record for a single quarter. Revenue reached $40.2 billion (about NT$1.2703 trillion), up 33.7% in dollar terms, with a gross margin of 67.7%. That marks nine straight quarters of double-digit profit growth.

The engine was unmistakably AI. High-performance computing (HPC), anchored by AI accelerators, grew 20% in a single quarter and now accounts for 66% of total revenue. TSMC raised its full-year revenue growth outlook (in dollars) to “slightly above 40%” and lifted its annual capex plan from $52–56 billion to $60–64 billion. It also announced an additional $100 billion investment in Arizona, taking cumulative investment there to $265 billion.

To put it plainly: contrary to the market’s fear that “AI investment is rolling over,” the world’s No. 1 foundry — the company that actually makes the AI chips — instead raised both its outlook and its investment.

🎬 Netflix: Results Met Expectations, So Why Did the Stock Drop?

Netflix left the opposite lesson: the results were solid, yet the stock plunged. Its Q2 revenue, reported July 16 local time, was $12.6 billion (about ₩18.6 trillion), up 13% year over year and in line with expectations. Earnings per share came in at $0.80, edging past the $0.79 estimate, with an operating margin holding at 33.4%. Membership growth, price increases, and expanding ad revenue drove the growth.

Even so, the stock fell 8–9% after hours. The culprit was expectations. Q3 revenue growth guidance of +11.7% (roughly $12.8 billion) came in below the market’s bar (around $13 billion), and the full-year outlook was merely narrowed to $51.0–51.4 billion rather than raised. It’s a clear case that merely “meeting” expectations isn’t enough — a stock reacts only when guidance clears the forward bar.

Put TSMC and Netflix side by side and the grammar of this earnings season comes into focus: the market is reacting more to “how much more will they earn going forward” — guidance and outlook — than to “how much did they just earn.”

📦 A Supercycle Proven in Numbers — H1 ICT Exports

Separate from the bubble debate, Korea’s trade data put the semiconductor supercycle into hard numbers. First-half ICT exports, announced July 14 by the Ministry of Science and ICT and the Ministry of Trade, came to $253.86 billion — up 120.5% from a year earlier ($115.12 billion) and a record for any half-year. ICT’s share of total exports reached 51.1%, crossing the halfway mark for the first time ever.

By category, semiconductors led at $192.43 billion (+162.5%), an all-time high, followed by SSDs at $19.94 billion (+317.5%) and mobile phones at $8.4 billion (+38.0%). June alone saw monthly ICT exports top $50 billion for the first time. The half-year ICT trade surplus reached $160.65 billion, surpassing the previous annual record set in 2018 ($113.22 billion) in just six months.

One caveat is worth noting: much of this surge rests on rising memory prices for AI servers, so if memory prices adjust or investment slows, the growth rate could fall off quickly. That is precisely the fault line between “bubble” and “supercycle.”

🤖 The Quietly Escalating AI Price War

While the market churned, a price war intensified among AI model makers. Through early in the week (as of July 13), major models — OpenAI’s GPT-5.6 (Sol, Terra, Luna), xAI’s Grok 4.5, and Meta’s Muse Spark — launched in quick succession within a week, kicking off a “half-price model” contest.

The biggest shift is that as performance levels out at the top, the battleground has moved to “price” and “actual task performance.” The per-million-token price gap between top premium models and low-cost ones has widened to roughly 5x, making cost a core variable in corporate adoption decisions. Because this trend feeds directly into demand for AI infrastructure (chips and data centers), it moves as one with the chip earnings and bubble debate above.

🔭 Next Week’s Outlook — The Big Tech Earnings Rush and the FOMC

The stretch from next week through the end of July is the real turning point in this AI debate. If TSMC and Netflix gave individual companies’ answers this week, the next test is Big Tech — the deep-pocketed spenders behind AI investment.

  • Big Tech earnings rush: Around July 29 local time, Microsoft, Alphabet, Meta, Amazon, and Qualcomm are due to report, with Apple on July 30. Whether their AI capex plans hold or expand is the key gauge that will set the direction of the “AI bubble” debate. If cloud giants keep their investment scale intact, the chip-demand narrative gains support.
  • FOMC (Federal Open Market Committee): The U.S. rate decision meeting is held on July 29 local time, with the outcome landing early on July 30 Korea time. The rate path directly affects tech-stock valuations, so it’s a variable to watch alongside earnings.
  • Nvidia: Nvidia’s earnings — the symbol of the AI cycle — are slated for late August (expected Aug 25), so the debate is likely to run into August.

From a Korean-market perspective, these Big Tech firms’ capex and guidance act as the rudder for domestic supply-chain demand across HBM, advanced packaging, and foundry.

The Bottom Line

This week’s tech market was a clash of “psychology” versus “results.” Stocks were shaken by “AI bubble” fear — the KOSPI fell 8.77% for the week — yet the numbers from companies that actually reported (TSMC net profit +77.4%, H1 ICT exports +120.5%) backed up the reality of a supercycle. At the same time, the Netflix case showed today’s jittery market psychology: even good results sink a stock if they don’t clear the bar.

Three points to watch: First, this upswing leans heavily on a single axis — AI server memory prices — so growth rates could wobble if prices adjust. Second, the market has grown more sensitive to “forward guidance” than to results themselves. Third, next week’s Big Tech earnings and capex should deliver a substantive answer to the bubble debate. Ultimately, “bubble or supercycle” will be largely settled by Big Tech’s investment plans next week.

※ This article is for informational purposes only and is not investment advice.

Sources