TOP STORY
📉 Chip Stocks Have Worst Week in Over a Year - Here’s Why

📉 If you’ve been reading the Buzz these past few weeks, you’ll know that since Micron’s ($MU) blowout earnings back in late June (which we covered here), chip stocks have been facing rough waters.

📊 And this week, those rough waters turned into the sector’s worst week in over a year, with the Semiconductor Index ($SMH) dropping 9%, now ~20% below its June highs.

🇨🇳 Add on top of that a ‘DeepSeek 2.0’ scare from China, after a new Chinese AI model was released that competes with top American models at half the cost… and you get the Nasdaq’s worst week since April. Across the indexes:

• The S&P 500 fell -1.55%

• The Nasdaq 100 fell -4.13%

• The TSX fell -0.12%

🤿 So let’s dive into what’s going on, what the new Chinese AI model means for your portfolio, and a few other top stories of the week (like $IBM crashing 25% after earnings)…

👾 Chip Stocks Have Their Worst Week in Over a Year

🌼 With nearly 1/3 of Blossomers holding chip stocks (such as Nvidia, Micron, AMD, Broadcom, and TSM), many of you are likely wondering what caused this crash and where we’re going from here, so let’s talk about it.

😰 The drop started on Monday after the investment firm Korea Investment & Securities published its Q2 2026 profit estimate for SK Hynix, which came in 8% below analyst consensus due to “slower-than-expected” shipments of its newest AI memory chips.

📉 This sent SK Hynix shares down 15% in a single day (its worst day ever), just days after its blockbuster U.S. IPO (which we covered here), setting off a domino effect, with stocks like Micron and Sandisk ($SNDK) falling ~6% each.

🩸 $ASML helped stop the bleeding a bit on Wednesday after reporting blowout earnings, raising its 2026 revenue guidance to €43-45 billion and pledging 30% more manufacturing capacity to meet “extremely strong” AI demand, driving a brief rally.

📉 But then Thursday reports that CXMT, a state-backed Chinese memory maker, was prepping for a massive IPO drove chip stocks down again (due to fears of an incoming supply glut).

⏰ A Wake-up Call for Beginner Investors

🚀 The thing is, even with this ‘crash’, chip stocks are still up massively.

• Micron is up 650% in the past year

• AMD’s up 216% in the past year

• TSM’s up 67% in the past year

⚠️ So if this drop is causing you to panic, it probably means you bit off more risk than you have an appetite for.

🎢 Chip stocks are famously more volatile than the broader market and are traditionally quite cyclical, driven heavily by supply and demand dynamics, meaning these sorts of big swings are par for the course.

💡 And according to many analysts, the fundamentals remain strong:

📈 According to FactSet, high-bandwidth memory is sold out through most of 2027, and second-quarter semiconductor earnings are on pace to grow 131% year-over-year… not exactly the sign of a sector in crisis.

✨ On another bright note, we’re officially in earnings season, and so far, ~90% of the S&P 500 companies that have reported have beaten expectations. Next week, another ~80 report, including Tesla ($TSLA) and Alphabet ($GOOG).

💡 All in all, this week looked more like a reset than a fundamentals problem, profits are at record highs, memory is still supply-constrained, and the AI spending behind it all is still flying high. If the demand story is actually cracking, it’ll show up in the earnings numbers within the next three weeks, so we won’t have to guess for long…

👀 But the biggest new threat to the chip trade this week didn’t come from an earnings report. It came out of China from a company named after a Pink Floyd album… so let’s get into it.

😎 But first, a quick word from this week’s sponsor, Plynk!

PRESENTED BY PLYNK

Get a 25% cash match on your dividends with Plynk

Want to earn 25% more on every dividend you earn? With Plynk, you’ll automatically receive a 25% cash match on your dividends earned on the app. If you earn $1,000 in dividends, that’s an extra $250 in your pocket—which you can take as cash or reinvest to help grow your wealth. No opt-ins or complex steps required. Just hold eligible dividend-paying stocks, ETFs, or funds, and your match is applied monthly.

Browse our intuitive categories to discover dividend-paying stocks and ETFs, which you can sort by yield, performance, and size. Plus, track everything in the Income Hub which gives you one clear view to see your match earnings, as well as upcoming dividend and interest payouts.

Want even more? Upgrade to Premium for just $3.99 to double your boost to a 50% match, plus receive a 0.5% cash rate boost to the current rate.

Stop leaving money on the table. Download Plynk today and start earning!

^{*See Plynk Disclaimer at the end of the newsletter}

TOP STORY
🤖 China’s Kimi K3 Gives Markets A Second “DeepSeek Moment”

😳 On Friday, Nvidia briefly lost its crown as the world's most valuable company to Apple. The trigger? A three-year-old Beijing startup released a near-frontier AI model that anyone can download and run for free.

🚀 The Chinese AI start-up “Moonshot AI” unveiled “Kimi K3” on Thursday night: a 2.8 trillion-parameter AI model, making it the largest open-source AI system ever released that competes head-to-head with top American models… at roughly half the cost.

🌙 Moonshot AI is a 3-year-old Beijing startup founded in 2023 by Yang Zhilin (a Carnegie Mellon graduate who previously worked at Meta and Google), backed by Alibaba, Tencent, and Meituan.

📊 On independent Arena benchmarks, the model beat both Anthropic’s Fable 5 (widely considered the current best model) and OpenAI’s GPT-5.6 Sol on front-end coding tasks, and outscored Anthropic’s previous flagship model, Opus 4.8, on general text performance, at a price of just $3 per million input tokens and $15 per million output tokens, ~40% cheaper than Opus 4.8, and a fraction of Anthropic’s Fable 5’s $50 per million output.

💬 According to Anastasios Angelopoulos, co-founder and CEO of AI evaluation platform Arena:

🩸 Markets did not take it well. By Friday's close:

• Nvidia fell -2.2%, briefly handing the world's-most-valuable-company title to Apple

• Taiwan's benchmark index sank more than -6%; Japan's Nikkei fell -4.1%

• TSMC dropped -7% (despite reporting a 77% jump in quarterly profit the same day)

• SoftBank (often traded as an OpenAI proxy) fell -9%

• Chinese AI rivals Z.ai and MiniMax cratered -28% and -16% in Hong Kong

🤔 Why Does It Matter?

🏗️ Kimi K3 attacks the single assumption the entire AI trade has been built on, which is “staying at the frontier requires massive, sustained infrastructure spending.”

💵 It’s this idea that’s given Nvidia, memory makers, and hyperscalers (i.e., Amazon, Google, Microsoft, etc.) such enormous valuations over the past few years, because the logic was that only the deepest pockets can afford to compete at the top.

💰 But Moonshot is showing that might not be true. And if frontier-level intelligence can be built more cheaply than the market assumed, then the massive amounts of CapEx being spent by Big Tech may be on shaky ground.

🍔 And real US companies are already switching. DoorDash's CTO says using Moonshot's model is "better quality" at "cheaper cost". Airbnb and Siemens are also both experimenting with Chinese models to cut their rising AI bills.

⚖️ Deja Vu, or Something New?

📅 Now, if this all feels familiar, it is. Back in January 2025, DeepSeek’s original release (another low-cost AI model on par on certain performance metrics with American models) wiped roughly $600 billion off Nvidia’s market cap in a single day.

🔄 But if you remember, stocks recovered within weeks afterward, AI spending never actually slowed, and the Nasdaq went on to set dozens of new highs.

🧑‍🔬 Tech analyst Patrick Moorhead called this week’s reaction an “overreaction eerily similar to that earlier episode.” Morgan Stanley’s Gary Yu framed it as a sign Chinese labs are simply catching up to U.S. leaders on size, performance, and price. And Bernstein’s Robin Zhu called the release simply “confirmatory” of China's status as a real rival in the space.

🧑‍⚖️ There’s also a real fight over how Chinese labs got so good so fast. Anthropic accused Moonshot, DeepSeek, and MiniMax back in February of running campaigns to "illicitly extract Claude's capabilities" through distillation (i.e., training a cheaper model on a stronger model's outputs). Beijing calls the claims "groundless," but if US regulators act on it, that's a whole new front in the AI trade war.

📊 What This Means for Your Portfolio

🧩 The trick to thinking about this week cleanly is remembering that "AI stocks" aren't a single bet. They're three different layers, and Kimi K3 hits each one very differently:

• 1️⃣ The intelligence sellers (OpenAI, Anthropic): this is where the real potential damage is. When a near-frontier model shows up at half the price, and goes open-weight, charging a premium for raw intelligence gets harder to justify. Analysts note OpenAI and Anthropic have already been quietly loosening prices and rate limits in recent weeks. These are mostly private companies, so they don’t impact our portfolios directly, but they are big spenders on infrastructure so could drive a risk there if capex slows.

• 2️⃣ The infrastructure builders (Nvidia, memory stocks, chip stocks): this is where the core debate lives. The bear case is that if frontier AI is cheaper to build than assumed, the ~$700 billion the hyperscalers are spending this year is overkill. The bull case is the "Jevons paradox": when something gets cheaper, we use dramatically more of it, not less. That's what happened after DeepSeek: a panic, then compute demand hit new records within months.

• 3️⃣ The intelligence users (Apple, and most of the S&P 500): These are the quiet winners of cheaper models. If intelligence is more commoditized, value flows to whoever owns the customer and deploys AI cheaply. DoorDash, for example, is already cutting its AI bill with Kimi. Falling AI prices are a cost cut for every company that buys intelligence rather than sells it.

💡 So don't confuse "AI is getting cheaper" with "AI demand is dying", but make sure you know which layer each of your holdings lives in. If you own the chipmakers, you're betting usage grows faster than efficiency. If you own the hyperscalers, listen closely to the capex discussion on upcoming earnings over the next few weeks.

🇨🇳 As for me, this is also another reason I don’t like to keep all my eggs in the US basket and keep ~5% of my portfolio in $CHQQ, which tracks the Hang Seng TECH Index (holding companies like Alibaba, Tencent, and Meituan, which all have stakes in Moonshot).

📊 That said, China is very volatile, and so far my bet hasn’t paid off all that well (I’m down ~18%), but I’m in it for the long-term!

🗞️ Ok, let’s move on to our final big story of the week, IBM’s shocking 25% crash! But first, a quick word from our other sponsor this week, Range Finance!

EARNINGS
🩸 IBM Crashes 25% After “Pre-Announced” Earnings Lead to Its Worst Trading Day Ever

📄 On Tuesday, IBM ($IBM) pre-announced its Q2 earnings a full week ahead of its scheduled July 22 report, and the results weren’t good…

📊 By the numbers:

• ❌ Revenue hit $17.2 billion vs. $17.86 billion expected

• ❌ Adjusted EPS came in at $2.93 vs. $3.01 expected

• ❌ Infrastructure revenue fell 7%, Software revenue grew just 5% (well short of the double-digit target), and Consulting revenue was flat

📉 Shares fell a massive 25% on the news, wiping out roughly ~$70 billion in market cap in one session, and setting a record for the company’s single worst trading day in its entire 115-year history as a public company, even worse than Black Monday in October 1987.

🧠 More Memory, Less IBM

🤔 So why did a ~4% revenue miss trigger a 25% crash? Because of why IBM missed. IBM reported that clients had dramatically reprioritized their tech budgets in the final weeks of June, pulling money from software and services contracts to stockpile memory chips before prices rise further.

🎙️ From IBM CEO Arvind Krishna:

🔍 What Analysts Are Saying

✨ Many analysts view the selloff as an overreaction, saying that “if clients are simply front-loading purchases to beat memory price hikes, then demand will eventually come back to IBM once the shortage eases,” with IBM sitting at a moderate buy with a price target ~40% above its current price on TipRanks.

⚠️ But not all analysts agree, with BofA Securities’ Wamsi Mohan trimming his price target, saying he was “surprised by the magnitude of the topline miss,” and even cut IBM’s full-year revenue estimate to $69.6 billion (from $71.4 billion).

🔻 HSBC went further, downgrading IBM to “Reduce” with a $191 target, the lowest on the market, and roughly 11% below where the stock traded even after the crash.

😰 BNP Paribas' Stefan Slowinski (one of the few with an outright Underperform) cautions that "of all the companies I cover, IBM probably has the lowest organic growth currently and the lowest organic growth outlook."

👀 One risk is, of course, that this shift in spending lasts longer than a single quarter. But perhaps an even bigger risk is AI dissolving the moat that makes it hard to migrate off the IBM systems.

💡 And whether you hold IBM or not, this is a clear sign that this earnings season will be a lot less forgiving, and as O'Regan put it, "the days of getting the benefit of the doubt are probably over for now", for IBM and for any richly valued name that stumbles.

FROM THE BLOSSOM COMMUNITY
⭐️ Featured Posts of the Week

👇 Click to see the full post!