🚀 SK Hynix $26B IPO, Meta's Comeback Arc & China's Reusable Rocket Win

💎 SK Hynix Makes History with Record-Breaking NASDAQ Debut

SK Hynix completed the largest-ever first-time US share sale by a foreign company, raising $26.5 billion in its NASDAQ debut. The South Korean memory chip giant opened at $170 per share, marking a 14% pop above the $149 offering price and delivering a market cap of approximately $1.03 trillion—another trillion-dollar company minted in the AI boom.

Demand for the offering was exceptional, with subscriptions reportedly exceeding available shares by more than sevenfold. While heavily oversubscribed IPOs often see larger first-day gains, at this scale and price point, a 14% jump represents a strong debut. The company is trading under the temporary ticker SKHYV before transitioning to SKHY on Monday, as shares settle on a when-issued basis—a tool used to ease foreign companies into American markets.

The blockbuster listing caps an extraordinary rally for memory manufacturers, driven by surging demand for high-bandwidth memory (HBM) used alongside advanced AI accelerators. SK Hynix's Korean-listed shares have surged over 600% in the past year as investors rushed to gain exposure to one of the AI supply chain's most critical bottlenecks. SK Hynix, Samsung, and US-based Micron dominate the global memory market, and American investors can now access SK Hynix directly via ADRs without navigating South Korean exchanges.

"Congratulations to the whole team over there. A massively successful IPO."

🤖 OpenAI Drops GPT 5.6 Soul & Soul Ultra—Speed, Style, and Controversy

OpenAI launched GPT 5.6 Soul and Soul Ultra, delivering significant upgrades in speed, reasoning, and design capabilities. Early adopters are sharing demos across the timeline, including attempts to recreate entire films like Interstellar using prompt-to-3D workflows. While impressive, the consensus is clear: merely replicating existing content won't move the needle—creativity and novel storytelling will define success in AI-generated media.

One standout demo showed Soul building a 3D cannon model in Blender at blistering speed—750 tokens per second on Soul Ultra in fast mode. The model's ability to scaffold projects, write code, and execute tasks via computer use (not direct API integration) is turning heads. Watching an AI operate a computer at superhuman speed feels like an early "AGI moment" for many users.

However, the rollout wasn't without friction. OpenAI unified the ChatGPT desktop app with Codex, retiring the native Mac app in favor of an Electron-based interface. Power users expressed frustration over the loss of native integrations and familiar workflows, prompting OpenAI CTO Tibo to acknowledge missteps:

"We spent the last 24 hours reading feedback, looking at usage patterns, and talking with many of you. We made it too easy to use the highest compute settings without making the impact on usage limits sufficiently clear... our launch framing was focused on work, and to some of our Codex fans, it made it feel like Codex was going away over time. Absolutely not our intention. We love Codex and it is here to stay."

OpenAI reset usage limits twice in one day to let users keep experimenting, and the team is working to clarify the distinction between ChatGPT Work and Codex. The long-term vision appears to be a single omnibox—one interface for all tasks—but the transition has highlighted the challenge of balancing power user workflows with mainstream accessibility.

📊 Meta's Comeback Narrative: Semi Analysis Says Zuck Can Catch Up

Meta stock surged over 6% following a detailed research note from Semi Analysis, which argued that Meta is "the only hyperscaler/neo-lab on track to be world-class" at all three pillars required to build a frontier AI model: data, talent, and compute.

The report emphasized Meta's underappreciated advantage in human data and reinforcement learning (RL) environments. According to Semi Analysis, Meta reassigned 3,000 engineers—including 70% of new grads and a significant number of senior staff—to a new Applied AI Engineering org focused full-time on creating RL tasks and environments. This move effectively gave Meta an in-house version of Mercor, the RL data startup valued at over $1 billion ARR, for free.

"Meta is already in the same ballpark as Mercor's 2.5 million expert hours logged in Q2 2026, and their average quality is likely higher. Additionally, they have another 70,000 people to pull into that organization if needed."

Semi Analysis also dispelled the myth that 3,000 Meta engineers would be relegated to "mindless low-level data labeling." The report noted that creating high-quality training data has become a real intellectual challenge, with top expert contractors at leading data companies earning seven figures annually—a far cry from the $10/hour labeling tasks of the past.

The bull case hinges on Meta's structural willingness to make aggressive moves that competitors like Google may hesitate to execute. Zuckerberg's motivation is clear: he refuses to be at the mercy of platform owners ever again. Meta has spent billions paying taxes to Apple—both literal (via in-app purchases, which account for 2% of ad revenue and carry a 30% fee) and figurative (via iOS privacy changes that cost Meta billions in ad tracking capabilities).

"Zuckerberg has never held a job in his entire life. Being Meta's CEO is all he's ever known since he was 18 years old. He has nothing to lose. He's the exact type of person you should never bet against."

The market appears to be warming to the narrative. As one observer put it, Meta is shaping up to be "one of those setups where it's going to rip so hard that when it's done, everyone acts like it was obvious the whole time and pretends they were buying the whole way down. We can call this the Google at $150 phenomenon."

🚨 FIA Gate: Phoebe Gates's Shopping App Accused of Affiliate Fraud

Bloomberg published an explosive investigation into FIA, the buzzy shopping app co-founded by Phoebe Gates, daughter of Bill Gates. The report alleges that FIA's browser extension is claiming credit for online sales it didn't actually drive—a violation of affiliate marketing policies that could constitute unlawful fraud.

Affiliate marketing is straightforward: a publisher (like a product review site) refers a customer to a retailer, and if the customer completes a purchase, the publisher earns a commission. The cardinal rule is that commission is only paid if a user clicks. FIA is accused of violating this rule by opening background tabs and injecting its own referral code without user interaction, effectively stealing attribution from legitimate publishers.

Bloomberg tested FIA's mobile browser extension across more than 50 websites and found that during checkout, FIA opened a background tab and overrode legitimate referrals. These findings were consistent with testing by Capital One Shopping and independent researcher Ben Edelman, who has spent decades exposing deceptive practices in digital advertising.

"The most fundamental requirement in affiliate marketing is that commission is only paid if a user clicks. The rules don't allow fake clicks, simulated clicks, imaginary clicks, or hypothetical clicks. Only a real click will do."

Capital One Shopping sent a Tuesday email to retailers raising concerns about what it described as FIA's "fake clicks" or "cookie stuffing," attaching videos showing FIA's extension silently opening background tabs to ensure cookies were set. The email stated:

"Publishers like us are having material revenue taken, and advertisers like you are losing money to fake clicks."

FIA's alleged behavior is potentially worse than Honey's, the plugin that sparked a viral scandal in 2024. Honey at least required users to interact with a prompt—even if only to dismiss it—before updating affiliate tracking. FIA reportedly fires without any user interaction at all, silently redirecting revenue away from the publishers who drove the customer to the retailer in the first place.

Bloomberg's report stops short of alleging criminal conduct, noting that civil theories such as deceptive practices, unjust enrichment, interference with contracts, and unauthorized alteration of tracking data are "quite plausible." Deliberate forced click cookie stuffing can even become wire fraud, though proving intent would require additional facts not yet available.

The story has ignited intense speculation online, though observers caution against assuming intent before more facts emerge. It's entirely possible the behavior stems from a bug rather than deliberate design. Regardless, it's a brutal day on the internet for the FIA team—and a stark reminder of how quickly reputation can collapse in the affiliate marketing space.

🛰️ China Launches Reusable Rocket, Catches Booster in Giant Net

China successfully launched and partially recovered a reusable rocket, marking a milestone for Beijing's space program as it competes with the US and SpaceX. The Long March 10B was launched into orbit from southern Hainan province, and minutes later, a giant net mounted on a seaborne platform caught the rocket's booster upon descent.

The netted recovery represents the first net-based rocket launch recovery in history—a novel approach compared to SpaceX's precision landing systems. The Long March 10B stands approximately 200 feet tall with a payload capacity of more than 16 metric tons. The recovery net was mounted on a seaborne platform, and the company expects to reuse the recovered booster by the end of the year. Two previous attempts in December failed.

"It would be a big deal if China's program can regularly recover and reuse parts of its rockets. It pretty radically decreases the cost of building out these giant satellite constellations that China has made very clear that it wants to build out."

The successful launch is part of China's broader push to accelerate its space access capabilities, which have both economic and military applications. The Long March 10B's netted recovery raises questions about scalability and thermal management—observers wonder how the net withstands the heat of reentry without melting, suggesting advanced materials like titanium alloys may be in play.

💰 Costco's Quiet Millionaires: $5.85/Hour to Seven Figures

The Wall Street Journal profiled Tony Barzar, a Tucson Costco cashier who started gathering shopping carts at Price Club in 1968 for $5.85 per hour. After Price Club merged with Costco in 1993, Barzar stayed with the company, regularly contributing a small portion of his paycheck to his 401(k). Today, he has accumulated more than $1 million.

Barzar's story reflects Costco's long-standing philosophy: pay workers more than most retailers, reduce turnover, lower training costs, and improve customer service. The strategy has worked. Costco's one-year employee turnover sits at approximately 7%—far below typical retail levels, which range from 40% to 70% annually, according to Mercer's 2025 survey.

Costco's business has boomed alongside this approach. Annual sales have grown for nearly two decades, and the stock has risen more than 2,000% since the depths of the 2008 financial crisis. The company's founders believed higher wages and benefits would pay for themselves through operational efficiency, and the data confirms it.

Barzar's story also highlights the critical role of Costco's benefits during personal crises. When his wife of 26 years was diagnosed with stage three brain cancer, Costco's health insurance covered the full cost of her three brain surgeries. He took nearly a year of paid leave to care for his family and used the company's therapy benefits to cope. This year, he returned to work part-time, citing pay and benefits as major reasons he stayed with Costco for nearly four decades.

"The pay and benefits are a major reason why he stayed with Costco for nearly four decades."

The value of tenure extends beyond financial security. A tenured cashier can make an entire store run more efficiently, learning subtle tricks and optimizations that might seem trivial but compound across thousands of transactions. Costco's bet on retention has clearly paid dividends—literally and figuratively.

🎮 Love Island Drives Prediction Market Surge Among Women

Prediction markets are experiencing an unexpected demographic shift: Love Island, the reality dating competition where singles live together in a villa and compete for a cash prize, is driving a surge of women onto platforms like Kalshi. According to Barron's, the current season drove a 106% increase in weekly active female mobile users between June 8 and June 28, compared with a 54% increase among men.

Love Island markets have generated approximately $52 million in trades and have three times as many female traders as male traders. The show's format makes it almost perfectly designed for prediction markets: episodes air six days a week and are released close to real time, creating a constant stream of new information and betting opportunities.

The phenomenon is spreading beyond digital platforms. Bars have started treating Love Island like a live sporting event. Tom's Watch Bar in Sacramento reportedly generated $30,000 in a single night during a watch party—an otherwise slow sports season.

"I need a market on how long until there's an insider trading scandal on Love Island... production staff has their phones. It's just the contestants that don't."

The risk of insider trading looms large. While contestants are isolated from the outside world, production staff retain access to phones, creating potential avenues for information leakage. As prediction markets grow in scale and sophistication, Love Island may become the test case for how platforms handle insider trading in entertainment.

The Love Island USA finale airs Sunday, July 12th. We'll see if Kalshi can maintain momentum—and whether the prediction market industry can handle the inevitable regulatory scrutiny that comes with mainstream adoption.

🎬 Hollywood's Horror Meme Gold Rush: Siren Head Sells for $1M+

The Wall Street Journal reports that Trevor Henderson, a 40-year-old illustrator from Toronto, sold the movie rights to his internet-born monster Siren Head to Warner Brothers for more than $1 million. Henderson drew Siren Head eight years ago and watched it spawn YouTube films, knockoff Amazon merchandise, and video games—but never made a penny until now.

"I wasn't seeing any financial benefit from Siren Head for years, and I just made peace with that idea. So the last few weeks have been disorienting."

The deal is part of a broader Hollywood gold rush to find concepts and talent online that could fuel the next horror hit like Backrooms (which grossed over $100 million). Studios are scouring YouTube, Reddit, and even Roblox for IP that has already proven its appeal to Gen Z.

The race for hot digital properties resembles the kind of dogfights for buzzy scripts that were common in Hollywood during the 1980s and 1990s. 11 studios recently bid on the film rights to the psychological horror YouTube series The Mandela Catalog. Amazon-owned United Artists and Steven Spielberg's Amblin Entertainment ultimately agreed to pay millions of dollars and let the series' 22-year-old creator, Alex Kister, direct the movie.

Another filmmaker, Aaron Coons, raised a Kickstarter budget for Shelby Oaks two years ago because no one in Hollywood was interested. When he prepared to sell Mandela earlier this year, he decided to wait until Backrooms came out. As the bidding war escalated, Coons and his partner started eliminating studios that didn't seem familiar with the material, saying:

"They basically went into ChatGPT and said, 'What's the next Backrooms?'"

Numerous agents and producers are now preparing pitches based on internet horror concepts, planning to sell them to studios in the coming weeks. The meme-to-movie pipeline is officially open for business.

⚡ Odds & Ends: Sequoia's Million-Fold Gain, Sonos Layoffs & More

Nvidia's Origin Story: In June 1993, Sequoia Capital and Sutter Hill Ventures wired $2 million to a three-person chip startup called Nvidia. The round valued the company at roughly $6 million. Don Valentine famously told Jensen Huang that if he lost the money, he would kill him. Jensen took it to heart. Today, Nvidia is worth approximately $5 trillion—a million-fold gain on Sequoia's investment.

Sonos Restructures: Sonos pushed out several design and product leaders as part of recent job cuts, including a VP of design who spent 12 years at the company and a top UX executive with a decade of tenure. CEO Tom Conrad framed the cuts as a way to reduce management layers and accelerate decision-making. The company reported 8% year-over-year revenue growth in the fiscal quarter ending last March, stabilizing after a disastrous app rollout battered customer satisfaction.

Yan LeCun Raises AI Fund: Meta's former Chief AI Scientist is raising a $1 billion VC fund focused on AI, following a $1 billion raise earlier this year for his startup, AMI Labs. LeCun left Meta to build and invest in AI, continuing his decades-long role as one of the field's most prominent voices.

Kim Jong-un's Renovation Spree: Satellite imagery reveals that North Korea's leader has begun renovating at least nine luxury compounds simultaneously, flaunting spending power amid an economic revival. One villa features ongoing roof work at a private beachfront complex where his $7 million yacht typically docks. The regime spends up to $1.8 billion annually on perks like cars, luxury goods, and medical services for elites, despite nearly half the population being malnourished.

Record Companies Push for AI Labels: A coalition of groups representing music labels and artists is pushing to label AI-generated songs on streaming platforms, citing fan demand for transparency. The challenge lies in defining where the line is—acoustic recordings are clearly not AI, but what about heavily autotuned tracks processed with generative filters? The debate is heating up as artists and producers navigate the blurry boundary between production tools and AI synthesis.

🎓 The Great Midterm Heist: When AI Cheating Meets In-Person Finals

A Brown University professor gave students a take-home midterm exam. Suspecting widespread AI cheating, he made the final in-person. The results were damning: orange dots (midterm scores) clustered near the top, while gray dots (final scores) plummeted for many students. One student scored 100% on the midterm and 0% on the final—the most brazen example of AI-assisted cheating imaginable.

Only one student at the bottom of the distribution improved between exams, moving from a 55 on the midterm to a 59 on the final. That student—clearly not cheating—earned respect for grinding through the material and actually learning something. The chart serves as a stark reminder of the challenges educators face in an era where AI can effortlessly generate A+ work, but students still need to pass in-person exams.

🔮 Final Thoughts

This week delivered a mix of breakthroughs, scandals, and curiosities. SK Hynix's record-breaking IPO underscores the insatiable demand for AI infrastructure. Meta's potential comeback arc—fueled by a data moat and structural aggression—could reshape the frontier AI race. FIA's affiliate fraud allegations serve as a cautionary tale about the perils of growth-at-all-costs tactics. And Love Island's prediction market surge proves that culture, not just politics, can drive adoption of decentralized platforms.

As always, the markets reward preparation, execution, and the willingness to lean into chaos. Whether it's Zuckerberg betting billions on AI, China catching rockets in nets, or filmmakers turning internet memes into million-dollar deals, the throughline is clear: bold moves win.

See you next week. 🚀