Managing Elon

Key Person Risk

Elon Musk is launching a new political party called the "America Party." This is awkward, as it comes just days after a public feud with President Donald Trump, who threatened to cut off the government subsidies Musk's companies have built an empire on. For Tesla investors, who watched the stock fall nearly 8% on the news, this is not ideal. Their CEO now has a second, presumably less profitable, full-time job.

In a normal company, if the CEO announced he was starting a political party and feuding with the President of the United States, the board of directors would have a quiet word with him. And if he didn't stop, they would fire him. But Tesla is not a normal company. Here is Reuters on the board's unique challenge:

The company's shares and its future are seen as inextricably tied to Musk. He is Tesla's single largest shareholder, according to LSEG data.

But such a move [to remove him] is highly unlikely considering the board has often defended Musk.

"The Tesla board has been fairly supine; they have not, at least not in any demonstrable way, taken any action to force Musk to limit his outside ventures, and it's difficult to imagine they would begin now," [said Ann Lipton, a professor at the University of Colorado Law School].

The standard tools of corporate governance don't really apply here. The board’s job is to represent the shareholders and protect the value of the company. But what, exactly, is Tesla? Is it a collection of factories and employees and intellectual property? Or is it just Elon Musk?

The market has consistently voted with its money that it’s the latter. Tesla’s trillion-dollar valuation isn't a bet on car manufacturing margins; it’s a bet on Musk's visionary, world-changing genius. This works out well when the genius is focused on building robotaxis. It is less optimal when he is feuding with the person who controls the regulatory bodies that will approve those robotaxis.

This distraction now has a measurable cost. The stock is down 35% from its December high. An investment firm, Azoria Partners, pulled a planned Tesla-focused ETF, with its CEO saying they have "real concerns about Elon's ability to be a full-time CEO." Investors are openly asking whether his political ambitions are compatible with his duties to the company.

But the board is in a bind. Their fiduciary duty is to maximize shareholder value. Firing Elon Musk, the storyteller-in-chief, would almost certainly destroy more value than his political activities. And so they do... nothing. Their inaction isn't necessarily a failure of governance; it's a rational response to an impossible situation. They are trapped. Their job isn't to oversee a car company. It's to manage the emotional and political whims of one very important, very online man. Good luck putting a value on that.

Memory Is Everything

Samsung, a company that makes approximately a million different things, reported earnings this week, and its profit fell in half. This was surprising. The general vibe in the chip world has been pretty good lately, particularly for anything with "AI" stamped on it. Samsung, a global leader in semiconductors, should be doing great.

So what went wrong? It wasn't that people stopped buying its fancy folding phones or its high-end TVs. No, the problem was that Samsung, one of the most sophisticated electronics manufacturers on the planet, stumbled in the race to produce a very specific kind of memory chip that is, it turns out, absolutely essential for the AI revolution. And Bloomberg reported:

Samsung Electronics Co.’s profit fell for the first time since 2023, reflecting the deepening market share losses clouding the memory chipmaker’s prospects in the AI era. …

The disappointing results underscore how South Korea’s largest company has ceded leadership in the AI market to SK Hynix Inc. in the post-ChatGPT infrastructure boom. Its longstanding rival — along with Micron Technology Inc. — now sells more of the cutting-edge high-bandwidth memory chips paired with Nvidia Corp.’s AI accelerators.

This is a wild story. For decades, if you wanted to understand the technology landscape, you looked at CPUs and Moore’s Law. Then, for the last few years, all anyone could talk about were GPUs, the specialized processors that power artificial intelligence. But it turns out that GPUs are just one part of the equation. To do all the magical things that AI does, these powerful chips need to be fed a constant, fire-hose stream of data. The bottleneck is no longer just the processor; it's the memory.

And not just any memory. High-Bandwidth Memory, or HBM, is a special kind of memory where chips are stacked vertically, like tiny skyscrapers, creating an ultra-wide highway for data to move. This is crucial because a multi-billion dollar AI data center is only as good as its slowest component. If your expensive Nvidia GPUs are sitting around waiting for data, you are lighting money on fire. HBM is the solution.

This has turned the once-sleepy market for memory chips into a high-stakes, winner-take-all battleground. And for now, the winner is SK Hynix, which cleverly aligned its production with Nvidia’s roadmap and became the primary supplier of the specific HBM needed for the H100 and new Blackwell chips.

Samsung had long been the king of memory. But its current crisis is rooted in its failure to get its latest 12-layer HBM3E chips certified by Nvidia. This validation is the golden ticket to the AI boom. Since Nvidia controls the vast majority of the AI accelerator market, its certification effectively determines which memory gets used in the world’s most advanced data centers. Without Nvidia's stamp of approval, a cutting-edge HBM chip has no mass-market buyer. The delay forced Samsung into a massive inventory writedown on unsold AI chips. And while the company has secured some orders from Nvidia's rival, AMD, it’s a much smaller prize.

The power dynamic here is fascinating. The "Nvidia Effect" is now so total that it doesn’t just dictate the market for AI accelerators; it dictates the market for all the critical components that plug into them. Samsung's failure to get a single product generation right cost it billions in profit and its leadership position. It’s a brutal reminder that in the AI arms race, coming in second can be the same as not showing up at all.

The Scoreboard

• AI: Meta Hires Top Apple AI Expert, Continuing Zuckerberg’s Recruitment Push (WSJ)
• Cloud Infra: Coreweave to Acquire Core Scientific in $9 Billion All-Stock Deal (CNBC)
• Data Center: Japan, UK Firms Seek to Build ‘World’s First’ Floating Data Center (Nikkei Asia)
• E-commerce: Amazon Sellers Curb Prime Day Discounts With Tariffs Taking Bite (Bloomberg)
• Self-Driving: China’s Car Inc Launches Self-Driving Rental Service With Baidu’s Apollo (Reuters)