02-07 Gathering of Heroes: Career Paths and Industry Insights from Tech Professionals

The Randomness of Finance Careers

Careers in finance are full of chance. One friend shared a dramatic case: a woman who used to be a car show model ended up as a fixed-income manager at a bank. Her path went like this: from car show model to bond sales, and because she was attractive and had a mentor who noticed her, she got into the bond trading department of a bank and eventually became a fixed-income manager. A career path like this shows just how much opportunity matters in finance.

Thinking from Traditional Trading to Quantitative Investing

When the conversation turned to investing, someone offered a sharp insight: money will become less and less valuable over time. That judgment reflects a deep understanding of the modern financial system. As for investment strategy, one person said they're not really a trader. They make decisions mostly based on understanding a project's performance.

Notably, even private equity funds managing tens of billions of yuan have been losing ground to quantitative trading in recent years. In 2024, for example, some fund manager friends, such as the boss of Black Wing Asset, who manages 20 billion yuan, started using large language models for quantitative trading. Their interns even published a paper in a top journal, applying AI models to quantitative strategies, and the returns far exceeded traditional discretionary trading.

The Truly Scarce Resources: Compute and Energy

From this angle, human traders are no longer the core competitive advantage. What's truly valuable? Compute, energy, and electricity. In the future, money may not even be the scarcest resource.

Based on this thinking, as early as around 2020, some people began investing in compute-related assets such as GPUs and HBM. This investment logic reflects an important methodology: keep reasoning forward, and work out what is truly scarce and what only looks scarce.

Investment Opportunities from a US-China Competition Perspective

Looking back, the investment logic at the time was somewhat naive, but it did catch some key trends. After Trump was elected in 2018, the US-China trade war escalated, and China's semiconductor industry faced enormous pressure. The thinking at the time was from China's perspective: the core of the US-China rivalry has shifted to AI competition, and the US is far ahead in crypto financial systems and AI.

For China to win, it might need to take a "brute-force" route: building data infrastructure. Another idea was that, facing the contradiction of a shrinking population but rising productivity, one could build a "double-deck elevated highway" of a digital world alongside the physical world, using virtual worlds to solve the problem of excess productivity and insufficient labor demand.

Musk's Grand Loop

By comparison, Musk came up with a much grander solution: Mars. Instead of creating a virtual demand, he opened up an entirely new scenario of real demand. Musk's Optimus V3 robot is said to be astonishing, and people in Silicon Valley who've seen it say Tesla stock is worth buying. More importantly, V3 uses only silicon carbide devices.

Musk's whole loop of thinking goes like this: send robots to Mars to transform the planet. For compute, deploy it in space, where there's neither a heat dissipation problem nor light pollution. For energy, solve it with space solar power. Finally, use robots to terraform Mars. This loop is far more macro than the purely digital-world vision. At the time it was hard to see that far, and only now are we slowly beginning to understand it.

The Underlying Logic of Infrastructure Investment

The investment logic at the time was: to counter America's AI advantage, China could only take the infrastructure route. Infrastructure means generating electricity. With electricity you get compute, with compute you can counter the models, and then you rely on China's huge data advantage.

Based on this logic, they took a server apart completely and found that GPUs and storage were the two largest components, followed by communication connection chips and various interface chips. So they bought all of these semiconductor targets. Then they added the energy side: photovoltaics, lithium batteries, energy storage including liquid energy storage, all configured. The judgment at the time may not have been fully mature, but these investments have performed quite well so far.

Investment Discussion

Over dinner, everyone enjoyed their food while discussing recent investments. Someone mentioned that the market had dropped sharply a couple of days earlier, causing considerable losses, and that they were now trying to buy the dip. The conversation turned to a specific investment: TowerJazz.

The Investment Value of TowerJazz

Why pay attention to TowerJazz? Because it has scale advantages. Currently, the photomask process for all CPUs needs silicon photonics. Previously it used EML, but now it has to use silicon photonics. And all silicon photonics chips, no matter which design vendor, have to be fabricated at TowerJazz, because it's the only one with that capability. This is a classic case of a monopoly advantage in a "chokepoint" technology.

When someone asked about the code, the speaker said it wasn't convenient to share publicly, because it was already known to too many people. But the logic itself is worth watching, so everyone can do some homework first. The whole dinner had a relaxed atmosphere while people exchanged these deep investment insights.

举