ASML Bets Big on AI Chips While Cutting 1,700 Jobs

The world's most critical chipmaker—the one that makes the machines that make the chips—just delivered a mixed message. ASML, which controls 90% of the extreme ultraviolet lithography market, says AI semiconductor demand will grow over 20% annually for years to come. At the same time, it's cutting 1,700 jobs to make its engineering teams more "agile."

This isn't just another earnings report. When ASML speaks, the entire semiconductor industry listens. Their EUV machines are the only way to manufacture the most advanced chips powering everything from ChatGPT to autonomous vehicles.

The AI Chip Gold Rush is Just Beginning

ASML's forecast extends far beyond typical corporate optimism. The company explicitly states that advanced logic and memory chips for AI computing will remain the "most important driver" in semiconductors until 2030. That's not a quarterly projection—it's a decade-long bet.

The math makes sense. Every major tech company is racing to build larger AI models, and each generation requires exponentially more computing power. OpenAI's GPT-4 reportedly used 25,000 GPUs for training. Future models could need 100,000 or more.

But here's what's interesting: ASML specifically mentioned both logic and memory chips. While Nvidia's GPUs grab headlines, the real bottleneck is often memory—specifically high-bandwidth memory (HBM) that can feed data to AI processors fast enough. Companies like SK Hynix and Samsung are scrambling to meet this demand.

Why Cut Jobs During a Boom?

The 1,700 layoffs seem counterintuitive during a growth phase. But ASML's reasoning reveals something crucial about the AI era: speed trumps scale.

Traditional semiconductor development cycles span years. You design a chip, test it, manufacture it, and hope the market still wants it when you're done. AI chips operate on a different timeline. Nvidia releases new architectures every two years, and AI software evolves even faster.

ASML is betting that smaller, more agile engineering teams can respond to this accelerated pace better than large, hierarchical organizations. It's a page from the Silicon Valley playbook: move fast and iterate.

The Memory Wars Heat Up

While Nvidia dominates AI processors, the memory game is wide open. SK Hynix currently leads HBM production, but Samsung and Micron are investing billions to catch up. Micron just announced a $24 billion memory plant in Singapore—a clear signal that memory makers see the same growth ASML is predicting.

The stakes are enormous. HBM prices are roughly 10 times higher than standard memory chips, and demand is outstripping supply. If ASML's 20%+ growth prediction holds, the memory market could see a fundamental shift in profitability.

The Geopolitical Wild Card

There's one factor ASML's forecast doesn't fully address: geopolitics. The company can't sell its most advanced EUV machines to China due to export restrictions. Meanwhile, China is investing heavily in domestic alternatives and building its own AI infrastructure.

This creates an interesting dynamic. If Chinese companies develop competitive AI chips using older technology, it could fragment the market ASML is betting on. Alternatively, if restrictions tighten further, it could concentrate even more demand among allied nations.