The world’s biggest chipmaker just admitted it can’t make chips fast enough. Taiwan Semiconductor Manufacturing Co. is struggling to meet demand from American customers, even with billions pouring into new US factories, according to The Verge AI, which cites reports from Reuters and Bloomberg. That’s a striking confession from the company that builds the silicon behind nearly every major AI chip on the market.
What CEO C.C. Wei actually said
Speaking after a shareholder meeting on Thursday, TSMC chief executive C.C. Wei put it plainly: “Customer demand is so high, and we can only support so much.” He added that the company is “doing our best to ensure TSMC does not become a bottleneck.”
That word, bottleneck, is the whole story. When the company that makes the chips becomes the constraint, everyone downstream feels it: Nvidia, Apple, AMD, and every startup trying to get compute for its models.
Why this matters
TSMC isn’t just another supplier. It manufactures the most advanced chips for Nvidia’s AI accelerators and Apple’s processors. When its CEO says demand outstrips what the company can physically produce, that points to a supply ceiling on the entire AI buildout.
What stands out here is the timing. The AI boom is already squeezing the memory market. The Verge AI notes that a widespread shortage of RAM and NAND Flash memory is expected to last for years. Now the logic-chip side is flashing the same warning. Two of the most critical components in any AI system are both constrained at once.
The money behind this is staggering. Citing research from Deloitte, The Verge AI reports the AI surge could push semiconductors toward a $1 trillion industry by 2027. Demand isn’t the problem. Capacity is.
The US factory math doesn’t add up fast enough
TSMC has leaned hard into American manufacturing. Here’s where things stand, per The Verge AI:
- A factory is already open in Arizona.
- The company plans to invest $165 billion to build three more US plants.
- That includes two advanced packaging facilities and a research and development center.
That’s a serious commitment. But Wei warned it could take a “very long time” to meet customer needs through US-based production. Building leading-edge fabs is slow, expensive work. You don’t spin up a chip plant in a quarter. You spend years on construction, equipment, and yield tuning before a single usable wafer ships at scale.
So even with the buildout, the gap between what customers want and what TSMC can deliver isn’t closing soon.
The pricing question
Wei said he’d “like” to raise TSMC’s prices, which is what you’d expect when demand runs this far ahead of supply. But he signaled the company won’t do an abrupt hike. The Verge AI contrasts that with the sharp jumps we’ve already seen in DRAM and SSD prices, where shortages translated into fast, painful cost increases.
For now, TSMC appears to be holding the line on sudden price shocks. Whether that holds as demand keeps climbing is another matter.
What to expect next
A few things follow from this, and they’re worth watching:
- Longer lead times. If TSMC is the constraint, anyone ordering advanced chips should plan for waits, not quick turnarounds.
- Upward price pressure. Wei wants to raise prices. Even a gradual climb feeds into the cost of every AI chip and, eventually, the cost of running AI.
- More squeeze on smaller players. When supply is tight, the biggest buyers get served first. Startups and smaller firms may find themselves at the back of the line.
- The memory and logic shortages compound. RAM, NAND, and now leading-edge logic are all tight. That’s a broad supply problem, not a single choke point.
The honest read is that the AI industry’s growth is now bumping against physical limits. You can raise billions and announce data centers, but someone still has to etch the silicon. Right now, the company that does that better than anyone is telling the world it can only do so much.
More details are available in the original report from The Verge AI.