American investors have done something in September that they rarely manage twice in one month: they’ve minted another of the trillion-dollar chipmakers. Meanwhile, the firm that would most like to join them can’t get enough of its own chips to its own customers.
Key takeaways
- AMD crossed $1 trillion in market value on September 21, joining the trillion-dollar chipmakers alongside Nvidia, Broadcom and Micron among US chipmakers.
- Nvidia, at roughly $5.3 trillion, is larger than the other three US members combined, several times over.
- Huawei says it lacks the capacity to meet even Chinese demand, and has pulled its next Ascend chip forward to Q1 2027.
- US and Chinese chipmakers face two different scarcities, and the AI supply chain is splitting because of it.
How AMD Joined the Trillion-Dollar Chipmakers
AMD crossed the threshold on September 21, with shares jumping nearly 10% in a single session. The push came from 107% year-on-year growth in its data-centre business, and from a chief financial officer willing to put a number on the industry’s ambition: a $3 trillion total addressable market for AI by 2030.
AMD now sits alongside Nvidia, Broadcom and Micron as domestic trillion-dollar chipmakers. Taiwan Semiconductor and Samsung hold similar territory abroad.
Nvidia still dwarfs everyone
The scale gap is striking. At roughly $5.3 trillion, Nvidia is not just the biggest of the trillion-dollar chipmakers. It’s larger than the other three combined, several times over. That matters, because the more interesting geopolitical story this month isn’t about the club’s newest and smallest member. It’s about the company still standing outside.
Huawei’s Uncomfortable Admission About Chip Capacity
On September 17, at Huawei Connect in Shanghai, rotating chairman Eric Xu said something chip executives rarely say willingly: Huawei can’t make enough of its own product. Because it doesn’t have the capacity to meet demand inside China, he said, there’s no plan to expand internationally in a full-fledged way.
Pulling the Ascend 960DT forward
Huawei’s answer wasn’t to slow down. It moved its next flagship chip, the Ascend 960DT, forward by three full quarters, to the first quarter of 2027. At the same time it quietly raised prices on the current 950DT by 20% to 50%, as scarce high-bandwidth memory pushed costs up.
What the bottleneck really says about export controls
You can read this as a plain supply-chain bottleneck. But it’s better understood as a statement about where China’s AI buildout stands against Washington’s export controls, which have tightened steadily since 2022 and cut Huawei off from the most advanced Western chipmaking equipment.
Unable to buy its way around the limits, Huawei is trying to out-engineer them. Its Atlas cluster architecture links nearly 15,500 chips using optical networking to reach a claimed 120 exaflops, a brute-force answer to a precision problem. Xu also claimed that Huawei’s domestic Ascend market share already exceeds Nvidia’s inside China, though the company offered no figures to back that up.
Two Different Scarcities Facing Trillion-Dollar Chipmakers and Huawei
The American trillion-dollar milestones are mostly about AI infrastructure spending that has been building for two years. The bigger point is that the world’s two largest chip ecosystems are now limited by very different things, and neither problem is quick to fix.
The US problem: capacity
American trillion-dollar chipmakers are held back by demand outrunning fabrication capacity, which takes years and tens of billions of dollars to build. It’s a scarcity that money can eventually solve.
The Huawei problem: a technology ceiling
Huawei’s constraint looks more like a ceiling. It can’t source memory components at the volumes or performance that Western suppliers can, under an export regime designed to keep that gap open. Money alone doesn’t close a gap like that. State-directed capital, redirected supply chains and time do.
Why the Split AI Supply Chain Matters Beyond Chip Investors
If your business relies on AI infrastructure, whether you’re buying cloud compute, negotiating with a chip-dependent vendor or budgeting for the next generation of AI tools, the takeaway isn’t “US chips are hot.”
It’s that the global AI supply chain is quietly splitting into two systems, with different hardware, different memory suppliers and different regulatory assumptions, and few procurement teams have priced that in. Planning around a single global chip market means planning around something that, as of this month, both your biggest supplier and your biggest potential rival are telling you no longer exists.
Frequently Asked Questions
1. Which trillion-dollar chipmakers are there?
In the US, Nvidia, Broadcom, Micron and, since September 21, AMD. Taiwan Semiconductor and Samsung hold similar valuations abroad.
2. Why did AMD’s stock jump on September 21?
Shares rose nearly 10% on 107% year-on-year growth in its data-centre business and a $3 trillion AI market forecast for 2030 from its CFO.
3. Why can’t Huawei meet chip demand?
Huawei says it lacks the capacity to satisfy even Chinese demand, with export controls limiting access to advanced equipment and high-bandwidth memory driving costs up.
4. When is the Huawei Ascend 960DT coming?
Huawei pulled the chip forward by three quarters, to the first quarter of 2027.
Explore more expert insights, leadership stories, and business strategies at GlobeVox Leaders
