TSMC Doubles Down on America
The day belonged to Taiwan Semiconductor Manufacturing Company. The world's largest contract chipmaker is pouring another $100 billion into its Arizona operations, pushing its total U.S. investment to $265 billion. CFO Wendell Huang confirmed the company is speeding up construction at its Arizona factories to keep pace with surging demand for AI chips — the processors that sit inside a huge share of today's phones, computers, and AI systems.
The timing is no accident. TSMC just posted a record quarter of roughly $40.2 billion in revenue, and analysts are reading the result as a clear signal that the AI boom isn't slowing down. It wasn't alone: equipment maker ASML, another linchpin of the global chip supply chain, also notched record results. Together, the two underline just how thoroughly artificial intelligence is reshaping the semiconductor industry — one company building the chips, the other building the machines that make them possible.
The Buyers Are Writing Enormous Checks
If TSMC's numbers show the supply side booming, the demand side is matching it dollar for dollar.
Elon Musk's SpaceX is reportedly placing one of the largest AI hardware orders yet: a $52 billion deal, struck with manufacturer Foxconn, for 13,000 racks of Nvidia's GB300 AI servers. The rocket company doesn't just plan to use all that horsepower itself — it intends to rent the chips out. Worth noting: the figure comes with an asterisk, circulating through secondhand reporting rather than official confirmation, so treat the precise number with some caution.
Amazon, meanwhile, wants investors to see its AI spending as strategy, not speculation. The company is committing roughly $200 billion to artificial intelligence, and CEO Andy Jassy insists the bet isn't being made "on a hunch." His evidence: Amazon's homegrown AI chip is already a $20 billion business — proof that building silicon in-house can pay off at scale.
Nvidia Everywhere, and the Push to Loosen Its Grip
Nvidia's gravity was felt across the globe today. CEO Jensen Huang has been in Japan signing a wave of technology agreements that reportedly span the country's entire tech ecosystem rather than any single partner — a sign of how central Nvidia has become to national AI ambitions.
But dominance invites resistance. Alibaba has open-sourced its chip software as Chinese GPU makers work to weaken the grip of Nvidia's CUDA platform, the software layer that keeps developers locked into Nvidia hardware. The move follows similar plays by Huawei, and it points to a slow, deliberate effort inside China to build an alternative software stack that doesn't depend on the market leader.
On the Ground: Faster Laptops, Pricier Memory
Away from the mega-deals, two developments hit closer to everyday buyers. A leaked benchmark suggests AMD's next-generation laptop processor — a 10-core mobile chip from the "Medusa Point" family — has set a new x86 mobile speed record, hinting at a standout performer in the next round of laptops.
The less welcome news is in memory. One of the industry's most powerful figures, the chair of SK Group, has publicly called RAM prices "abnormally high" — an unusually candid admission that confirms what buyers already suspected: the cost of computer memory has gotten out of hand.
The Bottom Line
The throughline is unmistakable. Record revenues at TSMC and ASML, record-sized orders from SpaceX and Amazon, and Nvidia stitching together deals across Japan all point the same direction — the AI chip cycle is still accelerating, not cooling. The open questions are about the edges: whether China's open-source counterattack can chip away at CUDA, and whether soaring memory prices become the first real friction in an otherwise runaway boom.