Key Highlights

● Alphabet introduced two eighth-generation Tensor Processing Units.

● The TPU 8tx is built for secure, confidential AI inference.

● AMD's EPYC server CPUs power the TPU 8tx host.

● Nvidia tumbled to $178.67, its lowest close since April 7.

Alphabet (NASDAQ:GOOGL) used Monday, November 9, 2026, to show off the next step in its custom silicon push. According to news reports, the company introduced two eighth-generation Tensor Processing Units: the TPU 8tx, built for secure, confidential AI inference, and the TPU 8i Max, designed to address the so-called "memory wall," the bottleneck created when processors wait on data from memory.

Google framed the launch around access. "Our aim is to make TPUs more easily accessible to every developer and Cloud customer," the company said in its announcement, according to market coverage. Market data quoted Google chip chief Amin Vahdat.

TPUs are Google's own chips for Training and running AI models, offered through Google Cloud. Every workload that runs on them is one that does not necessarily need a graphics processor from Nvidia.

Partners and rivals in the same breath

The announcement scrambled the usual lines between friend and foe. Advanced Micro Devices (NASDAQ:AMD) said in a press release that "fifth-generation AMD EPYC server CPUs are powering the host foundation for Google Cloud TPU 8tx," according to market reports. That link added to a big day for AMD, which soared after Oppenheimer upgraded the stock to Outperform with a $600 price target.

Nvidia (NASDAQ:NVDA), meanwhile, tumbled to $178.67, its lowest close since April 7 according to market data, amid an independent review by its board involving CEO Jensen Huang. Market data flagged the new TPUs as added competition for the market leader. Broadcom (NASDAQ:AVGO) jumped with the rest of the sector.

How Alphabet's own stock reacted

Alphabet shares climbed sharply in early trading, according to news reports, before giving back much of the gain. Class A shares closed modestly higher at $288.15, while the Class C shares edged up a little less, so the market's reaction to the chips themselves was measured rather than euphoric. The rally was broad: the PHLX Semiconductor index rocketed to 10,033.93, its best day since July 30, and the S&P 500 climbed 173.08 points to 7,549.94.

Alphabet's lift was more modest than the chip names. The company is still digesting a third quarter in which, according to market coverage, big negative surprises from Alphabet and Meta helped make communication services the only sector with slowing Earnings growth.

Why the race is heating up

The Economics are simple. AI data centers are some of the most expensive buildings on Earth, and chips are a large share of the bill. Cloud giants that design their own processors can tailor them to their workloads and reduce reliance on any single supplier. That does not end Demand for Nvidia's hardware, but it could give buyers more options.

The spending itself is under scrutiny. Goldman Sachs sees AI capital-expenditure growth peaking, and a research firm has warned that spending has outpaced operating cash flow, according to market reports. Chicago Fed President Austan Goolsbee said Monday the AI data-center build-out is "starting to look like old fashioned demand overheating."

Custom Silicon Raises the Stakes

Nvidia's November 17 earnings report is the obvious next marker, as investors listen for any comment on custom-chip competition. Before that, Applied Materials reports Thursday, giving a read on the equipment that builds every kind of AI chip. Watch, too, for how quickly Google Cloud customers adopt the TPU 8 family and whether other hyperscalers answer with chips of their own.