Chip race: Microsoft, Meta, Google, and Nvidia battle it out for AI chip supremacy

After decades of only licensing its chip designs for others to use, UK-based Arm revealed the first chip it’s producing on its own, and the first customer. Dubbed the Arm AGI CPU, it’s another chip designed for inference, or running the cloud processing for AI tools like AI agents that can continue to spawn more and more tasks to run at once. The first company in line to use it is Meta, which has reportedly struggled to launch its own AI chips.

Meta says it’s both the lead partner and co-developer, and plans to work on “multiple generations” of the data center CPUs, for use along with hardware from other vendors like Nvidia and AMD. Arm customers like Amazon AWS, Microsoft, Google, Marvell, Nvidia, Samsung, and others included congratulatory notes with the announcement. However, Qualcomm, which said it had achieved “complete victory” over Arm with a court ruling last fall in their case over the terms of licensing agreements, was not one of them.

Nvidia announced on Monday that it’s investing $2 billion each into Lumentum and Coherent, which are both developing photonics technology for data centers, like optical transceivers, circuit switches, and lasers, which are used to move data at high speeds over long distances. Their tech could improve energy efficiency, data transfer speeds, and bandwidth in future AI data centers, after Nvidia already capitalized on its 2020 acquisition of the network hardware company Mellanox to beef up NVLink and increase the amount of data moving between its GPUs.

For Lumentum, the nonexclusive multiyear deal includes a “multibillion purchase commitment and future capacity access rights for advanced laser components,” as well as support for expanding R&D and manufacturing. Coherent’s deal is described similarly, with a “multibillion-dollar purchase commitment and future access and capacity rights for advanced laser and optical networking products.”

OpenAI has closed another round of funding, totalling $110 billion being newly committed to the maker of ChatGPT, which it says has more than 900 million weekly active users and over 50 million consumer subscribers. Amazon is investing $50 billion and striking a deal that includes plans for custom models and more. Nvidia and SoftBank are each contributing $30 billion, as well, even as the Wall Street Journal notes that Nvidia’s previous $100 billion investment plan is “on ice.” This marks another massive influx of cash for the company that’s now valued at $730 billion, and previously closed a $40 billion round in 2025. At the time, it was the largest private tech deal on record.

The investment from Amazon is more than just an injection of cash. The companies are entering a partnership that will potentially allow Amazon to play catch-up in the AI market. The two companies will be collaborating on custom models intended to power “customer-facing applications” like Alexa.

The AI data center build-out, as it currently stands, is dependent on two things: Nvidia chips and borrowed money. Perhaps it was inevitable that people would begin using Nvidia chips to borrow money. As the craze has gone on, I have begun to worry about the weaknesses of the AI data center boom; looking deeper into the financial part of this world, I have not been reassured.

Nvidia has plowed plenty of money into the AI space, with more than 70 investments in AI companies just this year, according to PitchBook data. Among the billions it’s splashed out, there’s one important category: neoclouds, as exemplified by CoreWeave, the publicly traded, debt-laden company premised on the bet that we will continue building data centers forever. CoreWeave and its ilk have turned around and taken out debt to buy Nvidia chips to put in their data centers, putting up the chips themselves as loan collateral — and in the process effectively turning $1 in Nvidia investment into $5 in Nvidia purchases. This is great for Nvidia. I’m not convinced it’s great for anyone else.

AMD has sealed a $1 billion deal with the US Department of Energy to develop two supercomputers, Lux and Discovery, in collaboration with Oracle and Hewlett Packard Enterprise (HPE). Both supercomputers will live at Oak Ridge National Laboratory (ORNL) in Oak Ridge, Tennessee. Lux is slated to come online fairly soon in early 2026, with Discovery following in 2029.

Both build on the work that went into the Frontier supercomputer, which is also housed at ORNL and was the fastest in the world until El Capitan came online last year at Lawrence Livermore National Laboratory. AMD also helped develop those supercomputers, so this isn’t its first time working with the US government on a project like this.

Qualcomm is launching a pair of new AI chips in an attempt to challenge Nvidia’s dominance in the market. On Monday, Qualcomm announced plans to release its new AI200 chip next year, followed by the AI250 in 2027 — both of which are built on the company’s mobile neural processing technology.

The new chips are built for deploying AI models, rather than training them. The launch marks a notable change for Qualcomm, which has primarily made processors for mobile phones, laptops, tablets, and telecommunications equipment.

With Windows 10 on its last legs, Intel is looking forward to the PC industry growing more than it has in years — the most since 2021, when the covid-19 pandemic revived industry growth by creating a huge surge in demand. But it seems the struggling Intel, which just received lifelines from Nvidia, Softbank, and the US government, isn’t fully ready to take advantage and is prioritizing AI instead.

Today on the company’s Q3 2025 earnings call, where Intel saw its first profit in nearly two years due primarily to those lifelines, CEO Lip-Bu Tan and CFO David Zinsner explained how the company doesn’t yet have enough chips. It’s currently seeing shortages that it expects to peak in the first quarter of next year — in the meantime, leaders say they’re going to prioritize AI server chips over some consumer processors as it deals with supply and demand.

OpenAI is teaming up with Nvidia via a “strategic partnership” that will get the ChatGPT-maker more compute and more cash to develop new models on the road to superintelligence.

The partnership, announced Monday, will allow OpenAI to “build and deploy at least 10 gigawatts of AI datacenters with NVIDIA systems,” which translates to millions of GPUs that can help power OpenAI’s new models. One of the most important points here, besides more data centers and compute — which are always in high demand for companies like OpenAI — is that as part of the deal, NVIDIA “intends to invest up to $100 billion in OpenAI progressively as each gigawatt is deployed,” per the release. The details will be finalized in the next few weeks, according to the companies.

Alibaba, ByteDance, and other Chinese technology companies are barred from purchasing Nvidia’s latest AI chips custom-made for China, the Financial Times reported on Wednesday. The Cyberspace Administration of China banned buying and testing the RTX Pro 6000D chips, despite companies having ordered thousands of the chips in the months since their introduction in July, according to three unnamed people with knowledge of the matter and who spoke to the FT.

Nvidia CEO Jensen Huang said he was “disappointed” by news of the ban during a press conference with reporters in London, where he is set to join President Donald Trump at a state dinner on Wednesday night. In August, Nvidia worked with Trump to broker a deal allowing the sale of H20 chips to China in return for a 15 percent cut of the profit.

President Donald Trump has called for Lip-Bu Tan to immediately resign as Intel’s CEO over his reported ties to Chinese tech firms. The demand follows Arkansas Senator Tom Cotton questioning Intel’s board chairman whether Tan’s alleged connections to China would conflict with security regulations.

“The CEO of INTEL is highly CONFLICTED and must resign, immediately,” Trump posted on Truth Social. “There is no other solution to this problem. Thank you for your attention to this problem!”

Nvidia has revealed its new DGX Spark and DGX Station “personal AI supercomputers” at today’s GTC conference, which are both powered by the company’s Grace Blackwell platform and designed for users to work on large AI models with or without a connection to a datacenter. The Spark is going up for preorder today.

The Spark is the new name for Nvidia’s $3,000, Mac Mini-sized “world’s smallest AI supercomputer” that was announced with the name “Digits” at CES earlier this year. Its larger Station counterpart, currently with no price tag, is aimed at “AI developers, researchers, data scientists and students to prototype, fine-tune and inference large models on desktops.”

Nvidia now makes $2,300 in profit every second on the back of the AI revolution. Its data center business is so gigantic, even its networking hardware now rakes in more money than its gaming GPUs. Now, the company is announcing the AI GPUs that it hopes will extend its commanding lead: the Blackwell Ultra GB300, which will ship in the second half of this year, the Vera Rubin for the second half of next year, and the Rubin Ultra that will arrive in the second half of 2027.

This year’s Blackwell Ultra isn’t what we originally expected when Nvidia said last year that it would begin producing new AI chips on a yearly cadence, faster than ever before, as Blackwell Ultra is not on a new architecture. But Nvidia quickly moved on from Blackwell Ultra during today’s GTC keynote to reveal that next architecture, Vera Rubin, whose full rack should offer 3.3x the performance of a comparable Blackwell Ultra one.