Women Building the AI Era, Part 3- Renee James
While Nvidia became the face of the AI boom, Renee James spent seven years quietly building a company to unseat Intel from the AI data center's other chip, and SoftBank just paid $6.5 billion to bet she was right.

For twenty-eight years, Renee James's career was Intel's career. She started as an engineer, became the executive assistant to co-founder Andy Grove, and by 2013 was president of the company, one rung from the top job.
In 2015, she left. Two years later, she used money from The Carlyle Group to buy the leftover intellectual property from a struggling Arm chip startup nobody in the industry took seriously, and built a new company around it.
That company, Ampere Computing, spent the next seven years selling a simple, contrarian bet: that Intel's grip on the processor sitting inside nearly every cloud data center wasn't as unbreakable as Intel itself believed. In 2025, SoftBank paid $6.5 billion to agree with her.
The Chip Nobody Talks About
Ask someone what powers the AI boom and they'll say Nvidia, and they wouldn't be wrong. Nvidia's GPUs do the actual model training and inference that makes a chatbot answer a question, and the company's valuation reflects it, the backbone Nvidia never planned to become is now propping up much of the stock market.
But a GPU doesn't run alone. Every server rack in a data center also needs a general-purpose processor, a CPU, to manage the workload, move data around, and handle everything the GPU isn't built for. For decades, that CPU was almost always made by Intel, and cloud providers paid Intel's prices because there wasn't a serious alternative.
Think of the AI rack as a small kitchen during a dinner rush. The GPU is the chef, doing the one specialized task everyone showed up for. The CPU is the person running everything around that chef: taking orders, plating, keeping ingredients moving, making sure the chef never has to stop and do prep work. Nobody photographs the person running the kitchen, but dinner doesn't happen without them, and for twenty years, that job belonged to Intel almost by default.
Ampere's entire business is built on being an alternative to that default. Its chips use the Arm architecture, the same lower-power chip design family that runs most smartphones, scaled up to server size. Arm-based chips are generally built to do more work per watt of electricity than a traditional Intel-style design, which matters enormously when a cloud provider is running tens of thousands of these processors around the clock. Shaving even a modest amount off the power bill of every server in a data center adds up to real money at that scale, and it hands a customer leverage to negotiate a better price from Intel even if they never switch a single workload to Ampere.
Cloud giants including Oracle, Google Cloud, Microsoft Azure, and Tencent Cloud now run workloads on Ampere processors. One industry newsletter that tracks the chip market closely has described Ampere as the leading merchant supplier of Arm-based server chips sold to outside customers, though Ampere stayed privately held throughout this period and never disclosed audited revenue, so any specific market-share or revenue figure attached to it should be read as an outside estimate, not a confirmed company figure. Some of the same cloud giants also design similar Arm chips solely for their own internal use, a separate pool of hardware that never shows up in any estimate of Ampere's business but rests on the same underlying bet: that Intel's decades-old default doesn't have to stay the default.
Did You Know?
Ampere's founding technology didn't start as Ampere's. It began as "X-Gene," an Arm server chip built by a company called Applied Micro. When Applied Micro was sold off in 2017, a Carlyle-backed shell company bought just the X-Gene chip technology and the engineers who built it, then hired Renee James to turn it into a real company. Ampere is, in a sense, a chip design that outlived the company that invented it.
An Unusual Investor With a Direct Interest
Ampere's early backing came with a twist: Oracle, one of the biggest customers for server chips in the world, invested directly in Ampere rather than just buying its processors. Oracle first disclosed a stake worth about $40 million in 2019, well before Ampere had shipped its first chips at scale.
That arrangement gave Ampere something most chip startups never get: a guaranteed, high-volume customer with a financial reason to see the company succeed. It also meant Oracle had every incentive to keep buying Ampere chips even if a comparable Intel processor were slightly cheaper, since Oracle's own stake gained value every time Ampere won a new customer. That isn't necessarily a conflict of interest so much as an alignment of one, but it's one more example of how blurry the line between customer and investor has become across the AI supply chain, the investors financing this buildout are increasingly the same companies buying what gets built with it, whether that's Oracle backing a chip supplier or a hyperscaler taking an equity stake in the AI lab it also pays to use.
The Sale That Validated the Bet
Estimates of Ampere's actual revenue vary widely and conflict with each other, from figures in the tens of millions of dollars a year to claims of a billion-dollar run rate, depending on the source and what's being counted. What isn't in dispute is the outcome: seven years after starting from a discarded chip project, Ampere had become valuable enough, on the strength of its cloud partnerships and its Arm-based challenge to Intel, that SoftBank agreed to buy the company outright for $6.5 billion in cash in March 2025. The deal formally closed on November 25, 2025, after clearing a review by the US Federal Trade Commission.
The price was actually a step down from the roughly $8 billion valuation Ampere had commanded in earlier private funding talks, a reminder that even a genuine success story in AI infrastructure doesn't always move in a straight line upward. Still, for a company built from a discarded chip project and a contrarian bet against her former employer, $6.5 billion in cash was a clear vindication. James stayed on as chairman and chief executive, with Ampere continuing to operate as an independent subsidiary rather than being folded into SoftBank's other holdings.
That detail matters almost as much as the price tag. SoftBank didn't fold Ampere into a hardware division or swap out its leadership, the way an acquirer signals it bought a product line to absorb or wind down. Keeping James in charge and Ampere semi-autonomous looks more like what a buyer does when it believes the company's edge is the team and the cloud relationships it built, not just the chip designs sitting in a patent filing.
Interesting Fact: Intel's own CPU sales to Oracle for its cloud servers didn't just slow down after Ampere arrived, they nearly disappeared. Oracle's disclosed purchases of Intel processors for its cloud fell from tens of millions of dollars a year earlier in the decade to just a few million dollars annually by 2025, as Ampere chips took over more of that workload.
Knowlegic Perspective
It's tempting to file this story under "chip wars," but the more interesting detail is what Renee James actually bet on. She didn't try to out-invent Nvidia at the thing Nvidia was already best at. She looked at the part of the data center everyone assumed was permanently Intel's, and decided permanence was an assumption, not a fact.
That's a different kind of AI story than the one usually told. Most of the industry's founder narratives are about racing toward some new frontier, a bigger model, a stranger modality. James's story is about noticing that an old, boring, extremely profitable piece of infrastructure had quietly become contestable, and having the patience to contest it for the better part of a decade before anyone outside the industry noticed.
The AI boom rewards spectacle, the trillion-dollar GPU maker, the lab racing toward some new capability. It also, less visibly, rewards whoever is willing to spend years grinding away at the unglamorous layer everyone else assumed was already settled. Renee James built a chip company to take on Intel, not Nvidia, and the AI boom she quietly rode along the way was the thing that finally made that bet worth $6.5 billion.
Sources & References
- SoftBank Group to Acquire Ampere Computing, SoftBank Group Corp. (2025)
- SoftBank to acquire chip designer Ampere in $6.5 billion deal, CNBC (2025)
- Former Intel executive Renee James launches Ampere, a new Arm-based server chip company, GeekWire (2018)
- Oracle discloses $40 million stake in Ampere, a chip start-up founded by former Intel president Renee James, CNBC (2019)
- CPUs are Back: The Datacenter CPU Landscape in 2026, SemiAnalysis (2026)
Enjoyed this?
Get notified when a new Knowlegic story worth knowing is published.