9 AMD Acquisitions Fueling Its AI Rivalry With Nvidia

The chip designer’s acquisitions of software firms, an AI lab, a silicon photonics startup, a data center infrastructure provider and the team behind an AI chip startup have all served to improve and build out AMD’s AI capabilities and ramp up its competition against Nvidia.

While AMD has made significant investments to speed up development of its Instinct data center GPUs to go head-to-head with Nvidia’s most powerful AI chips, the company has also leaned heavily into acquisitions so that it can offer “end-to-end AI solutions.”

At its Advancing AI 2025 event next Thursday, the Santa Clara, Calif.-based chip designer is expected to share its “bold vision” for AI and announce its next generation of Instinct GPUs, including the MI400 slated for release next year, among other things—efforts of which have been boosted by several acquisitions the company has made over the past three years.

[Related: Analysis: Nvidia Vs. Intel Vs. AMD Q1 Earnings Face-Off]

“Early customer feedback [for the Instinct MI400] has been very positive, marking a major step forward in our Instinct roadmap and significantly expanding our AI accelerator [total addressable market] as customers plan broader Instinct deployments to power a larger share of their AI infrastructure,” AMD CEO Lisa Su said on her earnings call last month.

AMD and Nvidia have been long-time rivals, but competition between the two companies kicked into overdrive when the latter a few years ago began to see its data center business grow by major leaps thanks to its GPUs being adopted for generative AI development.

Forrest Norrod, the head of AMD’s data center solutions business unit, told CRN last year that he believes Nvidia decided to accelerate its GPU road map to an annual release cadence in response to increasing competition, particularly from AMD.

AMD, in turn, announced last year that it would do the same, with the Instinct MI325X launching later in 2024 and the MI350 on track to begin production soon.

“We think we are closing the gap, narrowing the gap between the introduction of Nvidia's part and the introduction of our same generation part,” Norrod said at the time.

To flesh out its GPU, system and software capabilities, AMD has made several acquisitions since 2023, starting with software firms Mipsology and Nod.ai in 2023 and then continuing with AI lab Silo AI and data center infrastructure provider ZT Systems last year.

That acquisition spree has continued into this year, with the company announcing that it scooped up silicon photonics startup Enosemi, compiler software startup Brium and the team behind AI chip startup Untether AI in the last nine days.

These acquisitions have all served to improve and build out AMD’s AI capabilities, particularly when it comes to the data center market, where the greatest potential for revenue growth and profitability lies.

But the company’s AI strategy and its ability to compete with Nvidia has also benefited from acquisitions prior to its latest shopping spree. This includes the company’s 2022 acquisitions of programmable chip designer Xilinx and networking chip designer Pensando, which have both served to expand AMD’s opportunities with new products and markets.

What follows is a roundup of nine acquisitions AMD has made over the past three years to fuel its AI strategy, including two that were announced this week.

Untether AI Team

AMD confirmed Thursday that it acquired the employees behind Untether AI, a developer of AI inference chips marketed as faster and more energy-efficient than rival products for edge environments and enterprise data centers.

“AMD has entered into a strategic agreement to acquire a talented team of AI hardware and software engineers from Untether AI,” an AMD spokesperson told CRN in a statement.

Financial terms of the deal were not disclosed.

“The transaction brings a world-class team of engineers to AMD, focused on advancing the company’s AI compiler and kernel development capabilities as well as enhancing our digital and SoC design, design verification, and product integration capabilities. We are excited to welcome the team’s unique expertise to AMD,” the representative added.

Founded in 2018, the Toronto, Ontario-based startup had been selling AI inference chips for edge environments and data centers based on its “at-memory” architecture that it said “dramatically” improved performance and reduced power consumption.

“We’ve proven that when you reduce data traffic in the chip, you’re improving both throughput and energy efficiency, and so the difference between what you need to do [at the edge or in a data center], it’s really more about how many rows and columns do we put down in the memory interface,” Chris Walker, the former Intel executive who was Untether AI’s CEO until he stepped down last month, told CRN in an April interview.

Brium

AMD announced on Wednesday that it acquired Brium, which is made up of “world-class compiler and AI software experts with deep expertise in machine learning, AI inference and performance optimization.”

Financial terms of the deal were not disclosed.

Anush Elangovan, corporate vice president of software development at AMD, said Brium’s advanced software capabilities will “strengthen our ability to deliver highly optimized AI solutions,” including its Instinct GPUs that have been key to its fight with Nvidia.

“Their work in compiler technology, model execution frameworks and end-to-end AI inference optimization will play a key role in enhancing the efficiency and flexibility of our AI platform,” he wrote in a blog post.

The major advantage AMD sees with Brium is the startup’s “ability to optimize the entire inference stack before the model reaches the hardware,” according to Elangovan.

Enosemi

AMD announced in late May that it acquired silicon photonics startup Enosemi to “support and develop a variety of photonics and co-packaged optics solutions across next-gen AI systems.”

Enosemi’s photonic integrated circuits will help AMD enable “faster, more efficient data movement” within server racks that is required by ever-growing AI models, AMD executive Brian Amick said in a blog post.

“Co-packaged optics can deliver higher bandwidth density and better power efficiency than traditional approaches, representing a transformative step in system architecture where tighter integration between compute and networking is enabled to support the performance and scale that advanced AI workloads require,” he wrote.

The chip designer did not disclose how much it paid to acquire Enosemi.

Founded in 2023, Enosemi’s “elite team of experts and PhD-level talent” has a “proven record of building and shipping photonic integrated circuits in volume, a unique feat that few select teams have accomplished,” Amick said in his blog post.

ZT Systems

AMD said it completed its $4.9 billion acquisition of data center infrastructure provider ZT Systems in late March to combine the firm’s “industry-leading systems” and “rack-level expertise” with its silicon portfolio to offer a “new class of end-to-end AI solutions.”

The chip designer said ZT Systems’ rack-scale AI solutions design and customer enablement teams joined AMD’s data center solutions business, which is led by Executive Vice President Forrest Norrod.

In a statement, Norrod said the ZT Systems acquisition will allow AMD to reduce the time it takes to design and deploy “cluster-level” data center AI systems for customers.

“Acquiring ZT Systems is a significant milestone in our AI strategy to deliver leadership training and inferencing solutions that are optimized for our customers’ unique environment, ready-to-deploy at scale, and based on our open ecosystem approach that combines open-source software, industry standard networking technologies and now ZT Systems’ leadership systems design and customer enablement expertise,” he said.

More than a month later, AMD said reached an agreement to sell ZT Systems’ server manufacturing business to U.S. electronics manufacturing services giant Sanmina for $3 billion. The chip designer had sought a buyer for the server manufacturing business since it announced the ZT Systems deal last August.

As part of the deal with Sanmina, AMD said the company will become a preferred new product introduction (NPI) manufacturing partner for its cloud rack and cluster-scale AI solutions.

Silo AI

AMD announced last August that it had completed its $665 million all-cash acquisition of Silo AI, which it called the largest private AI lab in Europe and a developer of open-source multilingual large language models.

When the company announced the deal earlier in July, it said the acquisition represented “another significant step in the company’s strategy to deliver end-to-end AI solutions based on open standards and in strong partnership with the global AI ecosystem.”

Peter Sarlin, Silo AI’s CEO and co-founder, continues to lead the lab’s team and reports to Vamsi Boppana, senior vice president of AMD’s AI group.

Boppana, who joined AMD via its Xilinx acquisition in 2022, said the Silo AI acquisition will help enterprises develop and deploy AI solutions faster and more effectively.

“Silo AI’s team of trusted AI experts and proven experience developing leadership AI models and solutions, including state-of-the-art LLMs built on AMD platforms, will further accelerate our AI strategy and advance the build-out and rapid implementation of AI solutions for our global customers,” he said in a statement in July of last year.

Founded in 2017 and headquartered in Helsinki, Finland, Silo AI employs “world-class” AI scientists and engineers who have “extensive experience developing tailored AI models, platforms and solutions for leading enterprises spanning cloud, embedded and endpoint computing markets,” according to AMD.

Silo AI’s offerings are part of “end-to-end AI-driven solutions that help customers integrate AI quickly and easily into their products, services and operations,” and the company’s customers include major corporations like Allianz, Philips, Rolls-Royce and Unilever.

Part of the lab’s offerings are open-source multilingual large language models (LLMS) that run on AMD platforms as well as its own SiloGen model platform.

Nod.ai

AMD announced in October of 2023 that it had acquired Nod.ai to enhance its open software capabilities to better compete with Nvidia.

Founded in 2013, Noda.ai is a developer of compiler-based automation software to optimize AI solutions for hyperscalers, enterprises and startups.

Nod.ai’s software “accelerates the deployment of AI solutions optimized” for AMD’s portfolio of chips, including the Instinct accelerators, EPYC processors and Versal system-on-chips for data centers and edge computing as well as the Ryzen AI processors and Radeon GPUs for PCs, according to AMD.

This reduces the need to manually optimize and the time it takes to deploy “highly performant AI models” on AMD’s processors, the chip designer added.

“The acquisition of Nod.ai is expected to significantly enhance our ability to provide AI customers with open software that allows them to easily deploy highly performant AI models tuned for AMD hardware,” said Vamsi Boppana, senior vice president of AMD’s AI group, in a statement when the deal was announced.

Mipsology

AMD announced in August of 2023 that it had acquired French startup Mipsology to strengthen its AI inference software capabilities.

When the company announced the acquisition, it said the deal would “accelerate our customer engagements and expand our AI software development capabilities.”

“Specifically, the team will help develop our full AI software stack, expanding our open ecosystem of software tools, libraries and models to pave the way for streamlined deployment of AI models running on AMD hardware,” said Vamsi Boppana, senior vice president of AMD’s AI group, in a statement at the time.

Mipsology’s expertise is in developing “plug-and-play” software that speeds up AI inference performance without requiring new tools or changes to neural network models, which are key to AI applications.

The software, called Zebra, is focused on enabling “world-class speed” on FPGAs, short for field programmable gate arrays, that AMD gained with its acquisition of Xilinx, according to its website.

Pensando

Before the AI computing race started to heat up with the introduction of AMD’s Instinct MI300X GPU in 2023, the company completed its $1.9 billion acquisition of networking chip designer Pensando Systems roughly a year before.

When the company announced the closing of the acquisition in May of 2022, it said Pensando’s distributed services platform would expand AMD’s data center portfolio with a “high-performance data processing unit (DPU) and software stack that are already deployed at scale across cloud and enterprise customers.”

As AMD doubled down on its AI strategy over the following years, Pensando’s products would grow to serve the networking side of AI infrastructure.

This was highlighted late last year when the company announced the 400-Gbps Pensando Salina DPU that it called a “critical component in AI front-end network clusters, optimizing performance, efficiency, security and scalability for data-driven AI applications.”

AMD also announced the Pensando Pollara 400, which it called the “industry’s first Ultra Ethernet Consortium-ready AI NIC.” This component is “critical for providing leadership performance, scalability and efficiency of accelerator-to-accelerator communication in back-end networks,” according to the company.

Xilinx

AMD announced the completion of its $49 billion acquisition of programmable chip designer Xilinx in February of 2022.

When the closing of the deal was announced, AMD said the addition of Xilinx’s field programmable gate arrays (FPGAs) would significantly expand the company’s opportunities in data centers, embedded computing and telecommunications.

“The acquisition of Xilinx brings together a highly complementary set of products, customers and markets combined with differentiated IP and world-class talent to create the industry’s high-performance and adaptive computing leader,” AMD CEO Lisa Su said in a statement.

Xilinx would go on to fuel AMD’s AI strategy in various ways over the ensuing years.

On the personnel side, AMD ended up appointing Vamsi Boppana, a longtime Xilinx executive, as senior vice president of its new AI group.

In this role, Boppana is responsible for leading AI software and ecosystem development as well as AI hardware road maps. Then last year, he took over leadership of AMD’s Instinct data center accelerator chip business when AMD President Victor Peng retired.

The company has also taken the AI engine developed by Xilinx and began incorporating the technology into AMD products as a neural processing unit (NPU), most recently with the AMD Ryzen AI 300 series. AMD has previously said it plans to incorporate the AI engine into a future, unannounced EPYC CPU lineup.

In addition, AMD has also said that Xilinx’s would give the company much broader coverage in the AI market, with the Zynq adaptive chips being used for industrial use cases, the Versal adaptive chips being used for telecom use cases, and the Alveo accelerator chips and Kintex FPGAs being used in cloud data centers.