Advanced Micro Devices Inc. (AMD) has officially acquired Enosemi Inc., a discreet startup based in Sacramento that specializes in developing cutting-edge photonics chips. The move underscores AMD’s aggressive push to strengthen its artificial intelligence (AI) hardware ecosystem and close the gap with competitors like Nvidia Corp. Although AMD did not reveal the financial terms of the deal, the strategic intent behind the acquisition sends a clear message.
AMD aims to build a tightly integrated AI infrastructure solution. Enosemi’s expertise in photonics and optical interconnects will play a pivotal role in achieving this goal.
AMD Expands AI Hardware Strategy
AMD continues to broaden its AI ambitions. The acquisition of Enosemi follows AMD’s $4.8 billion purchase of ZT Systems last year, a company known for its capabilities in designing and building rack-scale data center systems. With these moves, AMD no longer positions itself solely as a semiconductor designer. The company now actively builds a full-stack AI infrastructure platform.
In large AI clusters, engineers must interconnect graphics cards, processors, switches, and networking components across multiple server racks. This complex arrangement demands high-speed, low-latency communication between systems—something traditional copper wiring often fails to deliver efficiently. Photonics offers a solution.
Photonics leverages light instead of electricity to transmit data. As AI models become larger and more compute-intensive, the speed and efficiency of photonic communication become indispensable. AMD understands this shift and integrates photonics into its infrastructure design to meet the rising performance demands of AI applications.
Enosemi’s Photonic Expertise Meets AMD’s Ambition
Enosemi focuses on developing photonic chips tailored for fiber-optic data center networks. These chips support high-bandwidth, low-latency communication between servers, making them ideal for AI workloads. AMD plans to embed these capabilities into its infrastructure to support emerging needs in AI model training and inference.
Brian Amick, AMD’s Senior Vice President of Technology and Engineering, highlighted the importance of fast and efficient data movement. “As AI models grow larger and more complex, the need for faster, more efficient data movement is accelerating,” he stated. “To meet these evolving demands, particularly at the rack scale, optical interconnects offer a compelling path forward.”
AMD will now scale its photonics development by absorbing Enosemi’s talent and intellectual property. This strategic move not only deepens AMD’s in-house capabilities but also accelerates time-to-market for new AI-centric hardware products.
Building AI Systems That Compete with Nvidia’s DGX
AMD’s path appears increasingly similar to Nvidia’s. Nvidia has long offered integrated systems like the DGX series, which combine GPUs, networking equipment, cooling systems, and other components in a single turnkey solution. These systems eliminate the complexity of manually assembling server clusters and appeal to enterprises and research institutions working on high-performance AI projects.
AMD seeks to replicate—and eventually outdo—that success. The acquisition of ZT Systems gave AMD the capacity to design and manufacture rack-scale solutions. With Enosemi now in the fold, AMD can elevate the networking backbone of its systems by embedding optical technologies into the very core of its architecture.
This will allow AMD to ship AI server clusters with tightly integrated optics, ensuring high performance, reduced latency, and improved energy efficiency.
Co-Packaged Optics: The Next Frontier
A key technology that AMD will now aggressively pursue is co-packaged optics (CPO). This emerging hardware innovation replaces the need for separate, pluggable transceivers by embedding them directly into the switch or processor chip.
In conventional systems, servers convert electrical signals into light using pluggable transceivers, which then transmit data across fiber-optic cables. However, as AI clusters scale, the number of transceivers needed increases exponentially, adding to both cost and complexity.
CPO technology addresses this by fusing optical transceivers directly with the chips that drive the system. This tight integration results in several advantages:
- Lower Latency: With the transceiver built into the switch, data travels a shorter distance, reducing delay.
- Reduced Power Consumption: Less energy is needed to move data between components.
- Cost Efficiency: Eliminating pluggable modules decreases overall hardware costs for data center operators.
Enosemi brings critical know-how in this space, and AMD now positions itself to deploy CPO across its next-generation AI systems. The company aims to design more streamlined and powerful server clusters that scale effectively and operate with higher efficiency.
From Partnership to Integration
Enosemi had already been working with AMD as an external development partner in the photonics domain. According to Amick, the collaboration delivered successful results. “Enosemi has collaborated with us as an external development partner on photonics, and this acquisition extends that successful relationship,” he said. “Now as part of AMD, the team will help us immediately scale our ability to support and develop a variety of photonics and co-packaged optics solutions across next-gen AI systems.”
This seamless transition from partnership to acquisition ensures continuity and accelerates innovation. AMD can now synchronize engineering roadmaps more tightly, share intellectual property freely, and consolidate R&D investments to speed up product development.
Industry Implications and Competitive Landscape
The semiconductor industry currently stands at the heart of the AI revolution. Companies such as AMD, Nvidia, and Intel are racing to build chips and systems that can support the massive computational requirements of large language models, computer vision, and generative AI.
Nvidia dominates the AI accelerator market with its GPUs and integrated DGX systems. However, AMD has begun to close the gap with the launch of its MI300 series and recent acquisitions. The addition of Enosemi signals AMD’s recognition that AI hardware must evolve holistically—not just with better GPUs but also with faster networking, improved memory hierarchies, and smarter power distribution.
By acquiring companies that specialize in rack-scale design and photonics, AMD gains a structural advantage. Instead of relying on third-party vendors for critical system components, AMD now builds more of the AI stack in-house.
This vertical integration enables better optimization, tighter security, and quicker deployment—all factors that appeal to enterprise clients and hyperscalers.
The Road Ahead for AMD
AMD has made its intentions clear. It wants to dominate the AI infrastructure market not just by selling chips, but by delivering complete, optimized systems. With ZT Systems and Enosemi under its umbrella, AMD accelerates toward that future.
The company will continue to expand its engineering teams, develop AI-specific architectures, and strengthen its ecosystem of data center partners. The focus now shifts to execution—integrating new technologies and delivering on the promise of faster, more efficient AI systems.
As AI models grow in size and complexity, the winners in this space will be those who combine hardware innovation with system-level vision. AMD has now laid the foundation. The next wave of competition will prove whether it can fully capitalize on this strategic positioning.
Conclusion
The acquisition of Enosemi marks a turning point in AMD’s AI strategy. By bringing advanced photonics and co-packaged optics technology in-house, AMD reinforces its ability to design high-performance, cost-effective AI systems. As the battle for AI infrastructure dominance intensifies, AMD now holds more cards—and the expertise to play them well.