Salience & Tower Partner on Optical Switches for Next-Gen AI

MIGDAL HAEMEK, Israel, and OXFORD, England – February 25, 2026 – By Kevin Lee

In a move set to address the mounting data and power challenges of the artificial intelligence boom, UK-based Salience Labs and Israeli foundry Tower Semiconductor have announced a strategic partnership to mass-produce a new generation of optical circuit switches. The collaboration transitions the cutting-edge technology from development into pre-production, aiming to equip future AI data centers with ultra-fast, low-latency, and power-efficient connectivity.

The partnership combines Salience Labs’ innovative photonic switching architecture with Tower Semiconductor’s high-volume silicon photonics manufacturing platforms. This alliance is poised to tackle the critical bottlenecks that threaten to stall progress in large-scale AI, offering a path to industrialize a technology essential for the next wave of computational infrastructure.

The Optical Revolution in AI Infrastructure

As AI models grow in complexity, the infrastructure supporting them is straining under the pressure. Traditional data center networks rely on Electronic Packet Switches (EPS), which convert optical signals from fibers into electrical signals for processing and then back again—a process known as optical-electrical-optical (OEO) conversion. This constant conversion creates latency and consumes a significant amount of power, creating a major roadblock for performance as AI clusters scale.

Optical Circuit Switching (OCS) represents a paradigm shift. By creating an all-optical path between network endpoints, OCS eliminates the need for OEO conversions for data traffic. This approach, where light carries data from input to output without being converted to electrons, drastically reduces latency and power consumption. A study by one hyperscale operator suggested that implementing OCS could slash a data center’s energy consumption by as much as 40% compared to traditional architectures.

Salience Labs' approach is built on a proprietary design using Semiconductor Optical Amplifiers (SOAs). This technology delivers exceptionally low latency and minimal signal loss, making it ideal for the demanding, high-throughput communication required between GPUs in large AI training clusters. Furthermore, OCS technology is inherently speed-agnostic, meaning the same hardware can support future upgrades from today’s 800 Gbps speeds to the 1.6 Tbps and 3.2 Tbps rates expected in the coming years, protecting infrastructure investments.

A Strategic Alliance to Scale Innovation

This partnership is a textbook example of synergistic collaboration. Salience Labs, a 2021 spinout from the University of Oxford and the University of Münster, brings over a decade of foundational research and a novel, validated architecture. The company has already attracted significant investor confidence, raising $45.3 million to date, including a recent Series A round.

Tower Semiconductor provides the crucial path to commercial scale. By leveraging Tower’s mature and differentiated silicon photonics platforms—specifically the PH18DA platform with integrated III-V lasers and the TPS45PH platform with low-loss nitride waveguides—Salience Labs can move its designs from the lab to high-volume manufacturing. The ability to integrate lasers directly onto the silicon chip is a key manufacturing advantage that simplifies packaging and improves overall efficiency.

“Tower is a key partner for Salience Labs, supporting our roadmap with its silicon photonics and switching technology platforms,” said Vaysh Kewada, Founder and CEO of Salience Labs, in the announcement. “Our collaboration builds on our deep expertise in silicon photonics and specialty platforms, strengthening our ability to deliver optical switch technology optimized for the performance and power demands of AI data centers.”

For Tower, the partnership solidifies its position as a critical enabler in the rapidly expanding AI hardware market. “Silicon photonics with integrated light sources is a key enabler for scaling next-generation optical connectivity, and our collaboration with Salience Labs reinforces our strong momentum in AI and data-center infrastructure,” commented Dr. Ed Preisler, Vice President and General Manager of RF Business Unit at Tower Semiconductor.

Navigating a Competitive and Exploding Market

The market opportunity is immense. Market research firm Dell’Oro Group projects that spending on switches for AI back-end networks will soar past $100 billion by 2030. This explosive growth has attracted a host of major players and innovative startups, all racing to provide the interconnects for the AI revolution.

Established foundry giants like GlobalFoundries are aggressively expanding their silicon photonics offerings, while tech titans like Intel, despite divesting parts of its photonics business, continue to innovate in optical I/O for AI. The competitive field also includes a new class of startups such as Ayar Labs and Celestial AI, which are developing their own unique photonic interconnect solutions. The Salience-Tower alliance aims to carve out a dominant position by combining Salience’s low-latency SOA-based architecture with Tower's proven, at-scale manufacturing processes, offering a clear and rapid path to deployment for data center operators.

Building a Greener Future for Artificial Intelligence

Beyond raw performance, the partnership addresses one of the most pressing concerns of the AI era: its enormous energy footprint. The U.S. Energy Information Administration has projected that data centers could consume over 6% of the country’s total electricity by 2028. AI workloads are the primary driver of this surging demand.

By dramatically reducing the power consumed by network switching, OCS technology offers a tangible solution to improve energy efficiency. Lowering the energy-per-bit for data transmission not only leads to direct operational cost savings but also contributes to building a more sustainable digital infrastructure. While GPUs will remain the dominant power draw in AI clusters, optimizing the network fabric is a critical piece of the puzzle. An efficient network ensures that expensive, power-hungry GPUs are utilized to their full potential, reducing idle time and maximizing the computational work done for every watt consumed.

As the industry converges for the upcoming OFC 2026 Conference in Los Angeles, where both companies will be present, the conversation will undoubtedly center on how to sustainably scale the infrastructure for AI. The move by Salience Labs and Tower Semiconductor to bring OCS technology into the mainstream manufacturing ecosystem represents a vital step in ensuring the future of AI is not only powerful but also practical and efficient.