In a bid to tackle the massive energy demands of AI data centers and speed up artificial intelligence processes, investors are pouring funds into photonics startups. These innovative companies use light instead of electricity to transmit data, offering a promising solution to some of the tech industry’s biggest challenges.
The Rise of Photonics
What is Photonics?
Photonics is the science of generating, manipulating, and detecting light. As AI technology expands, the need for more efficient data transmission has made photonics a hot topic among tech companies.
Major Investments
Recently, British company Oriole Networks raised $22 million from London-based venture capital firm Plural. This funding aims to develop a new networking infrastructure for AI supercomputing clusters that relies on light for data transmission.
Other significant investments in the field include:
- Lightmatter, which secured $400 million in funding, valuing the company at $4.4 billion.
- Xscape Photonics, which raised $44 million with participation from Cisco and Nvidia.
Challenges in Traditional Data Centers
As AI models grow in size and complexity, traditional data centers face several hurdles:
- Energy Consumption: Conventional networking equipment, primarily using copper wiring, consumes a lot of power and generates heat, leading to higher cooling costs. In some cases, networking equipment can account for 20% of a data center’s total energy consumption.
- Data Transmission Bottlenecks: Large AI models often require data to travel through multiple network switches, leading to delays—up to 90% of training time can be spent waiting for data.
How Photonics Can Help
Photonics offers a way to solve these issues by using fiber optics to transmit data as light, which is more efficient and produces less heat. This technology allows for:
- Direct Connections: AI chips can connect directly to each other, reducing the need for switching equipment and minimizing delays.
- Higher Data Capacity: Companies like Xscape are developing systems that can transmit hundreds of colors of light through a single cable, significantly increasing data flow.
The Vision of Oriole Networks
Oriole Networks is taking a bold approach, aiming to connect every AI chip in a supercomputing cluster directly to each other. This could lead to training times for AI models like GPT-4 being reduced by 10 to 100 times and cut energy usage significantly.
To achieve this, Oriole plans to introduce:
- New Communication Equipment: This will focus solely on photonics, eliminating the need for electrical packet switching.
- Innovative Software and Hardware: A new software system will manage data flow, acting as the “brain” of the network.
A Radical Shift
As Oriole CEO James Regan puts it, “It’s completely radical. There’s no electrical packet switching in the network at all.” This transformation could redefine how data centers operate, making them faster and more energy-efficient.
Conclusion
With the tech industry increasingly looking to photonics to solve pressing issues in AI data processing, the influx of investment signals a promising future. As these startups continue to innovate, they may very well lead the way in creating more sustainable and efficient data centers.
I am Aparna Sahu
Investment Specialist and Financial Writer
With 2 years of experience in the financial sector, Aparna brings a wealth of knowledge and insight to Investor Welcome. As an accomplished author and investment specialist, Aparna has a passion for demystifying complex financial concepts and empowering investors with actionable strategies. She has been featured in relevant publications, if any, and is dedicated to providing clear, evidence-based analysis that helps clients make informed investment decisions. Aparna holds a relevant degree or certification and is committed to staying ahead of market trends to deliver the most up-to-date advice.