microLEDs: from headlamps to the data center
With new optical interconnect technology, data centers can increase bandwidth, efficiency and reliability
microLEDs: from headlamps to the data center
When we think about the evolution of AI technology, developments in machine learning and large language models come readily to mind, as do the latest graphics processing units (GPUs), high-bandwidth memory (HBM), and exotic semiconductor technologies such as chiplets and heterogeneous integration.
But car headlamps?
Automotive lighting is not usually seen as an inspiration for the next big thing in AI. Interestingly however, high-tech ‘adaptive beam’ front lights are proving the reliability and scalability of a valuable optical connectivity technology which could help data center operators to meet today’s challenges of increasing network bandwidth, efficiency and reliability.
But let’s take a step back to understand how headlamps could possibly be relevant to the technology of AI data centers.
The data transfer bottleneck
The steeply rising demand for AI – for both training and inference – is imposing huge pressure on equipment manufacturers to increase system compute capacity. Today, the main bottleneck is actually not in the compute function, thanks to the high performance of the accelerators (‘xPU’ devices such as GPUs and neural processing units, or NPUs) which process AI data inputs at astonishingly fast rates. The problem is the rate at which data can be shifted between xPUs, between an xPU and memory, and between servers in a rack.
Traditional copper electrical interconnects, which are familiar, cheap, and easy to integrate into electrical circuits, scale badly: over the distances required to scale up AI compute systems – sometimes up to 30m – copper interconnects require higher transmit energy, stronger equalization, and more complex signal conditioning, in part to overcome problems with electromagnetic interference (EMI). This in turn causes copper links to consume more power and generate more heat. This hamstrings equipment manufacturers’ efforts to increase bandwidth density (Gbits/s/mm) as well as system efficiency and reliability.
So, in order to increase network bandwidth, data centers have turned to optical interconnect technology. To do so, they have imported technology from the internet backbone: the companies which shuttle internet traffic between continents via a limited number of undersea cables – which are very difficult and expensive to install and maintain – have mastered the art of maximizing throughput per cable. Today’s intercontinental optical transport networks have links operating at high per-lane rates of up to 1.6Tbits/s.
This is the ‘fast-and-narrow’ approach to maximizing bandwidth: shoot as much data down each optical ‘pipe’ as fast as possible. But such high-frequency systems are complex, power-hungry and expensive. What is more, each link is a hugely consequential single point of failure which poses a high risk to system availability. And it is unclear how much further the fast-and-narrow architecture can be scaled, as every additional rise in per-lane speed and throughput becomes exponentially more difficult and expensive to implement.
In internet infrastructure, the high cost of cabling means that operators have no choice but to maximize data throughput per cable. But data centers do not face the same constraint.
From a single optical channel to thousands
This is why there is increasing interest in the data center world in a new approach: instead of going fast-and-narrow, why not go slow-and-wide? This means replacing a single, ultra-high speed link with hundreds or thousands of slower parallel optical channels, to give higher total bandwidth while using simpler, cheaper, lower-speed components.
In such a system, the data transmitters can be hundreds of microLEDs, replacing the single, high-power laser light source used in internet infrastructure. Data communications equipment manufacturers have not previously assembled parallel optical channels served by hundreds or thousands of microLED transmitters, so this slow-and-wide concept is unproved in data centers.
What a slow-and-wide architecture requires is the ability to deposit hundreds of optical emitters very close to a server’s data processors and data storage components. And because demand for AI never sleeps, data center operators need these miniature emitters to support reliable, 24/7/365 operation. But the deployment of chip-scale arrays of thousands of microLEDs has been validated in the demanding automotive market – and this is where the car headlamp can offer inspiration to data center equipment manufacturers.
It is precisely these capabilities which are in evidence in the microLED technology in ams OSRAM’s EVIYOS™ light source for adaptive beam headlights.
Fig. 1: In the EVIYOS product, ams OSRAM integrates the microLED array and driver circuitry into a single compact package
A single EVIYOS chip contains an array of 25.600 microLEDs, each in the size of half of a human hair and integrated with a CMOS driver chip in a single package measuring just 22.0mm x 17.5mm. Each of the 25.600 ‘pixels’ is individually controllable, so that the headlamp can cast complex patterns on the road surface. The unique integration technology between microLEDs and CMOS driver was awarded the German Future Prize as the “Digital Light” and products based on that proved to be reliable and robust in current production vehicles.
Now this same technology is being adapted for use in ultra-high bandwidth optical interconnects for AI data centers.
Proven micro-emitter technology
For use in optical interconnects, a similar fabrication technology is used as for headlamps, but whereas in headlamps the microLEDs are configured in dense monolithic arrays, for data connections the microLEDs are singulated – cut from the wafer on which they are fabricated – and mounted on a substrate to allow each emitter to interface to its own fiber optic cable or waveguide. This substrate can then be assembled on a target CMOS wafer.
Fig. 2: Enabling manufacturability for microLEDs in high-speed optical links
Because the microLEDs are so small, data communication transceivers based on them can achieve very high bandwidth density. Internal ams OSRAM studies have shown that microLED emitters can achieve per-lane data rates of 3.0Gbits/s at very low energy levels (smaller than 2pJ/bit) for the entire 10m link, while meeting the bit error rate of less than 10⁻¹⁵ (BER) specifications of industry standards.
And by replacing a single ultra-high speed line with hundreds of parallel connections, AI equipment manufacturers can gain other benefits which are valuable in data center applications:
- Reliability - microLEDs can be configured in systems with multiple redundant channels. This means that a single faulty emitter can be allowed to fail gracefully and be replaced by a spare channel.
- Efficiency – in a slow-and-wide architecture, each emitter operates at relatively low switching frequency – typically around 1GHz – which reduces power consumption compared to ultra-high frequency laser transmitters in internet infrastructure. Lower power consumption also results in lower waste heat generation, giving equipment operators more headroom in their thermal budget.
- Simplicity – the fast-and-narrow architecture requires the parallel streams of data processes in an AI accelerator to be serialized for transmission and deserialized after reception. The slow-and-wide architecture based on ams OSRAM microLEDs is inherently parallel, eliminating the need for complex and costly serialization.
Partnering with the datacom world
ams OSRAM is the only volume manufacturer of microLEDs in Europe. Its production capability is proven – it has been supplying volume shipments of the EVIYOS product to the automotive industry since 2023.
Now ams OSRAM is ready to collaborate with data center equipment manufacturers to adapt the technology for use in optical interconnects – for instance, to integrate the driver circuitry appropriate to high-frequency data transmitters, to optimize the package design, and to ensure interoperability with commercial sources of optical connectors, cables and fibers.
In this development endeavor, manufacturers will be helped by the broad optical system capabilities of ams OSRAM: the company manufactures photodiodes (optical receivers) as well as microLEDs, so can play a central role in the production of fully integrated optical data transfer systems.
Fig. 3: A concept for integration of a microLED array with photodiode receivers into a high-bandwidth data transport solution for AI data centers
Early concept designs for integrating microLEDs with optical interconnects are already available. If you’d like to learn more or explore co creation opportunities with ams OSRAM, feel free to contact us.