360 gigabits per second. That’s the speed a new chip achieved in lab tests—using beams of light instead of radio waves to transmit data. The number isn’t a typo. It’s not a theoretical peak. It’s what researchers measured when they fired up a tiny silicon chip embedded with dozens of microscopic lasers, each tuned to carry a separate data stream. This isn’t incremental progress. It’s a hard pivot from how we’ve moved data wirelessly for over a century.
Key Takeaways
- The new laser-based wireless chip reached 360 Gbps in controlled tests—over 10 times faster than the fastest Wi-Fi 6E systems.
- It uses half the energy of conventional Wi-Fi for the same data throughput, a critical gain as network demand skyrockets.
- The chip integrates dozens of micro-lasers on a single silicon die, enabling parallel data transmission via light beams.
- Transmission is line-of-sight and short-range, limiting use to indoor environments like data centers or offices.
- The technology avoids radio frequency congestion, operating in an unregulated part of the optical spectrum.
The End of Radio’s Reign?
Wi-Fi has been on life support for years. Not dead—far from it. But it’s wheezing. Every new device, every video call, every smart thermostat adds pressure to a system already stretched thin. Radio spectrum is finite. Channels overlap. Interference is constant. We’ve patched it with beamforming, channel bonding, dual-band routers. But physics hasn’t changed. There’s only so much data you can shove through 2.4 GHz and 5 GHz.
That’s why this shift to light matters. It’s not just a faster lane—it’s a different highway. The chip’s array of micro-lasers emits tightly focused beams in the infrared range, each carrying a unique data stream. These don’t interfere with Wi-Fi, Bluetooth, or cellular signals. They don’t require FCC licensing. And because light has a much higher frequency than radio waves, it can oscillate faster—packing more data into the same amount of time.
The researchers didn’t name a specific use case, but the implications are obvious. Data centers. Server racks. High-frequency trading floors. Places where speed, low latency, and power efficiency are non-negotiable. These environments already use fiber optics for backbone links. This chip could eliminate the last few meters of cabling—replacing patch cords with invisible beams of light.
How the Chip Actually Works
The core innovation isn’t the laser itself. It’s the integration. The chip combines photonic and electronic circuits on the same die, allowing data to switch between electrical and optical domains without external converters. Each micro-laser is tuned to a slightly different wavelength, a technique known as wavelength division multiplexing. That lets them transmit simultaneously without crosstalk.
On the receiving end, another chip captures the beams using tiny photodetectors, converting light back into electrical signals. Alignment is critical—these aren’t diffuse signals like Wi-Fi. They require line-of-sight, which limits range but enhances security. An eavesdropper would need to physically intercept the beam, making passive snooping nearly impossible.
Speed vs. Practicality
360 Gbps sounds absurd until you contextualize it. That’s enough to download a 4K movie in under a second. It’s triple the peak speed of Thunderbolt 4. But raw speed doesn’t matter if you can’t use it. The system tested on April 2, 2026, as detailed in the original report, only worked over a few meters. Walls block the signal. Even a hand waved between transmitter and receiver can disrupt the link. That’s not a flaw. It’s the trade-off.
What’s impressive is the energy efficiency. The chip used just 0.5 picojoules per bit—half the energy of modern Wi-Fi. At that rate, a full day of continuous 360 Gbps transmission would consume less power than a single incandescent nightlight. For data centers, where networking can account for up to 20% of total energy use, that’s not just nice—it’s material.
Why Silicon Matters
Previous optical wireless systems relied on discrete components—bulk lasers, external modulators, separate receivers. They were fast but expensive, fragile, and power-hungry. What makes this chip different is that it’s built using standard silicon fabrication techniques. That means it can be mass-produced alongside conventional chips, without exotic materials or custom processes.
Silicon photonics has been a quiet revolution for years—used in high-end data center interconnects, but rarely in wireless. This is the first demonstration of a fully integrated silicon laser array capable of wireless transmission at these speeds. It’s a milestone not because it’s flashy, but because it’s manufacturable.
The Hidden Bottleneck: The Last Meter
We’ve spent decades optimizing wireless for mobility. Phones. Laptops. Wearables. But in high-performance environments, mobility is a liability. Every time you add multipath signaling or frequency hopping to avoid interference, you pay a tax in latency and complexity. What if, instead of making wireless more robust, we made it simpler—faster, but only over fixed paths?
That’s the philosophy behind this chip. It’s not meant for your living room. It’s for the server rack where every nanosecond counts. Right now, many of these systems still rely on copper cables or short-reach fiber. This chip could replace both. Imagine a blade server that talks to its storage unit via invisible laser links—no connectors, no wear, no signal degradation from electromagnetic noise.
And because the beams are directional, you can pack dozens of them into the same physical space without interference. The researchers demonstrated 32 parallel channels, each running at over 11 Gbps. That’s 32 lanes of optical data, all from a chip smaller than a postage stamp.
What This Means For You
If you’re building infrastructure for AI clusters, high-frequency trading, or scientific computing, this changes the calculus. You’ll be able to design systems with lower latency and far less power draw for inter-node communication. The need for physical connectors drops. Maintenance costs could follow. And because the system operates in the optical spectrum, you won’t have to worry about RF compliance or spectrum licensing.
For developers, the bigger impact is indirect. Faster, cheaper, lower-power interconnects mean that distributed systems can scale more efficiently. Training clusters can grow without hitting I/O bottlenecks. Real-time analytics pipelines won’t choke on data movement. The chip won’t show up in your laptop, but it might power the backend that serves your app.
Of course, that assumes the technology moves beyond the lab. There’s no word on commercialization timelines. No partnerships announced. No cost estimates. But the fact that it works at all—on silicon, at room temperature, at 360 Gbps—suggests it’s not just a lab curiosity. It’s a proof that light, not radio, might handle the heaviest lifting in future networks.
Will we all have laser internet? No. But the machines that run our digital world might.
Competing Approaches in Optical Wireless
While this new chip marks a leap, it’s not the only effort pushing optical wireless forward. Companies like Facebook (now Meta) explored free-space optical communication years ago through its Connectivity Lab, aiming to beam internet from drones. Their prototypes achieved multi-gigabit speeds over kilometers, but required precise tracking systems and clear atmospheric conditions—impractical for indoor use. More recently, startups like Airtame and Light Fidelity (LiFi) Ltd. have commercialized visible-light communication using LED bulbs. LiFi, championed by Harald Haas, can reach up to 224 Gbps in lab settings using off-the-shelf LEDs and photodetectors. But those systems still rely on bulky external components and struggle with signal integrity under ambient light.
The silicon-integrated laser array bypasses many of these issues. Unlike LiFi, which modulates existing lighting, this chip generates its own infrared beams, avoiding interference from room lights. And unlike drone-based optical links, it operates at centimeter-scale distances where alignment is manageable. Intel and GlobalFoundries have also invested in silicon photonics, but their focus has been on fiber-coupled transceivers for data center racks—not wireless. The breakthrough here is combining integration, directionality, and wireless freedom into one manufacturable package. Other labs, including MIT’s Photonic Microsystems Group and TU Eindhoven, are exploring similar architectures, but none have demonstrated 32-channel parallelism on a single die with sub-picojoule efficiency.
The Bigger Picture: Energy, Heat, and AI’s Network Hunger
Data centers are hitting hard limits—not just in bandwidth, but in power and heat. In 2025, global data center electricity use surpassed 460 terawatt-hours, according to the International Energy Agency, rivaling the annual consumption of countries like Germany. A significant chunk—up to 20%, per a 2024 Uptime Institute analysis—goes to networking gear. As AI models balloon in size, with clusters like those running GPT-class models requiring tens of thousands of GPUs, the demand for interconnect bandwidth has exploded. NVIDIA’s latest NVLink 5 can move 1.8 terabits per second between GPUs, but it’s limited to direct, cabled connections. Scaling that across racks introduces bottlenecks.
This new laser chip could alleviate that strain. At 0.5 pJ/bit, it’s twice as efficient as today’s best-in-class Wi-Fi 6E radios, which consume about 1 pJ/bit. If deployed at scale, replacing even a fraction of copper interconnects in a 100,000-server data center, the energy savings could reach gigawatt-hours per year. That’s not just cost reduction—it’s a sustainability win. Cooling requirements would also drop, since less wasted energy means less heat. For AI workloads, where training runs can last weeks, shaving microseconds off communication latency between nodes can cut days off total training time. Companies like Google, Microsoft, and Amazon are already investing in optical circuit switches and in-rack photonics. Integrating wireless optical links at the chip level could be the next logical step—one that aligns with their broader push toward disaggregated, modular data center architectures.
Policy and Spectrum Implications
One of the quiet advantages of optical wireless is regulatory freedom. Radio frequencies are governed by national and international bodies—the FCC in the U.S., Ofcom in the UK, the ITU globally. Licensing, interference rules, and spectrum auctions add cost and complexity. Wi-Fi 6E’s expansion into the 6 GHz band, for instance, required years of lobbying and technical negotiation. Optical wireless, especially in the infrared range, operates in unregulated spectrum. There are no licensing fees. No risk of violating emissions limits. No coordination with neighboring bands.
This makes deployment faster and cheaper, particularly in dense environments like urban data centers or multi-tenant office buildings where RF interference is a constant headache. It also sidesteps geopolitical issues around spectrum control—something increasingly relevant as nations compete over 5G and 6G dominance. While optical systems aren’t immune to regulation—safety standards like IEC 60825 govern laser exposure—these are well-established and easier to comply with than dynamic RF policies. For enterprises, this means IT teams could deploy high-speed optical links without involving legal or compliance departments. That agility could accelerate adoption, especially in sectors like finance, where firms like Goldman Sachs and Citadel have already adopted custom low-latency networks for trading. If chip-scale optical wireless becomes available off-the-shelf, it could democratize access to speeds once reserved for billion-dollar infrastructure projects.
Sources: Science Daily Tech, IEEE Spectrum, International Energy Agency, Uptime Institute, IEC, ITU, NVIDIA, Meta Connectivity Lab, LiFi Ltd.
