A faster optical transceiver is often imagined as little more than a faster wire. The reality, as the engineers at Linkstar describe it, is closer to a rethink of how light itself is generated, shaped, and routed. The Singapore-based deep tech company has built its 1.6 terabit photonic engine around that idea, and the design offers a useful window into where high-speed data movement is heading.
At its core, the engine runs two parallel chains. On the transmit side, a laser source feeds a photonic integrated circuit, where modulators encode digital data onto light using a signaling scheme known as PAM4. Those separate channels are then combined by an on-chip multiplexer and launched into an optical fiber. On the receive side, the same process runs in reverse, with photodetectors and amplifiers recovering the signal and converting it back into electrical form. In broad strokes, it is a system for turning electronics into light, carrying that light efficiently, and turning it back again at the far end.
What makes the design notable is not any single component but the degree of integration. In older approaches, many of the functions involved in shaping and directing light were handled by discrete parts, including separate filters, lenses, and beam splitters bolted onto a board. Linkstar’s approach designs those functions directly into the chip itself. The assembly is then mounted on a through-silicon via interposer using bump interconnects rather than traditional wire bonds. Collapsing those connections shortens the electrical path that signals have to travel, which preserves signal integrity at extreme speeds and shrinks the overall footprint of the device.
The payoff of that integration shows up in numbers that the company treats as its competitive scorecard. Moving from an 800 gigabit photonic chip to the 1.6 terabit architecture roughly doubles bandwidth density, halves the physical footprint of the module, and cuts power consumption by around forty percent. Each of those figures is more than a technical bragging right. Each is an axis on which the product competes.
Higher bandwidth density means more data can be pushed through the same faceplate, the limited front panel area of a switch where ports are mounted. For the hyperscale operators who run the largest data centers, that density is one of the metrics they prize most, because physical space and port count are real constraints. A smaller footprint means more ports per switch and less material used per unit. Lower power means less heat to remove and a lower total cost of ownership when the same device is deployed across millions of units. In a business where small per-unit differences multiply across enormous volumes, those gains add up quickly.
The architecture’s scalability is what ties the advantages together. Because the same fundamental design spans from 100 gigabit to 1.6 terabit, the engineering invested in one generation carries forward rather than being discarded. A company that establishes a lead can extend it into the next generation instead of restarting the race each time the industry adopts a faster standard. That continuity changes the economics of competing in a field that moves as fast as optical interconnects do.
Linkstar argues that this combination of integration, efficiency, and scalability is what allows a focused team to compete with much larger incumbents. Rather than winning purely on manufacturing scale, the company aims to win on bandwidth per watt and bandwidth per millimeter, the measures of how much data a device can move for every unit of energy and every unit of space it occupies. Those are the metrics that matter most as data centers strain against limits on power and physical room, and they reward clever design as much as sheer size.
There is also an organizational advantage built into the approach. By anchoring design, fabrication, and testing within a single Singapore-based ecosystem, Linkstar keeps the loop between invention and validation tight. When the people designing the chip, the people building it, and the people testing it work closely together, problems surface faster, and improvements move more quickly from idea to silicon. Fragmented supply chains, spread across many vendors and borders, rarely match that speed of iteration.
The deeper message in the engineering is that progress in moving data is now as much about packaging and integration as it is about the raw physics of light. The most meaningful gains come from collapsing distances, removing discrete parts, and designing functions directly into the chip. For Linkstar, the 1.6 terabit engine is not a one-off achievement but a reusable platform, and the company is betting that the discipline of building light at scale, generation after generation, is what will keep it ahead.
