Source: Laser Manufacture News  13^th Aug 2026

As silicon-based computing approaches its physical limits, optical computing is accelerating from an academic vision toward industrial reality.

In 2025, a team of PhDs from leading institutions including MIT and Harvard founded Rationalite Photonics in Shanghai's Zhangjiang. With its globally pioneering "Implosion Fabrication" (Imp-Fab) 3D nanomanufacturing technology, the company has successfully broken through the "impossible triangle" of precision, throughput, and cost in micro-nano manufacturing. Just three months after its founding, Rationalite Photonics completed a seed round in the tens of millions of yuan, targeting edge-side scenarios such as AR/VR and embodied intelligence with its Application-Specific Optical Computing (ASOC) chips as the entry point.

Laser Manufacture News recently interviewed Yang Gaojie, founder of Rationalite Photonics, to hear her story of this frontier technology moving from the lab to industry, and to decode how this emerging company is redefining computing with light.

Yang Gaojie, Founder of Rationalite Photonics

Laser Manufacture News: Ms. Yang, hello. The core team at Rationalite Photonics comes from world-leading institutions such as Harvard and MIT. What prompted you to return to China to start a business and focus on the optical chip and optical computing track?

Ms. Yang: Our decision to return to China and start the company in 2025 came from several factors converging at that point in time.

First, the technology itself had reached a critical stage. At MIT, we explored many directions, but optical computing was the most rewarding one—it is not about making small improvements within the existing system, but about asking whether the architecture of next-generation computing hardware can fundamentally change. After exchanging views with leading international optical computing teams, we gained great confidence.

Second, we believed this technology could not remain confined to overseas laboratories forever. Global competition around AI computing, optical chips, and advanced manufacturing is accelerating. The sooner underlying original technology develops engineering and industrialization capabilities, the better the chance of seizing the initiative.

Third, the domestic environment has become increasingly mature. After returning, we received support from the national level as well as from Pudong and Zhangjiang in terms of policy, space, and industrial resources. For hard-tech startups, this support is crucial.

Our judgment was this: the technology window has arrived, industrial demand has arrived, and the domestic ecosystem has arrived. What Rationalite Photonics aims to do is truly transform the original 3D manufacturing capabilities accumulated at the frontline of scientific research into a chip-level 3D optical computing platform.

Laser Manufacture News: The company completed its seed round just three months after founding. Beyond the team's background, what technological potential at Rationalite Photonics impressed investors the most?

Ms. Yang: What investors value more fundamentally is the platform value of the technology itself. What Rationalite Photonics is building is not a single optical computing device, but an original 3D micro-nano manufacturing process and a chip-level 3D optical computing architecture. once established, it can support not only optical computing chips but also extend to advanced optical devices, optoelectronic packaging, and more novel chip structures.

At the same time, this technology is no longer at a purely conceptual stage. We have completed key proof-of-principle validation and mastered the complete core chain, from materials and processes to structural design and device testing.

Commercially, our advantages lie in two areas: first, the process route has scalable potential in terms of materials and processing costs; second, the technological barrier is very high—it cannot be replicated simply by buying equipment and tuning parameters, but requires long-term accumulated systematic capabilities.

This funding has helped us accelerate the construction of our Zhangjiang R&D platform, advance R&D on 3D optical computing chips and optoelectronic hybrid prototypes, expand our engineering team, and quickly begin validation in real-world scenarios.

Laser Manufacture News: The "Imp-Fab" technology was published in Nature Photonics, breaking through the "impossible triangle" of precision, throughput, and cost. Could you explain in plain terms this reverse-thinking process of "micron-scale sculpting first, then implosion shrinkage"?

Ms. Yang: Micro-nano manufacturing has long faced a practical dilemma: high precision often means high cost and low efficiency, while large-area, high-throughput production struggles to maintain nanoscale precision.

The idea behind Imp-Fab is this: first, write micron-scale structures into a larger 3D material system using relatively mature, lower-cost laser processing, then shrink the entire 3D structure down to the nanoscale through controlled material contraction. The advantage is that it lowers the difficulty of direct nanoscale fabrication while preserving full 3D structural freedom.

This matters especially for optical chips. Electronic chips are well suited to high-density integration on a 2D plane, but light has properties such as wavelength, phase, and diffraction. If you try to integrate it purely in 2D, you easily run into crosstalk, loss, and limited design freedom. If true 3D optical structures can be built, light can propagate, couple, and compute in three-dimensional space.

That is the significance of Imp-Fab—it gives us a chance to manufacture 3D nanoscale optical modulation structures that were previously very difficult to achieve, using a relatively low-cost, high-throughput approach.

Laser Manufacture News: based on its 3D manufacturing capability, Rationalite Photonics has chosen the Application-Specific Optical Computing (ASOC) route. Could you discuss ASOC's advantages in the context of specific scenarios?

Ms. Yang: ASOC, as we understand it, is not meant to replace GPUs or CPUs. Rather, it takes on a clearly defined class of tasks in edge AI systems: completing high-frequency, repetitive front-end inference computations that are extremely sensitive to power consumption and latency, using optical structures in advance.

Take AR/VR glasses as an example. They need continuous eye tracking, gesture recognition, and spatial perception, but the devices are highly sensitive to power consumption, weight, and heat. ASOC can sit closer to the sensor front end, using light propagation through 3D structures to complete part of the feature extraction first, compressing raw information into intermediate results before handing them to back-end electronic chips.

Embodied intelligence robots follow a similar logic. Many low-level perception tasks do not require a full GPU to be involved at every step. If part of the parallel computation can be done at the sensing front end using low-power optical devices, robots can achieve faster response times and lower overall system energy consumption.

Compared with other optical computing routes, our differentiation lies in 3D integration and specialization. Many silicon photonic MZI approaches offer programmability, but they rely on 2D waveguide networks and large numbers of modulation units, which increases system area, power consumption, and complexity. Our ASOC route does not pursue "computing everything," but instead uses the Imp-Fab process to push the part of computing best suited to light to the extreme. The optical computing that truly reaches deployment in the future is very likely to be a new optoelectronic hybrid architecture.

Laser Manufacture News: You mentioned the concept of "physics as algorithm" based on optical diffraction neural networks (ODNN). What challenges arise in engineering when turning AI inference into the natural propagation of light through 3D structures?

Ms. Yang: "Physics as algorithm" is not simply about transferring a software model into a material. It is about converting the most time-consuming optical transformations in AI inference into the natural propagation of light through precisely designed 3D structures. There are indeed several thorny issues in engineering this.

The first is the mapping between algorithm and manufacturing. In an algorithm, weights are numbers, but in a chip, they correspond to the position, refractive index, and height of every micro-nano unit in 3D space. A model being theoretically feasible does not mean it is manufacturable.

The second is nanoscale 3D fabrication and interlayer alignment. Visible light has short wavelengths, so the neurons in an ODNN must be sub-micron or even nanometer-scale. Traditional lithography can only fabricate each layer separately, then align and package them. Achieving 3D alignment of multilayer chips at tens-of-nanometers precision is an engineering disaster. The advantage of Imp-Fab is that multiple ODNN layers and their spatial relationships can be fabricated in a single 3D material volume in one go, without complex 3D alignment.

The third is scalable mass production. The Imp-Fab technology itself has strong potential for mass production—the material system is relatively simple, and laser processing technology is relatively mature. In the future, throughput can be improved through line scanning, parallel optical writing, and other methods.

Laser Manufacture News: The company plans to launch an "optoelectronic hybrid acceleration architecture," assigning linear operations to light and nonlinear operations to electronics. What kind of performance gains can this roadmap bring to customers, and when do you expect to see a demo?

Ms. Yang: From the very beginning, we were clear: we are here to solve industry pain points, not to show off technology. Why choose "optoelectronic hybrid"? In today's AI model computation, more than 80% of computing power is consumed by matrix multiplication. This is precisely where optical computing has an absolute advantage—our 3D optical chips can, like a wide-body jet, complete massive parallel data processing instantly at extremely low power consumption.

Compared with traditional electronic chips at the same process node, this architecture is expected to deliver leaps of one to two orders of magnitude in energy efficiency (TOPS/W) and computing density. For downstream customers, this means future lightweight AR glasses will be able to run high-parameter AI models locally and smoothly without carrying bulky batteries and cooling fans.

As for progress, we are actually moving faster than outsiders might expect. We have already secured commercial cooperation orders in the tens of millions of yuan and are conducting deep joint development with frontier customers such as brain-computer interface companies that have extremely demanding requirements for power consumption and latency. We expect to showcase our first engineering-grade demo to the industry between the end of this year and the first half of next year.

Laser Manufacture News: What is the evolutionary path from "optoelectronic hybrid" to "all-optical computing"? How will Rationalite Photonics play the role of an "enabler"?

Ms. Yang: All-optical computing is a long-term direction, but it will not replace electronic computing in one step. The more realistic path is to start with optoelectronic hybrid, letting light handle the parallel propagation and matrix transformations it excels at, while electronics continue to handle control, storage, and nonlinear activation.

As 3D optical structures, nonlinear optical devices, and packaging technologies gradually mature, optical computing will move from "auxiliary acceleration" to "core computing." In this process, Rationalite Photonics hopes to play the role of an ecosystem enabler. On one hand, Imp-Fab provides the manufacturing capability for complex 3D optical structures; on the other hand, we will also provide optical computing chip design, process design rules, device libraries, and system integration capabilities to help industry partners lower the R&D threshold.

Laser Manufacture News: What is the point of convergence between your company and the ultrafast optics industry?

Ms. Yang: Ultrafast optics technology is a critical underlying equipment enabler for the industrialization of 3D optical chips. Femtosecond laser nanoscale printing is not only an important micro-nano manufacturing method, but also a core process entry point for realizing complex 3D optical networks, spatially stacked structures, and optoelectronic hybrid packaging.

That said, when it comes to chip-level products and large-scale applications, existing ultrafast laser technology still has room for improvement in processing efficiency, long-term stability, and mass-production cycle time relative to our ultimate needs. Therefore, our relationship with the ultrafast optics industry is not simply that of a "equipment user" and a "process provider," but one of mutual traction and jointly defining a new track.

Going forward, we will continue to drive deep integration of ASOC with ultrafast optics manufacturing, inspection, and system validation. Our goal is to work with the ultrafast optics industry to open up an entirely new industrial track: 3D optical computing chip manufacturing.

Laser Manufacture News: Beyond ASOC, your technology also has enormous potential in areas such as quantum photonics and CPO. How is the company planning its commercialization path?

Ms. Yang: Our commercialization path is not a choice between "only making optical computing chips" and "only providing 3D manufacturing services." Optical computing represents the company's most important long-term strategic direction, but we will not stop at a distant vision.

Specifically, optical computing is our core main line in the medium to long term. But as we advance ASOC, we will also pursue 3D manufacturing services, joint R&D, and customized solution delivery around high-value directions such as optical quantum devices, 3D photonic crystals, and CPO optoelectronic co-packaging. This is not low-value-added contract manufacturing, but integrated capabilities spanning design, simulation, processing, and packaging validation for complex 3D optical field modulation structures.

Put simply, our long-term goal is a 3D optical computing chip platform. Our short- and medium-term path is to use advanced processes and solutions as the starting point, solving real problems and building industry trust first, then gradually moving toward scaled products.

Laser Manufacture News: China's optical chip industry currently faces issues such as technology homogenization and difficult mass production. How do you view the current state of the industry? What changes will the next three to five years bring?

Ms. Yang: The industry is currently in a phase of high development enthusiasm, with technological differentiation gradually becoming more pronounced. Demand from AI computing and data center interconnect is rapidly pulling optical technology forward. But homogenization does exist—many teams remain at the proof-of-concept stage, and few can truly cross from "laboratory results" to product-level validation.

Over the next three to five years, 3D optical computing will undergo three important changes. First, technology roadmaps will shift from "demonstrating that light can compute" to "proving that light is more valuable than electricity in specific scenarios." Second, competition will shift from single-point devices to 3D integration, optoelectronic co-packaging, and system-level collaboration. Third, commercial opportunities will first emerge in areas such as edge intelligent sensing, industrial vision, robotics, and low-power front-end computing.

Laser Manufacture News: What does the competitive landscape of the international optical chip industry look like? As a Chinese startup less than a year old, how is Rationalite Photonics building its technology moat?

Ms. Yang: The international optical chip industry has already formed a clear layered structure: top universities such as MIT and UCLA propose new concepts at the source; giants like Intel and NVIDIA drive large-scale engineering of silicon photonics and CPO around AI data centers; and a large number of innovative companies seek breakthroughs in niche directions.

The path Rationalite Photonics has chosen is to build differentiated advantages in the new dimension of 3D optical computing chips—not just making an optical device or a manufacturing process, but integrating 3D optical structure design, micro-nano manufacturing, algorithm training, optoelectronic packaging, testing and characterization, and application scenarios into a complete technology loop.

Our moat comes from three aspects. First, differentiated routing—focusing on chip-level 3D spatial optical field modulation, bypassing homogeneous competition in mature tracks. Second, full-stack capability—coordinating materials, processes, design, simulation, training, packaging and testing, and system applications. Third, supply chain security—insisting on multiple suppliers and multiple process paths for key links, avoiding building core capabilities on a single overseas equipment or platform.

Laser Manufacture News: As a young scientist moving from the frontlines of academia to the forefront of industry, what advice do you have for researchers embarking on hard-tech entrepreneurship?

Ms. Yang: My judgment is that the next five to ten years will be a very important window for entrepreneurship in China's optoelectronics and optical chip sectors. For researchers considering starting a business, my advice is: first, believe in original innovation; second, develop an engineering mindset early—don't just judge whether something can be made, but also think about whether it can be made reliably and at scale; third, truly step into industry scenarios and work with customers and supply chain partners to repeatedly refine real problems.