Reporter |
Editor | Wen Shuqi
On July 20, the four-day 2026 World Artificial Intelligence Conference (WAIC) in Shanghai wrapped up. Looking back at this year’s industry gathering, it wasn’t just about the buzz around large models, embodied intelligence, or domestic computing power. Two other “young” forces in the exhibition halls were fueling our imagination for the future.
Inside one hall, a player wearing a lightweight brain-computer interface device controlled the character in Black Myth: Wukong—moving, fighting, casting spells—all with their mind, hands free from any controller.
Not far away, quantum computing companies brought their full-scale machines to the show floor. Neutral-atom quantum computers were hooked up alongside CPUs, GPUs, and supercomputing clusters, trying to become a type of heterogeneous computing power that can be called upon.
Though their tech principles couldn’t be more different, and both are still in early commercialization stages, they’re being drawn onto the same map of future industries.
In China’s 15th Five-Year Plan outline, quantum technology and brain-computer interfaces are both listed as future industries for forward-looking deployment, with plans to explore diverse tech routes, typical application scenarios, and viable business models.
Beyond policy, market changes are becoming more concrete. BCI companies aren’t just showing off tech demos anymore—they’re pushing devices into broader consumer and industry settings. Meanwhile, quantum computing is catching investors’ eyes. Just on the neutral-atom route, the number of related domestic companies reportedly grew from one to 21 in half a year.
What WAIC showcased this year is exactly this picture: these future technologies trying to step out of labs and into application scenarios.
Brain-Computer Interfaces Hunt for New Scenarios
The “mind-controlled” demo of Black Myth: Wukong came from YanShan Brain, a subsidiary of YanShan Technology, showing off its next-gen brain-controlled game interaction system.
A staff member on-site explained to NUPIAO that as a non-invasive BCI product, the system captures the player’s visual reactions and brain neural signals in milliseconds—no physical operation needed at all.
According to them, this system’s whole unit is a hundred times lighter than similar top-tier devices, achieving the same brain-control performance with just about 1% of the volume.
Mind control doesn’t mean the device reads thoughts like “I want to move forward” or “I want to attack.” The staff member told NUPIAO that flashing dots on the screen correspond to commands like forward or backward. When a player stares at a dot, their visual cortex brainwaves change, and the system recognizes and converts that into an action. It takes about 10 minutes for an average person to get the hang of it.
However, the system can only execute one command at a time (like moving forward or attacking, not both). The team hasn’t ruled out exploring a mode that triggers commands directly through motor imagery in the future.
Mind-controlled gaming feels like sci-fi, but beyond the booth, the signals from this industry are a lot more grounded. Brain-computer interfaces are moving past pure tech demonstrations and starting to solve real problems in various real-world scenarios.
Embodied intelligence is a key crossover area for BCI deployment. As robots gain execution abilities, understanding human intent becomes a critical bottleneck.
At this year’s WAIC, BrainCo released what it calls the industry’s first integrated brain-controlled robot AI research platform, exploring human-robot collaboration through intent interaction.
Simply put, it lets people directly command robots through brain signals. After a user puts on an EEG cap, the system captures task-related brain signals in real time, uses algorithms to identify intent-related patterns, and converts them into specific commands for the robot to execute.
At the booth, a robotic arm completed actions like picking up a water cup or grabbing an apple after identifying user intent. According to staff, even researchers without a BCI background can learn to control a robot with their “mind” in about 10 minutes. The platform already supports connecting to third-party devices like humanoid robots, robotic arms, and robot dogs.
In BrainCo’s vision, BCI handles intent decoding and output, AI boosts decoding and task decomposition, and embodied intelligence handles physical execution. The trio—Neuro-Embodied-AI (BCI + embodied intelligence + AI)—works together to form a closed loop from collection to decoding to execution.
After robots understand what humans want, they still need to learn how to do the work—and what they lack most is data. OYMotion’s exhibit at WAIC focused on this.
Their gForce Ultra EMG wristband features 8-channel dry EMG electrodes and an IMU posture sensor with a 1000Hz sampling rate, capturing forearm muscle contraction and force patterns.
According to them, regular vision systems only record what a robot does. EMG data fills in the gap on how a person exerts force and when they make micro-adjustments—exactly the missing info for training robotic arms and dexterous hands to handle objects gently.
Additionally, OYMotion showcased collection devices like the OB series smart EEG machine and CeRelax EEG headband at WAIC, and opened up SDK interfaces. Their approach is to supply standardized data tools for embodied intelligence and research markets.
If BCI is on the hunt for new scenarios, quantum computing—another forward-looking industry—is thinking about how to plug into the real-world computing system.
Neutral-Atom Quantum Computing Hits Its Stride
“It only took half a year for neutral-atom quantum computing companies to go from one to 21.”
A technician from a Shanghai-based quantum computing company put it succinctly at WAIC, describing the heating up of this tech route.
Since 2026 began, multiple neutral-atom quantum computing companies have completed funding rounds. AtomMatrix, founded in 2025, closed three funding rounds in half a year, including a several-hundred-million yuan Series A round in May. Liangyi Wanxiang and Shanghai Taiyi Liangsheng also recently secured hundreds of millions in funding.
Neutral atoms have gained attention lately because they show potential for scaling up physical qubit counts.
Quantum computers try to build physical qubits by manipulating basic particles (like Josephson junctions in superconducting routes or ions in ion trap routes), then use error correction to achieve quantum computing. The neutral-atom route opts for uncharged atoms like rubidium or cesium.
The technician explained that the quantum computing system first cools atoms with lasers, then captures and moves them with optical tweezers, rearranging them into defect-free atom arrays. Neutral-atom quantum computing systems have inherent identity and full connectivity, giving them a natural edge in large-scale fault-tolerant quantum computing.
Notably, lasers run through almost the entire neutral-atom quantum computing process. This means China’s accumulated expertise in lasers and precision optics becomes a real boost for developing the neutral-atom route.
Zhongqi Wuliang, drawing on the Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences’ long-term work in lasers, precision optics, and atomic clocks, told NUPIAO that devices like cold atom setups and precision optics share strong technical overlap with neutral-atom quantum computing in parts of lasers and optical paths.
Moreover, neutral-atom physical qubit counts are easier to scale up. Li Ziliang, technical director of BuChou Quantum, a quantum computing company incubated by Fudan University, told NUPIAO that ion trap gate fidelity is usually high, but as the number of ions increases, control and system expansion become more difficult.
But more qubits don’t directly equal a more powerful computer. Since qubits are extremely susceptible to environmental disturbances and “decoherence” (losing their quantum state), many physical qubits are needed for error correction to form a few logical qubits for consistent and accurate computation.
That’s why, despite varying data disclosures from quantum computing companies, they all admit that even correcting one logical qubit is no small feat. It also explains why the full-scale quantum computers concentrated at WAIC this year all anchored their deployment first on “quantum-supercomputing integration”—collaborating with classical computing.
Zhongqi Wuliang released the neutral-atom quantum computer “Qinghe No.1” designed for computing centers and showcased a large-scale fault-tolerant quantum computing system architecture. AtomMatrix partnered with Suanfeng Information to seamlessly integrate its MatriQ neutral-atom quantum computer with the Shanghai Cube intelligent computing cluster via photoelectric connection, forming a super-hybrid training device where the QPU can run arrays of up to 2310 defect-free physical qubits. BuChou Quantum’s exhibited “Liangchou No.1” integrates a neutral-atom QPU, GPU cluster, and classic CPU. TuringQ, taking the photonic quantum route, also launched QAgent, aiming to break down tasks through intelligent agents and call on quantum and classical computing power.
According to NUPIAO, the algorithm pathways for connecting quantum computers to supercomputers are diverse, but the basic idea is for classical computing to handle most routine processes, while quantum computing takes on the “critical 1%” tasks that classical computing finds difficult or extremely costly.
Quantum-supercomputing integration can’t yet prove quantum computing has commercial value, but it offers the industry a more practical path forward: first, plug the QPU into the existing computing system, let companies and research institutions call on real machines and validate problems, and continuously improve hardware, algorithms, and software interfaces through real tasks. As real machines, computing power, and industry needs start to converge, quantum computing is getting its first clear path from research breakthrough to industry validation.
