Investment Trends in Hard Tech: RISC-V, Autonomous Delivery, and Quantum Computing

The Value Inflection Point in Hard Tech

The global technology industry is undergoing a profound transition from business model innovation to underlying technology-driven growth. Recent investment events indicate that hard tech sectors—represented by chip architecture, quantum computing, and brain-computer interfaces—are moving from the proof-of-concept stage into a critical window of engineering and large-scale deployment. Capital market activities reflect not only recognition of the commercial potential of frontier technologies but also a shift in industrial competition from the application layer to the underlying infrastructure.

This article examines several typical cases—including IC design, autonomous delivery networks, and quantum computing platforms—to analyze the common patterns of technological evolution and explore the opportunities and challenges for developers.

RISC-V and AI Integration: A New Paradigm in Chip Commercialization

A chip design enterprise focusing on the integration of RISC-V architecture and AI recently went public, marking a milestone event in the industry. The company’s core strategy revolves around “RISC-V + AI,” with product lines covering human-machine interaction, multimedia processing, interconnection, and computing. These products target diverse scenarios such as consumer electronics, automotive electronics, robotics, and industrial control. As of its IPO prospectus period, the company had accumulated over 620 IP modules, more than 20 serialized RISC-V cores, and over 1,740 patent applications. It has also launched over 150 software-hardware co-designed products, serving 220 global customers.

Financially, the company achieved revenues of 1.752 billion, 2.025 billion, and 2.431 billion CNY over the past three years, respectively, with a three-year compound annual growth rate of approximately 17.8%. Its latest quarterly revenue grew 18.4% year-over-year, demonstrating steady growth.

The deeper significance of this case lies in the fact that RISC-V, as an open instruction set architecture (ISA), is evolving from an “alternative” to a “mainstream choice.” Combined with the demand for AI acceleration, it breaks the ceiling of traditional closed ISAs in customization, power efficiency, and ecosystem scalability. For developers, the maturation of the RISC-V ecosystem means lower architecture licensing barriers, greater flexibility in IP combination, and the possibility of instruction-level optimization for specific workloads (such as Transformer inference and edge vision). Notably, the company emphasizes the strategy of “IP module reuse” and “serialized cores,” which essentially reduces marginal development costs through platform-based design, establishing a positive flywheel of “architecture-toolchain-application.”

Autonomous Delivery Networks: The Scaling Inflection Point Driven by Data Flywheels

In the autonomous driving sector, an enterprise focusing on unmanned delivery vehicles recently announced the completion of an approximately $400 million Series E funding round, setting a record for the largest single financing in the field this year, with cumulative financing exceeding 1 billion CNY. The company’s core strategy is not purely pursuing L4 technical metrics but focusing on the operational efficiency of the “transport capacity network.”

Its technical path is distinct: using a vision-centric end-to-end large model as the perception and decision foundation, while constructing an industry-leading “operations-data-algorithm” data flywheel through a unified architecture combining cloud, edge, and vehicle models. In this closed loop, real delivery orders generate massive edge-case data, which is filtered, labeled, and sent back to the cloud for training and iteration. The updated models are then deployed to the fleet via OTA, creating a continuously evolving capability cycle.

From a commercial perspective, the company has pioneered overseas markets, achieving large-scale delivery of unmanned vehicles and related services in regions such as the Middle East, Europe, East Asia, Southeast Asia, and Australia. This model of “scenario first, data feedback, global replication” provides a crucial reference for the autonomous driving industry: technological leadership must match the granularity of commercial scenarios to achieve continuous iteration within controllable costs.

Quantum Computing and AI for Science: From Proof-of-Concept to Fault-Tolerant Engineering

New capital signals have also emerged in quantum computing. A startup founded just eight months ago has accumulated over 1 billion CNY in financing. Its technology roadmap is highly representative, focusing on a neutral atom quantum computing platform centered on the Ytterbium (Yb) atomic system, with “large-scale fault tolerance” as the top-level goal guiding system design.

Technologically, the company’s layout covers five layers: the physical base, system architecture, quantum error correction, real-time control, and the software stack. This “full-stack in-house R&D + error-correction first” strategy contrasts with the mainstream approach of “increasing qubit count first, adding error correction later.” Its rationale lies in the fact that neutral atom systems possess natural scalability (via optical tweezer arrays or optical lattices), longer coherence times, and uniform interactions between atoms, making them strong candidates for fault-tolerant quantum computing.

Meanwhile, the AI for Science sector continues to attract capital. One enterprise focuses on microscopic world models (such as molecular dynamics, protein conformation, and material property prediction), aiming to reshape the drug discovery process using generative models and geometric deep learning. The core value of this technology is transforming traditional “wet lab trial and error” into a “dry-wet closed loop.” It uses AI models for large-scale virtual screening and property prediction, followed by a small number of wet experiments to validate and feed back into the model. For developers, this signifies a shift in scientific computing from traditional numerical simulation paradigms to a new paradigm combining data-driven approaches with physical constraints.

The Underlying Shift in Hard Tech Investment Logic

Analyzing these funding cases reveals three notable underlying shifts in investment logic:

1. From “Single Technical Metrics” to “System-Level Engineering Capabilities”
Chip companies are no longer solely competing on process nodes or clock speeds; they emphasize IP reuse rates, toolchain completeness, and depth of scenario adaptation. Quantum computing companies no longer use “qubit count” as their only selling point; they give equal weight to error correction architecture, control system stability, and software stack friendliness.

2. From “Laboratory Ecosystems” to “Commercial Closed-Loop Ecosystems”
The capital market’s valuation anchor for frontier technologies has shifted from the number of papers and patents to the rate of improvement in unit economic models (e.g., cost per delivery, chip tape-out yield, drug pipeline milestones). This requires technical teams to possess the ability to translate research prototypes into deliverable products.

3. From “Single Acronyms (e.g., RISC-V, AI, QC)” to “Cross-Disciplinary Integration”
The coupling of RISC-V and AI, the linkage between quantum computing and classical control systems, and the combination of biological computing with deep learning models all indicate that future technical barriers lie in “connection” rather than “isolated breakthroughs.”

Technical Implications for Developers and Architects

For developers or technical decision-makers involved in such projects, the following capability dimensions will become particularly critical in the next three to five years:

1. Heterogeneous Computing Architecture Design
Whether dealing with custom operators under the RISC-V ISA or real-time control programs in quantum-classical hybrid systems, developers must possess cross-ISA, cross-hardware abstraction programming mindsets. Mastering hardware description languages (such as SystemVerilog) and high-level synthesis tools will help identify innovation points at the hardware-software boundary.

2. Engineering Large-Scale Data Closed Loops
A data flywheel is not simply “collect-train-deploy.” It involves complex engineering tasks like data quality assessment, label system design, scenario distribution balancing, and model drift monitoring. Teams should build comprehensive observability and experiment management platforms early on, ensuring every model update is based on quantifiable metrics.

3. Fundamental Respect for Underlying Physics
Quantum computing software development cannot stop at “calling a quantum SDK.” Understanding decoherence mechanisms, gate fidelity limitations, and resource estimation methods is necessary to design genuinely valuable quantum algorithm prototypes. Similarly, in AI for Science, developers need to establish foundational physical intuition to prevent models from “getting lost in statistical correlations.”

Conclusion: The Compounding Effect of Long-Termism

From the autonomous selection of chip architectures to the advancement of quantum fault-tolerant engineering, and from autonomous delivery networks to microscopic world simulations, a main thread runs through these seemingly discrete funding events: the technology industry is returning to the roots of “hardcore innovation.” Capital is willing to bet not on fleeting trends, but on systematic engineering that can build technical moats, accumulate data, and continuously reduce costs and increase efficiency across cycles.

For developers, this is both a challenge and an opportunity. As the correlation between technical depth and commercial value becomes tighter, talents who can navigate complex systems, understand underlying principles, and communicate across disciplines will be the primary beneficiaries of this hard tech evolution. Every upgrade in underlying architecture signifies new interfaces, new scheduling logic, and new space for performance optimization—this is the perpetual soil for technical creativity.