The global semiconductor market is on track to exceed $1.3 trillion in revenue by 2026, according to IT research firm Gartner, marking what the firm calls the highest growth rate in two decades. That surge is fueling a wave of investment into startup chip companies, with billions in capital flowing to firms developing new architectures for AI and high-performance computing. Here is a look at ten of the most notable semiconductor startups of 2026 so far.
Gartner forecasts semiconductor revenue will grow 64 percent annually in 2026, with memory revenue expected to triple. “Amid high demand for AI processing, data center networking and power, and memory price inflation, the semiconductor industry is projected to achieve a third consecutive year of double-digit growth in 2026—a milestone that shows the sector’s central role in the AI technology stack,” said Rajeev Rajput, senior principal analyst at Gartner.
The scale of the opportunity is drawing both established players and newcomers. For startups, the challenge is not just building a better chip but getting it into production and winning customer commitments before the next wave of competitors arrives. Some have already secured hundreds of millions in funding and, in a few cases, signed contracts worth over a billion dollars.
Related: Microsoft’s AI Bug Disclosure Signals New Era For Security
Axelera AI and Cornelis: edge and networking
Axelera AI, based in the Netherlands, develops AI acceleration hardware for computer vision and generative AI at the edge and in data centers. Its Europa and Metis platforms focus on staying within real-world power and thermal limits. The startup has more than 500 customers across telecommunications, aerospace and enterprise, and it announced a $250 million funding round this year. Since its launch in 2021, it has raised over $450 million in equity, grants and venture debt.
Cornelis, headquartered in Wayne, Pa., takes a different approach by targeting networking for AI workloads. Its technology aims for lossless, congestion-free networking to reduce training time and improve compute utilization. In June, the company launched the Cornelis CN5000, a 952-node supercomputer cluster built with Dell PowerEdge servers and Intel Xeon processors. It also expanded its partner ecosystem this year, adding federal systems integrators and distribution channels. Whether these networking-focused startups can hold their own against incumbents like Broadcom and Nvidia’s networking division remains an open question, but the demand for specialized infrastructure is strong enough that multiple approaches are getting funded simultaneously.
d-Matrix, Etched, and Fractile: inference at scale
d-Matrix, based in Santa Clara, Calif., focuses on low-latency AI inference for data centers. Its product lineup includes inference accelerators, JetStream networking accelerators, Aviator software, and a rack-scale offering called SquadRack. In April, the startup acquired GigaIO’s data center business, a systems engineering organization with deep
Etched emerged from stealth in June by unveiling its Sohu chip and announcing over $1 billion in signed customer contracts. The San Jose, Calif.-based company develops custom chips optimized specifically for transformer models. The company says it has a path to gigawatt-scale operations by 2027.
Related: The technology used to protect a country’s border
Fractile, a London-based startup, is building chips that use in-memory compute, allowing calculations to run directly in computer memory. In May, the startup raised $220 million in a Series B round to fund development and commercialization of its upcoming AI inference chips. News reports indicate that Anthropic is in discussions with Fractile about purchasing those chips when the hardware becomes available in 2027. In February, Fractile announced plans to invest $135 million to expand its U.K. operations, including a new hardware engineering facility.
Lightmatter, MatX, and NextSilicon: photonics and alternative architectures
Lightmatter, based in Mountain View, Calif., owns Passage, a 3-D-stacked silicon photonics engine, and Guide, a VLSP light engine designed to connect thousands to millions of processors. The startup has raised a total of $850 million and now carries a valuation of $4.4 billion. Its technology aims to eliminate data bottlenecks in AI and high-performance computing workloads by using light rather than electrical signals to move data.
MatX, also in Mountain View, built an LLM chip called MatX One that delivers high throughput and low latency. The startup recently announced a $500 million Series B round to accelerate production of its flagship processor. MatX claims its chips are up to 10 times better at training large language models compared with Nvidia’s GPUs. The company was founded by two former Google engineers, Reiner Pope and Mike Gunter, and plans to start shipping chips in 2027.
Related: Microsoft invests $2.5 billion in new AI division
NextSilicon, based in Israel, builds computing infrastructure for algorithmically complex workloads. Its Maverick-2 accelerator chip uses a runtime reconfigurable dataflow architecture and claims up to 10X performance over leading GPUs at less than half the power, with no requirement to rewrite existing applications. The company has raised over $300 million. In June, it unveiled plans to productize its Arbel RISC-V core processor into a 64-core and a 128-core enterprise-grade processor, expected to be available in early 2028.
Tenstorrent and Xsight Labs: RISC-V and connectivity
Tenstorrent, led by chip architect Jim Keller, builds RISC-V-based AI processors and systems. The Santa Clara, Calif.-based company licenses its TT-Ascalon RISC-V CPU and Tensix AI cores to chip designers including Samsung and LG. Tenstorrent has raised over $1 billion and operates throughout North America and Asia. Last month, it launched TT-Ascalon S, a new compute-dense RISC-V CPU designed for agentic AI workloads.
Xsight Labs is a fabless semiconductor startup focused on intelligent connectivity for hyperscalers, edge computing and AI data center networks. The company is led by CEO Yossi Meyouhas and operates in a segment that has seen increasing demand as data center architectures shift to handle heavier AI traffic loads.
