Astera Labs is a fabless semiconductor company focused on purpose-built connectivity for rack-scale AI infrastructure. It designs chips, modules and software that move data between GPUs, CPUs, memory, networking and storage inside AI servers, racks and clusters. The company relies on third-party manufacturing and supply-chain partners, while keeping its value creation centered on architecture, silicon design, software, validation and customer integration.
Revenue comes mainly from sales of connectivity semiconductor products and related modules into AI and cloud data center supply chains. In Q2 2026, Astera reported revenue of $392.4 million, up 27% sequentially and 104% year over year. First half 2026 revenue was $700.8 million, compared with $351.4 million in the first half of 2025. The model has high gross margin characteristics for a fabless chip company, with Q2 2026 GAAP gross margin of 73.3% and GAAP operating margin of 22.7%.
The main product and service categories are
- PCIe and CXL connectivity: Aries smart retimers, signal conditioners and smart cable modules support high-speed data movement and signal integrity in AI servers and racks.
- Ethernet connectivity: Taurus smart cable modules target high-speed networking links in cloud and AI infrastructure.
- Memory connectivity: Leo CXL memory connectivity controllers address memory expansion and pooling needs as AI systems become more data intensive.
- Fabric switching: Scorpio smart fabric switches extend Astera from point-to-point connectivity into rack-scale AI fabric infrastructure. Management said Scorpio was expected to become the company’s largest product family in Q3 2026, one quarter ahead of prior expectations.
- Software and system management: COSMOS provides fleet management, diagnostics, monitoring, debugging and tuning across Astera connectivity products, giving customers a software layer around its hardware.
- Custom connectivity and reference designs: Astera offers custom solutions and, in October 2026, announced Orion, a compute-agnostic rack-scale reference architecture for AI inference factories targeted for availability through OEM and ODM partners starting in Q1 2027.
Astera’s sales motion is built around design wins with hyperscalers, AI platform providers, OEMs, ODMs and manufacturing partners. Revenue is often billed to contract manufacturers, distributors or supply-chain intermediaries rather than the final cloud or AI infrastructure customer. This makes customer concentration and geographic billing data important to interpret carefully. In Q2 2026, four customers each represented at least 10% of revenue, with disclosed shares of 29%, 25%, 15% and 13%. Some of those customers are manufacturing partners buying on behalf of end customers.
The company’s market position is that of a specialist in AI infrastructure connectivity rather than a broad merchant semiconductor supplier. Its products are tied to the shift from server-level acceleration toward rack-scale and cluster-scale AI architectures, where bandwidth, latency, signal integrity and power efficiency become limiting factors. Astera collaborates with major AI ecosystem companies including AMD, Arm, Intel and NVIDIA, and with Taiwan ODMs including GIGABYTE, Ingrasys/Foxconn, Inventec, Quanta Cloud Technology and Wiwynn.
China is a meaningful billing location, but not a pure measure of end-market demand. In Q2 2026, China-billed revenue was $152.7 million, about 39% of total revenue. In the first half of 2026, China-billed revenue was $242.3 million, about 35% of total revenue. These figures include billing addresses for manufacturing partners and distributors. Astera also identifies China-related export controls, tariffs, trade restrictions and adverse policy changes as business risks.
Astera’s competitive advantages come from focus, integration and ecosystem positioning. Its narrow emphasis on AI connectivity lets it move across PCIe, CXL, Ethernet, optical interconnects, fabric switching and rack-level designs. The COSMOS software layer supports diagnostics and fleet-level management, which matters for hyperscale deployments. Its Taiwan engineering expansion and cloud-scale interoperability lab also place the company near key ODM partners that build AI systems for global customers.
Direct competitors include Credo Technology Group in high-speed connectivity for AI and cloud data centers. Broader competition includes Marvell, Broadcom, Rambus and internal silicon or system-design teams at hyperscalers, depending on the product category. Compared with Credo, Astera has positioned itself more explicitly around an Intelligent Connectivity Platform spanning chips, cable modules, software and rack-scale fabric products. Compared with larger peers such as Broadcom or Marvell, Astera is more concentrated, with higher exposure to AI connectivity growth but also greater dependence on a smaller set of customers, products and deployment cycles.