← Portada | Mercado

Positron AI raises $875M for HBM-free inference appliances built on consumer-grade memory

2026-09-10 · Fuente:SiliconANGLE

Positron AI raises $875M for HBM-free inference appliances built on consumer-grade memory

NEA, Atreides, Valor, Andra, SemiAnalysis Capital and Netscape co-founder Jim Clark lead $875 million round for Positron AI's consumer-memory inference hardware.

Positron AI Inc. has raised $875 million in new funding as it pushes to accelerate artificial intelligence inference with hardware built around consumer-grade memory rather than HBM. The round, announced Thursday, was led by NEA, Atreides Management, Valor Equity Partners, Andra Capital, SemiAnalysis Capital and Netscape Inc. co-founder Jim Clark, with more than a dozen institutional investors joining alongside them.

The company develops inference appliances — systems purpose-built for running deployed AI models — and its defining design decision is the absence of HBM, the high-bandwidth memory that has become a fixture of mainstream AI accelerator architectures. By turning instead to consumer-grade memory, Positron is betting that inference can be sped up without the memory type that dominates leading accelerator designs, a positioning that sets the startup apart in a crowded AI hardware field.

The investor roster blends venture capital, investment managers and a recognizable name from computing history. NEA anchors the syndicate alongside Atreides Management, Valor Equity Partners, Andra Capital and SemiAnalysis Capital. Jim Clark's participation draws particular attention: the Netscape Inc. co-founder is placing a bet on an AI chip upstart, adding decades of entrepreneurial pedigree to the cap table. The breadth of the syndicate — more than a dozen institutional investors in total — signals substantial appetite for alternatives to incumbent approaches to AI compute.

The raise lands as competition intensifies to supply the compute behind inference, the phase in which trained AI models are put to work. With $875 million in hand and a memory strategy that departs from the HBM-centric norm, Positron now has the capital to press its argument that inference can run faster on appliances built from consumer-grade components.

Etiquetas: Positron AI · Jim Clark · AI Inference · HBM · Venture Funding


Lecturas relacionadas


萬安算交所 AIXX · AIXX · Exchange verificado · Ir al mercado