Artificial intelligence hardware startups are securing significant capital as investors shift focus toward specialized silicon capable of challenging Nvidia’s market dominance. Recent funding rounds for companies including FuriosaAI, Nuvacore, and d-Matrix underscore a broader trend: venture capital is flowing into firms designing custom AI inference chips to reduce energy consumption and operational costs for large language models.
## FuriosaAI Secures $80 Million for Next-Gen Chips
South Korean semiconductor startup FuriosaAI recently closed an $80 million funding round led by state-backed Korea Development Bank and investment firm Presto Ventures. According to the company’s official announcement, the capital will accelerate the mass production of its second-generation chip, “RUNGIE.”
The firm, founded by former Samsung engineers, intends to target the inference market—the phase where AI models process data and generate responses. By focusing on energy efficiency, FuriosaAI aims to provide a cost-effective alternative for data centers struggling with the high power demands of high-end GPUs.
## Nuvacore and d-Matrix Target Inference Efficiency
The capital influx extends to U.S.-based developers as well. Nuvacore, a startup focused on specialized AI compute architecture, recently raised capital to scale its engineering team and refine its chip design. Similarly, d-Matrix, which uses “digital in-memory computing” technology, has continued to draw interest from strategic investors, including Microsoft’s M12 venture fund.
These companies share a common technical objective: moving away from the general-purpose architecture of standard GPUs. By optimizing silicon specifically for the matrix multiplication tasks central to transformer models, these firms seek to lower the “total cost of ownership” for companies deploying AI at scale.
## Market Context: Beyond the Nvidia Monopoly
The current investment cycle is driven by the industry’s struggle to secure enough hardware to meet demand. While Nvidia remains the primary supplier of training chips, the market for inference—which is expected to account for the majority of AI compute spending—remains fragmented.
| Company | Focus Area | Key Technology |
| :— | :— | :— |
| FuriosaAI | Inference | High-efficiency ASIC |
| d-Matrix | Inference | Digital in-memory computing |
| Nuvacore | AI Compute | Specialized architecture |
Industry analysts note that while these startups face significant challenges in software ecosystem development, the sheer volume of capital indicates that hyperscalers and enterprise clients are actively seeking supply chain diversification.
## Strategic Outlook
The long-term viability of these startups depends on their ability to integrate with existing software stacks like PyTorch and CUDA. As these firms move from prototype to production, the focus will shift from fundraising to manufacturing yields and software compatibility. Investors are betting that the persistent shortage of high-performance chips will provide enough runway for these specialized players to capture market share in the inference sector, which is projected to grow significantly as generative AI models move from training to widespread commercial deployment.
Related reading