AI Venture Funding Surges: CuspAI, Onyx, Eliyan Raise Big

A New Wave of AI Capital Targets Infrastructure Bottlenecks

The AI investment boom is no longer just about building bigger models — it's increasingly about solving the physical and operational constraints that threaten to slow the industry down. A fresh round of venture funding illustrates this shift, with billions of dollars flowing into startups tackling everything from chip materials and data-transfer bottlenecks to enterprise AI security and industrial automation.

Bezos-Backed CuspAI Targets Materials Science

Among the most notable deals is CuspAI, an AI startup reportedly valued at $2.6 billion with backing from Jeff Bezos. The company is partnering with chip giants Nvidia and Meta to develop new materials aimed at addressing what has become an unglamorous but critical obstacle for the AI industry: the physical constraints of chip manufacturing and materials innovation 1. As AI infrastructure scales, seemingly minor engineering challenges — from thermal management to novel semiconductor materials — are emerging as serious chokepoints, and CuspAI's approach signals growing investor appetite for startups solving these foundational problems rather than just building applications on top of existing infrastructure.

Chip Bottlenecks Draw Investor Attention

That same infrastructure focus is visible in Eliyan's $145 million raise at a $1 billion valuation. The company is working specifically on easing data-transfer bottlenecks between AI chips inside data centers, a problem that has become increasingly urgent as AI workloads demand faster and more efficient chip-to-chip communication 7. Together, CuspAI and Eliyan point to a broader investment thesis: as AI compute scales, the plumbing connecting and constructing chips is becoming as valuable a target for venture capital as the models themselves.

Enterprise Security and Industrial Applications

Beyond hardware, investors are also funding the governance layer of AI deployment. Onyx Security raised $113 million in a Series B round — bringing its total funding to $153 million — to help enterprises control and monitor AI agents operating within their systems, reflecting mounting concern over the risks of autonomous AI acting inside corporate environments 2. Meanwhile, Arrakis emerged from stealth with $38 million to bring agentic AI into aerospace, energy, logistics, and manufacturing, betting that industrial sectors represent an underserved market for autonomous AI tools 3.

Signs of Strain Elsewhere in the Market

Not all the news signals unchecked momentum. DeepSeek reportedly paused a roughly 10 billion yuan fundraising round as it advances IPO plans, amid scrutiny and sensitivity around Nvidia chip access — a reminder that geopolitical and regulatory pressures continue to complicate AI fundraising, particularly for Chinese firms 4. Separately, the hedge fund Situational Awareness has been unwinding public AI-related stock positions, including an Anthropic stake, with Citadel purchasing the bulk of those holdings, as the fund seeks new capital following losses 6. These developments suggest that even as fresh venture money pours into infrastructure and enterprise AI startups, parts of the broader AI investment ecosystem are experiencing volatility and recalibration.

What It Means

Taken together, these deals show venture capital increasingly chasing the unglamorous but essential layers of the AI stack — materials, chip interconnects, and agent governance — even as broader market signals suggest growing caution among some institutional investors navigating an increasingly crowded and scrutinized field 5.