Euclyd's €200M Series A and Cornelis's $205M Signal EU Push to Break NVIDIA Dominance

EU-Anchored AI Silicon Gets Credible Funding

Euclyd, a Netherlands-based semiconductor startup, closed a Series A exceeding €200 million (approximately $231 million) on September 15, co-led by Samsung, Somerset Capital Partners, EQT's Scaleup Europe Fund, and Innovation Industries. The round positions Euclyd as a funded alternative to NVIDIA in enterprise, sovereign, and hyperscale AI markets.

For enterprise buyers, this matters because it validates a non-U.S., EU-jurisdictional AI compute stack. Euclyd is pre-commercial, but the capital and Samsung's involvement mean the company will deliver silicon and systems, not remain at concept stage. EU public sector buyers, national clouds, and regulated industries now have a credible path to AI infrastructure that satisfies data residency and export control requirements without relying exclusively on U.S. vendors.

The immediate impact is strategic, not operational. CIOs should add Euclyd to multi-year capacity planning scenarios and vendor diversification roadmaps. For organizations with "avoid single vendor lock-in" or "EU-sovereign AI" mandates, this round justifies budget allocation for pilot programs once Euclyd reaches commercial deployment, likely within 18-24 months based on typical Series A timelines.

Cornelis Networks Secures $205M to Challenge NVIDIA in AI Interconnects

Cornelis Networks, which builds high-performance networking for multi-GPU clusters, raised $205 million led by IAG Capital Partners. The company develops interconnect technology that enables AI chips to communicate at the scale required for clusters exceeding 10,000 GPUs—directly competing with NVIDIA's InfiniBand and Spectrum Ethernet offerings.

This capital event changes the calculus for buyers designing large on-premises AI infrastructure. NVIDIA has historically dominated both the accelerator and networking layers of AI data centers, creating cost and flexibility risks. Cornelis's funding validates an alternative interconnect stack with the capital to deliver long-term roadmaps and enterprise-grade support.

For enterprises evaluating on-premises clusters, this introduces a negotiating lever. Buyers can now credibly threaten to split the stack—procure GPUs from one vendor and networking from another—which typically drives 15-20% discounts in enterprise negotiations. More importantly, it reduces concentration risk. A single-vendor outage, licensing change, or supply constraint no longer paralyzes the entire infrastructure.

CIOs should update RFPs to include Cornelis as a networking option and model the total cost of ownership against full-NVIDIA stacks. The performance and support data will determine viability, but the vendor now has the resources to compete on both.

EasyStack's EAF Platform Targets Private AI Cloud Buyers

EasyStack launched EAF, an enterprise AI infrastructure software platform marketed as an "AI-native cloud foundation," with general availability set for September 30, 2026. The platform competes against hyperscaler AI offerings (AWS Bedrock, Azure AI, Google Cloud) and private cloud platforms with AI add-ons (Red Hat OpenShift, VMware Tanzu).

EAF is relevant for mid-to-large enterprises that want to run AI workloads on private infrastructure rather than rent GPUs from public clouds. The value proposition centers on long-term operational cost savings and control over data residency—critical for regulated sectors including finance, healthcare, and government.

The September 30 GA date is specific enough to influence Q4 planning. Buyers piloting on-premises AI should align proof-of-concept timelines and budget approvals with that availability window. The key questions are integration with existing Kubernetes and OpenStack estates, and pricing relative to hyperscaler managed services. EasyStack has not published pricing data, so buyers need to request direct quotes and model total cost of ownership over 3-5 year periods.

The risk is vendor viability. EasyStack is a regional player, primarily in APAC. Enterprises outside that geography should validate support SLAs, roadmap commitments, and financial stability before committing to a platform that sits below critical AI workloads.

What This Means for Enterprise AI Procurement

The combined $436 million flowing into Euclyd and Cornelis in a single week signals that capital markets are funding credible alternatives to NVIDIA's stack. For buyers, this creates three immediate actions:

First, update vendor risk assessments. Single-vendor concentration is now a choice, not a necessity. Multi-vendor strategies have support from funded competitors with roadmaps.

Second, revisit procurement timelines. Euclyd and Cornelis are not ready for production deployment today, but they will be within 18-24 months. Buyers planning infrastructure refreshes in 2027-2028 should include these vendors in evaluation cycles starting now.

Third, use these developments in negotiations. The existence of funded alternatives gives procurement teams leverage in pricing discussions with incumbent vendors, even if the enterprise ultimately stays with NVIDIA or hyperscalers. Vendors respond to credible competitive threats, not abstract ones.

The broader pattern is clear: AI infrastructure is fragmenting. The NVIDIA-everywhere model that dominated 2023-2025 is giving way to specialized vendors in silicon, networking, and platform software. Enterprises that build flexibility into their architectures now will have more options—and better pricing—in 2027.