Inside the Future of Open Infrastructure: Insights from our CEO, Ian Evans at OCP Summit 2025
Oct 23,2025
By admin
Inside the Future of Open Infrastructure: Insights from our CEO, Ian Evans at OCP Summit 2025
The halls of the San Jose Convention Center were populated with unmistakable cutting-edge open compute chassis, sophisticated cooling solutions and advanced power management solutions, a vibrant backdrop for the latest advancements in data center infrastructure. Thousands of engineers, architects, and innovators converged once again for the Open Compute Project (OCP) Global Summit 2025, the annual heartbeat of the open-hardware movement.
Among them was Ian Evans, CEO of Karios whose ipad was filled not with slides or marketing fluff but with firmware specs, rack diagrams, and power telemetry notes. What stood out this year wasn’t just the hardware—it was the conversation. From Redfish schema updates to AI-ready cooling solutions, the OCP floor felt like standing in the middle of a live blueprint for the future of infrastructure.
Redfish Telemetry: From Metrics to Meaning
One of Ian’s biggest takeaways came from his conversations with the DMTF, the official standards body behind the Redfish specification, which is a standard designed to deliver simple and secure management for converged, hybrid IT and the Software Defined Data Center (SDDC).
Redfish is now the standard for securely managing most servers through their Baseboard Management Controller (BMC). “With these new updates, we will continue to see improved control and observability into common server platforms. This will in turn give operators more control and visibility of their infrastructure.,” Ian noted. “It’s no longer just data collection, it’s actionable observability based on very specific information coming from these systems.”
The new Redfish telemetry model adds more sensors, deeper granularity, and real-time streaming across compute, storage, and power components. It’s a significant leap toward full-stack visibility, exactly what’s needed for AI and edge deployments where microseconds and milliwatts both matter.
DMTF’s growing influence was visible everywhere. Redfish, SPDM, and PLDM weren’t just logos on slides they were running inside production systems. These standards are now the connective tissue enabling open, vendor-neutral interoperability across servers, switches, and power gear.
Securing the Foundation: SPDM and Firmware Authenticity
Security at the firmware layer dominated hallway conversations. With vendors like NVIDIA and Microsoft collaborating on the SPDM 1.2 standard, a new level of trust is emerging for device-level communication.
SPDM now ties directly into Hardware Roots of Trust (HRoT), improving how devices authenticate firmware updates and configuration changes. As Ian put it, “Immediate firmware activation without a reboot—simple change, massive impact. With threats emerging so quickly, HRoT becomes a very important tool to ensure complete platform compliance and security.”
That feature alone can cut maintenance windows dramatically for hyperscale environments. Combined with new update agents and vendor-agnostic PLDM support, firmware management is finally escaping the era of “dumb binaries and manual updates.”
The ASPEED AST2700 BMC was another highlight. Its DDR5 memory controller, integrated HRoT, and OCP 3.x-compliant form factor signal a coming wave of decoupled BMC designs—hardware that’s modular, interoperable, and secure by default.
Karios Advantage #2: Security Beyond Patching
If one topic dominated every corridor, it was power and cooling. AI workloads are driving rack densities and power envelopes to extremes and OCP 2025 put those challenges on full display.
Dell’s edge-based solutions, Submer’s single-phase cooling systems, and Mitac’s rack-scale designs all showcased innovation aimed at making 2 megawatts of cooling at 3 °C approach temperature difference (ATD) completely achievable.
“Power and cooling are dominating. It’s the conversation everywhere,” Ian observed. “The way the industry talks about sustainability has shifted from slogans to schematics.”
Liquid-cooled GPU racks, OCP 3.0 cards with dual 25G and 400G ports, and smarter CDU (Cold Distribution Unit) management are pushing the boundaries of thermal autonomy. Liquid cooling is no longer a want, rather a need.
It’s About the Network
Beyond management, there were many notable network advancements such as new 800GB interfaces, Marvell liquid cooled switches and new achievements when it comes to breaking multi-terabit barriers. RDMA over Ethernet (ROCE) took center stage and has emerged as a key enabler of ultra-low-latency fabrics.
“We’re seeing open standards evolve from compliance checklists into innovation platforms,” Ian said. “That’s how open hardware matures—when standards enable creativity, not constrain it.”
Hardware Innovation on Full Display
Walking through the OCP floor felt like a guided tour of the near future:
- Axiado SCM3080-MT: is a compact, pluggable module leveraging FPGA technology to integrate essential networking and security features: BMC, TPM, AI/ML acceleration, and NC-SI. This solution streamlines infrastructure deployments and unlocks enhanced performance for modern networking environments.
- Mitac’s liquid-cooled rack solutions couples with a new and advanced BMC framework utilizing AMI’s OneTree framework: AMI’s OpenBMC-based next-generation open-source BMC manageability firmware solution that is optimized for scale-out AI/HPC optimized infrastructure.
- AMD Instinct GPU boards and Blackfield GPUs: rAMD Instinct GPUs and Blackfield GPUs deliver significant performance gains for HPC, AI, and data science. AMD Instinct provides powerful compute capabilities, while Blackfield GPUs offer ultra-high bandwidth for faster data processing and scalable clusters. This combination accelerates AI training, data analytics, and HPC simulations, enabling organizations to tackle complex computational challenges and drive innovation with a future-proof infrastructure.
- Broadcom Tomahawk 6 enabled switches: The demands of modern AI are pushing the boundaries of interconnect performance. Broadcom Tomahawk 6 represents a significant leap forward, delivering unparalleled bandwidth and efficiency for accelerating AI workloads. This report explores the key capabilities of Tomahawk 6, highlighting its advantages for researchers, developers, and organizations driving innovation in AI.
- AMD Helios and Facebook/Meta OCP racks: dual-width, modular, and breathtakingly engineered. AMD Helios and Facebook’s Helios are vital for accelerating AI & HPC. AMD Helios is an integrated solution combining CPUs and GPUs, while Facebook’s Helios is a custom supercomputer architecture. Both significantly speed up AI training and reduce costs, democratizing access to powerful computing for research and innovation.
Every product reflected a larger theme, openness, modularity, and intelligence at every layer. The move away from proprietary stacks to interoperable ecosystems is no longer a trend; it’s the baseline.
Power, Water and Land: The Sustainability Triad
In one hallway conversation, Ian joked, “Treat ours in the cube nicely!” It became a metaphor for mindful infrastructure design. Power, water, and land the triad shaping tomorrow’s data centers were discussed with urgency.
Vendors highlighted DC power distribution, CAN and LTPI protocols for smarter PSU management, and AI-assisted telemetry that predicts cooling needs before problems occur.
These insights resonate deeply with the philosophy behind Karios Cube and PowerLink, embedding sustainability directly into the infrastructure fabric rather than bolting it on later.
From Open Compute to AI-Native Hyperconvergence
As the conversations at OCP shifted from hardware to orchestration, one thing became clear: the next decade of infrastructure will be AI-native.
Karios embodies this shift, built on Bhyve, and ZFS, it integrates security, telemetry, and power analytics at the hypervisor layer. No per-core licensing, no bloat, just intelligent hyperconvergence engineered for transparency and efficiency.
The alignment between OCP’s direction and Karios’ vision couldn’t be clearer: open standards, lightweight architectures, and sustainability-by-design.
“The market has been asking for a new lightweight hypervisor platform —and we’re delivering,” Ian noted. “We’re redefining value in the HCI space, from edge to core.”
Collaboration Over Competition
What truly sets OCP apart isn’t just the technology, it’s the spirit. Engineers from competitors share specs, debug each other’s code, and debate thermal dynamics over coffee.
As Ian summarized, “Innovation isn’t just about faster chips, it’s about shared progress, better standards, and sustainable growth.”
That sense of collaboration underpins everything Karios stands for: open ecosystems, transparent design, and a commitment to making infrastructure smarter, not heavier.
Looking Ahead — The Road to OCP 2026
As the lights dimmed on the show floor, one takeaway stood out: the future of data centers is open, modular, and intelligent. The convergence of AI, power management, and firmware security is creating a new baseline for trust and efficiency.
With partners like Axiado, Mitac, AMD, Dell, and DMTF, the ecosystem is aligning faster than ever before. Karios will continue to build on these foundations—bridging observability, sustainability, and automation across the entire infrastructure stack.
Final Reflection
The OCP Summit 2025 wasn’t just an event, it was a statement. The industry has moved past proprietary silos and into an era where openness, security, and sustainability coexist.
The show delivered a dynamic look into the future of open hardware and infrastructure, with key highlights centered around advancements in data center efficiency, AI acceleration, and the evolving role of open standards.
- AI-Focused Innovation: A major theme was the acceleration of open hardware for AI. Several vendors showcased new developments in accelerator cards, networking solutions optimized for AI workloads, and advancements in power management for AI infrastructure. Expect to see more open designs tackling the growing demands of AI training and inference.
- Next-Gen Data Center Efficiency: Emphasis was placed on pushing the boundaries of data center efficiency. Discussions revolved around liquid cooling solutions, power management advancements, and innovative rack designs aimed at reducing energy consumption and operational costs.
- Expanding Ecosystem & Partnerships: The Summit highlighted the growing OCP ecosystem, with increased participation from hardware vendors, cloud providers, and technology partners. New collaborations were announced, focusing on accelerating the development and adoption of open hardware solutions.
- Open Networking Advancements: Significant progress was showcased in open networking technologies, including advancements in Ethernet, optical interconnects, and software-defined networking (SDN). The focus is on building more flexible, scalable, and interoperable networking infrastructure.
- Edge Computing Momentum: OCP’s commitment to open hardware for edge computing was evident, with several presentations on solutions tailored for deployments in remote locations and resource-constrained environments.
- Focus on Security: Security remained a critical topic, with discussions around open hardware security standards and best practices for building resilient data center infrastructure.
- Future Roadmap & Community Engagement: The Summit provided a platform for the OCP community to discuss future roadmap initiatives, foster collaboration, and drive innovation in open hardware.
OCP Summit 2025 reinforced the organization’s pivotal role in shaping the future of data center infrastructure. The focus on AI acceleration, efficiency, and open collaboration signals a continued drive towards building a more sustainable, innovative, and accessible computing landscape.
And as Ian Evans closed his ipad after another long day on the floor, the sentiment that lingered was simple: