As we move forward in 2026, it is natural to look ahead, but also to reflect on what was accomplished during 2025. It has been one of the strongest years for Netrouting so far. Throughout the year, we continued to gain traction and grew our margin to be able to heavily invest in our infrastructure and platform. Looking only at today’s demand was not enough; we looked ahead and tried to anticipate where the future of our market would be in a year or two and planned accordingly. This article is not meant to serve as a recap of all the announcements we made during 2025. The text will elaborate on the changes and decisions we have made during this year, the reasons behind them, and what this means for Netrouting in 2026.
A Strong Year, Backed by Real Investment
From a commercial perspective 2025 was an excellent year for the business, with significant growth in both revenue and profitability. We have used this increased headroom to invest further in our infrastructure.
We as a part of the infrastructure group were able to break through on the facilities side related to power, cooling and longer-term strategic planning. But from a customer’s perspective, you see continued progress throughout the year. And this is because we’ve made some very deliberate decisions for the growth and development of Netrouting, where we could easily be going bigger, but we wouldn’t want that to come at the cost of reliability, uptime, or performance.
My mindset during the year heavily influenced the types of technical and financial decisions I made.
Late 2025: Foundations, Not Flash
The final quarter of 2025 was less about big announcements and more about getting the fundamentals right.
Upstream, Peering, and Capacity
One of the biggest goals going into Q4 was to avoid having capacity get constrained. So we did the opposite and proactively invested in more infrastructure before links started to saturate:
- 500 Gbps of new capacity went live with NTT (AS2914)
- Contracted 500 Gbps with TATA (AS6453), now in implementation, ahead of a Q1 2026 go-live.
- Direct peering was established at the Equinix Hong Kong Internet Exchange, offering direct access to local networks and improving latency for APAC traffic.
Our updated numbers don’t reflect extras we’ve added to show off, but rather key improvements to the Stack Overflow platform that will continue to support growth as traffic, geography, and user workload evolve.
Facilities and Global Footprint
Also expansion of the physical installation went on.
Miami saw its largest upgrade to date, including:
- New cabinet deployments to support additional capacity
- Relocation into Equinix MI6
- On-site presence of our specialized and qualified team and round the clock presence from our facility.
- Local system maintenance for customer’s infrastructure from our Miami site.
- Full interconnection with Equinix MI1
- Direct connectivity to NAP of the Americas for local and global connectivity
In Europe and Asia, capacity was added in Frankfurt and Hong Kong, with bare-metal inventory having already been added to the platform. The company has also begun building out the infrastructure in Singapore with the intention to go live with first compute capacity by the end of Q1 2026.
Life After Launch: What Changed Day to Day
It becomes painfully obvious when going live that this work should have been completed previously.
Getting new servers provisioned in the key regions is faster. Routing behaves consistently during high load. Having seen inventory numbers allows for more informed planning. From the surface it looks smoother than it used to.
These are not platform features, like detailed answer templates in the product description, but rather fewer edge cases, fewer “We’ll get back to you” answers, and more confidence when speaking to customer queries regarding availability in specific regions.
Hardware That Actually Stands Out
Hardware decisions in 2025 were driven by one question: does this platform offer a meaningful improvement, or is it just incremental? That mindset shaped how we approach our bare metal servers focusing on systems that deliver real-world performance gains rather than marginal upgrades.
While many retailers spread out their product offerings across dozens of similar SKUs (Stock Keeping Units), our approach differs by featuring systems that clearly differentiate themselves from others on the market:
- AMD Ryzen 9 7950X3D – optimized for cache-sensitive and high-frequency workloads, making it particularly well-suited for latency-critical applications, real-time processing, and performance-driven environments where single-thread speed and cache efficiency matter most.
- AMD EPYC 7702 – designed for dense, memory-heavy compute, offering high core counts and large memory capacity for virtualization platforms, databases, containerized workloads, and large-scale infrastructure deployments.
- AMD Ryzen 3700X – a cost-effective, high-frequency processor that delivers strong single-core performance, making it a practical choice for general-purpose workloads, development environments, and smaller deployments that don’t require enterprise-level density.
- Dual Intel Xeon Gold systems – balanced, enterprise-grade platforms built for stability, scalability, and consistent performance across mixed and multi-tenant workloads, commonly used for virtualization, shared environments, and long-running production systems.
To explore available configurations and see which platform best fits your workload, view our bare metal server options. 👉 Explore Bare Metal Servers
These platforms continue to outperform alternatives in their respective categories, and based on demand, we’re currently building 60 new servers on these architectures.
We standardized on hardware platforms that offer reliable performance at scale, providing a stable foundation for future growth.
The Market Reality: Memory Pricing and Bare Metal
One topic that came up repeatedly throughout 2025 – and is becoming even more relevant in 2026 – is memory pricing.
RAM costs are rising quickly across the industry, reshaping the economics of VM-heavy platforms and short-term cloud deployments. As memory becomes scarcer and more expensive, predictable performance and long-term cost control start to matter a lot more.
This trend is one of the reasons bare metal continues to play a central role at Netrouting. When workloads are memory-bound, having direct access to physical resources often provides a clearer performance-to-cost advantage over shared environments.
We explored this in more detail here: Server RAM Prices Are Skyrocketing – Time for Bare Metal?
Road Ahead: What 2026 Is Shaping Up to Be
With the hard work laid out in 2025, 2026 is off to a great start!
Our next region to launch is Singapore in Q1. We will also be adding further server capacity across North America, Europe and APAC in the coming weeks. As previously stated we are continuing to add upstream and peering capacity to support growing traffic. But the really important thing is that our platform is designed to scale without affecting your experience.
The main point still stands: plan well, build carefully, and don’t make decisions that only make sense in the short term.
We should keep moving forward with early implementation, gradual scaling, and decisions respecting long-term infrastructure flexibility.