NovoServe

03/09/2026

What is defined as a high bandwidth port—10Gbps, 50Gbps, or even 100Gbps?

Many providers boast about high-bandwidth hosting, yet their instant self-service webshops cap dedicated server ports at 10Gbps. While a saturated 10Gbps pipe pushes around 3.2 Petabytes per month, high-concurrency applications can push past that ceiling.

Think live 4K media streaming platforms, edge CDN nodes distributing massive software updates, AdTech bid engines processing millions of real-time requests, and distributed AI clusters synchronizing multi-terabyte training datasets. All of these demand sustained throughput well beyond a single 10Gbps port. When peak traffic hits thousands of concurrent users, a 10Gbps line saturates instantly.

We checked several standard hosting webshops across the industry and found 10Gbps often to be a ceiling for self-service checkout. At NovoServe, we built our 18+ Tbps global network backbone specifically to eliminate port-level bottlenecks. You can configure and deploy dedicated servers up to 50Gbps unmetered directly from our webshop, with 100Gbps custom uplinks available for massive scaling requirements.

What's your high-bandwidth port demand? 10Gbps unmetered, or more?

01/09/2026

What's most important when choosing dedicated servers?

No need for guesswork. Our CRM, HubSpot, analysed over 3k recent chat & form inquiries and mapped the buying priorities of our prospects in no time. Note that to cleanly visualise the pie chart, we grouped related queries such as traffic port speed, networking IPs, datacenter locations into the Bandwidth & port speed category.

Server Specifications dominate at 32%. Your bare metal must precisely match your workload — processing speed, RAM density, and storage I/O are the absolute baseline.

Commercial questions and pricing follow at 18%, sitting right alongside network and infrastructure at 17%, which absorbs those bandwidth demands: locations, port speed, traffic capacity.

Deployment and availability drive 10% of all questions — rapid bare metal server provisioning matters when you're scaling. Use case fit accounts for another 6%, while the remaining 17% covers a mix of niche queries like reseller programs, testing and routing.

What are the most important factors for you when choosing a trusted dedicated server provider? Do they overlap with our data?

27/08/2026

👑 Our most popular category is absolutely unmetered dedicated servers.

It's our most visited webshop category and trusted by many streaming companies. From OTT platforms, CDN operators, video streaming software developers, media server toolkit builders, too video hosting providers, they all keep landing on NovoServe.

Businesses favour our unmetered server for our massive network, some of the lowest latency around, and 18Tbps of capacity behind every bare metal node. Most people expect an "unlimited" catch somewhere. That's why some clients are surprised when they push 10 Gbps, then 20 Gbps, and the connection doesn't throttle.

NovoServe's network consistently scores Grade A on Server Verify, placing in the top 7.56% worldwide, backed by a wide spread of Tier-1 transit providers, and a whole bunch of peering partners.

If you're after a truly unmetered server with real bare metal power behind it, send us a message and let's build.

25/08/2026

NovoServe is hiring for multiple IT & networking roles for our bare metal expansion.

We currently manage a fleet of over 7,000 bare metal servers across 5 TIER-III data centers in Europe and the US. Pushing this infrastructure to deliver 99.99% uptime takes a team that source it, build it, rack it, and get it talking to the rest of the network.

We're hiring across that entire chain for our bare metal delivery.

💪 IT Procurement Specialist — source and negotiate the hardware that powers our infrastructure
💪 Server & Datacenter Engineer — build, rack, and maintain the physical backbone of our network
💪 Network Engineer — keep traffic moving reliably across our global network
💪 Order & Delivery Coordinator — make sure every order lands exactly where and when it should

Based in Enschede or Amsterdam? We're informal, cool, hands-on, and growing fast. Open roles in our career page.

20/08/2026

Denk je bij webhosting automatisch aan een grote buitenlandse cloudpartij?

DE ZAAK heeft een Hosting & Webservices Vergelijker gebouwd die ondernemers in een paar vragen naar de Nederlandse aanbieder helpt die bij hun situatie past — of je nu alleen een domein en hosting zoekt, zelf een website wilt bouwen, of ook zakelijke e-mail nodig hebt.

Leuk om te zien dat NovoServe tussen die lijst met Nederlandse hostingaanbieders staat. Voor klein e-mailhosting en CMS zijn er andere Nederlandse partijen in het overzicht beter in thuis. Waar wij wél voor staan: infrastructuur voor websites en platforms met hoog verkeer. Denk aan streaming, iGaming, e-commerce, web-applicaties en crypto-gerelateerde bedrijven die een Nederlandse partij zoeken die met die schaal en hoge-bandbreedte verkeer overweg kan.

Op zoek naar de Nederlandse aanbieder die bij jouw bedrijf past?
👀 Check de vergelijker van De Zaak:

Hosting en Webservices Vergelijker De Hosting & Webservices Vergelijker van De Zaak helpt ondernemers om eenvoudig de juiste hostingpartij te vinden.

18/08/2026

We are all victims of the AI boom - and it is quietly destroying your infrastructure budgets.

The shift toward AI-driven High Bandwidth Memory (HBM) is choking the global supply of standard RAM and SSDs, causing hardware and hence server hosting costs to surge. However, memory remains the most critical parameter for application performance, as the speed difference between and modern is a factor of 1000.

To combat these rising CAPEX burdens, our Product Manager Sjoerd suggests the sustainability R's for infrastructure cost optimization.

✅ Reduce: Lower your memory footprint by shifting from heavy VMs to lightweight containers.
✅ Rethink: Consolidate your architecture by dynamically sharing resources on larger, centralized servers.
✅ Reuse: Deploy proven, cost-effective legacy hardware. For memory-bound applications, raw capacity matters more than bus speed, making massive 512GB DDR3/DDR4 incredibly cost-effective.

Read Sjoerd’s article on how to optimize your infrastructure costs while navigating today's hardware market.

13/08/2026

NovoServe just opened a direct 200G peering with Meta.🎉🎺🎉

We are proud to announce our straight link into the network behind , , , and . The direct path shortens the route between our infrastructure and Meta's platforms.

And it's only the first half. A second 200G connection is already underway, bringing our peering with Meta to a full 400G.

One more piece of what makes our global network stronger: over 18 Tbps capacity with lots of peering and transit partners.

For anything running on our network — streaming, gaming, or any platform where your users overlap with Meta's — you'll experience lower latency, more headroom, and fewer variables between your platform and the users. We are continuously making our network stronger for the bare metal clients.

Happy , everyone!

11/08/2026

How much GPU memory do you need to run AI locally?

Training and inference are two entirely different workloads. Training requires massive data centers—Microsoft, Meta, and OpenAI are pouring $100B+ into that end of the stack. But running a local model comes down to one hard constraint: fitting the weights in your GPU memory.

Take MiniMax M2.7, a 236B parameter model. In native BF16, that requires roughly 470GB of VRAM. No single GPU gets close.

But when you apply quantization and intelligently route the MoE experts, the hardware requirements change completely. Recent benchmarks on a single 96GB RTX PRO 6000 showed:

✅ Qwen3.5-122B running fully in VRAM with a massive 262K token context window.
✅ GPT-OSS 120B hitting 200+ tokens/sec decode speeds.
✅ MiniMax M2.7 maintaining usable real-time throughput despite offloading.

One single GPU. Frontier-class models. No multi-node cluster required.

At NovoServe, we have single RTX PRO 6000 configurations in stock for this exact workload. If you are figuring out what you can realistically run on sovereign, dedicated hardware instead of renting cloud GPUs by the hour, the full benchmark breakdown is worth a read.

Link in the comments. 👇

05/08/2026

Who is actually moving France's internet traffic?

According to Arcep's 2026 State of the Internet report, inbound traffic to major French ISPs reached 56 Tbit/s at the end of 2025, marking a 10.4% year-over-year growth. Hyperscalers dominate the top tier due to consumer streaming, with Akamai, Netflix, Amazon, Google, and Meta representing 49% of all traffic.

In the Cloud category, NovoServe ranks as the 4th largest contributor. We drive 0.16% of total inbound traffic to France’s top 4 ISPs, trailing only (0.73%), (0.33%), and (0.19%).

We are proud as a B2B Bare Metal provider to secure a 0.16% share of an entire country’s ISP traffic. We rank 20th overall, routing more traffic into French ISPs than global consumer platforms like Twitch (0.15%), Telegram (0.14%), and Alibaba Cloud (0.05%).

For our enterprise clients, this shows the pure structural capacity and peering density of our unmetered dedicated servers.

30/07/2026

Just another day walking the floor at our Amsterdam data center with Herke Plantenga.

😎 At NovoServe, we operate on a simple principle: everyone visits the data center. From our CEO to sales, marketing, dev, and product teams, we all frequent our DCs.

We ned to hear the fans and physically see the bare metal we deploy. Understanding dedicated hosting requires moving past spreadsheets and network diagrams. By touching the infrastructure, our entire team knows exactly what we deliver to our clients. It builds a unified respect for the high-bandwidth, unmetered environments powering our global network.

When you know the hardware, you build better solutions. It is that simple.