Leaseweb Launches New Dedicated Server Options, with Further Developments Planned for 2026

Enterprise infrastructure buyers keep running into the same trade-off: bare metal that’s competitively priced but light on compliance documentation, or managed cloud with a broader feature set and less visibility into where data actually sits. Two new Leaseweb Dedicated Server options address specific workloads inside that gap today, and a third wave built specifically to close it is in development for later in 2026.

View Our Dedicated Server Offering


An 8x GPU server for AI training and large-scale inference

The new option puts up to eight H100, H200, or RTX PRO 6000 GPUs on a single bare metal server, connected by NVLink at up to 900 GB/s. It’s built for training and inference work where shared infrastructure introduces latency spikes that are hard to plan around. You get exclusive use of the server: no other customer’s workload ever runs on that hardware, so there’s no competing for GPU cycles and no shared-tenancy question to answer in a compliance review.

Pricing is predictable: one flat monthly rate, no per-second billing, so you can forecast infrastructure spend without watching a usage meter. There are no egress charges on the Leaseweb Private Network either. At sustained utilization above 60%, that monthly bare metal cost comes in below cloud hourly billing over a 12-month period. The servers run in EU data centers under GDPR and NIS2 compliance.

Contact us for a quote


An AMD EPYC 4584PX server for latency-sensitive infrastructure

Not every workload needs more cores. Some need faster ones. The EPYC 4584PX packs 16 cores into a single-socket 1U chassis and outperforms the EPYC 7317P and EPYC 9124 on a per-core basis in the same footprint, boosting up to 5.6GHz.

Teams currently running the 7317P or 9124 get a direct upgrade path: same chassis format, same single-socket setup, no forced move to multi-socket hardware. ECC memory catches and corrects errors before they reach running processes, and IPMI gives full out-of-band remote management without opening a support ticket. It’s a fit for latency-sensitive databases, high-frequency application servers, CI/CD pipelines, and virtualization hosts running a moderate number of demanding VMs, or for anyone on a smaller single-socket server who wants more headroom without a multi-socket price jump.

Check configurations and availability


Building toward infrastructure with EU jurisdiction built in

The next wave in development responds to a specific problem: regulated enterprises, validator operators, and trading teams all need dedicated performance and a defensible compliance position, and today’s market usually makes them choose.

Three areas are driving the work. Financial services, healthcare, and public-sector teams need AI inference that stays inside the EU legal framework, with documentation their compliance teams can actually use. Validator operators need storage that scales with chain growth and uptime that protects against slashing, plus terms that explicitly support blockchain workloads rather than leaving it ambiguous. Algorithmic trading and quant teams need network consistency under load. A connection that’s fast on average but unpredictable during a trading window costs more than one that’s simply steady, and dedicated, unshared hardware removes that variance.

Leaseweb has run independent, EU-operated infrastructure since 1997, and this roadmap is a continuation of that position rather than a new one: dedicated hardware, EU-native operations, support with defined response times, and pricing customers can plan around, purpose-built for each of these three workload types.

If any of this matches what you’re running, get in touch for early access as these launch.

Leave a Reply

Your email address will not be published. Required fields are marked *