OVHcloud and Hetzner announced broad price increases that hit VPS, cloud instances, dedicated servers, and even add-ons like IPv4 addresses. This article explains what’s really driving the increases, why the two providers are reacting differently, what to expect next, and how hosting providers can reduce risk and costs.
AI changed the supply chain economics, and RAM is the new bottleneck
Both OVHcloud and Hetzner point to the same root cause: the cost and availability of core server components (especially RAM and flash storage) shifted faster than efficiency gains in virtualization, operations, and procurement. The trigger is not a single event, but a stack of reinforcing pressures that arrived at the same time.
1) The AI boom is consuming memory capacity, not just GPUs
AI infrastructure needs enormous amounts of memory bandwidth, which pushed manufacturers to prioritize high-bandwidth memory (HBM) used in AI accelerators. Producing HBM is complex (stacked dies, through-silicon vias, lower yields), so when wafer capacity is diverted toward HBM, it effectively squeezes the supply of conventional DRAM used in normal servers. The knock-on effect is exactly what hosting providers feel first: sudden DRAM scarcity and sharp contract price jumps. When RAM becomes the limiting component, every new server generation becomes significantly more expensive to build, even if CPUs do not rise as dramatically.
2) NAND flash pricing followed DRAM upward
Server NVMe and SSD capacity is also tied to the same constrained manufacturing ecosystem. With demand rising for high-performance storage to feed data-hungry AI and analytics workloads, and with manufacturers juggling production priorities, NAND pricing has also surged. Products that are storage-heavy (object storage, “storage optimized” dedicated servers, NVMe-centric VPS hosts) see immediate margin pressure because storage is a large fraction of bill-of-materials cost.
3) Data centers are power-constrained, and power density keeps rising
Even when you can buy hardware, you still need space and electricity to run it. Industry research shows occupancy in major hubs is extremely high, with long waits for new grid connections in some regions. At the same time, modern servers and AI-adjacent workloads increase rack density, driving higher cooling and power delivery costs. That combination raises operating costs and forces providers to fund capacity expansions earlier and at higher prices than in previous years.
4) IPv4 became a recurring “connectivity tax”
IPv4 exhaustion continues to translate into recurring rental fees. Both hyperscalers and European providers increasingly treat IPv4 as a scarce asset rather than a bundled utility. OVHcloud’s US notice reflects this directly: additional IPv4 increased from $2.00 to $2.40 per IP per month. Hetzner’s IPv4 pricing has similarly moved upward in recent cycles. For hosting providers managing many customer endpoints, load balancers, NAT gateways, or legacy IPv4-only integrations, this line item can become material surprisingly fast.
What OVHcloud and Hetzner are actually doing (and why it differs)
While the root causes are shared, OVHcloud and Hetzner chose different “blast radius” strategies. Understanding the difference helps predict what comes next and where your costs are most exposed.
OVHcloud: targeted increases, with a clear attempt to preserve entry tiers
OVHcloud’s communication (effective April 1, 2026) frames the update as applying to a limited set of Bare Metal Servers, the VPS 2026 range, and additional IPv4. Just as important is what OVHcloud explicitly excludes: Eco (including Kimsufi and So You Start), and many older generations or specific VPS families (Comfort, Elite, Value, Starter, Essentials, and others). This pattern strongly suggests OVHcloud is using hardware lifecycle management to protect price-sensitive segments. In other words, older amortized fleets can remain stable, while new builds based on DDR5 and high-density NVMe must be repriced.
Two changes stand out for general hosting buyers:
- VPS increases are steep: OVHcloud’s VPS-1 moved from $4.90 to $7.60, and VPS-4 from $26.00 to $43.50. That’s not a “small adjustment”; it indicates the cost of RAM per host has risen enough that low-end VM economics no longer work at prior price points.
- Modern bare metal sees moderate-to-high increases depending on tier: some Advance Gen4 servers rise modestly (e.g., ADV-1 Gen4 from $130 to $134), while others rise sharply (ADV-3 Gen4 from $210 to $255). This matches a reality where RAM-heavy or NVMe-heavy configurations are disproportionately affected.
Hetzner: a broader reset across regions and product families
Hetzner’s public statement is more sweeping: it cites “drastic” cost increases and applies new pricing broadly across products and geographies (Germany, Finland, US, Singapore). Reported changes commonly land in the 30 to 50 percent range, and in some US items even higher. That breadth fits Hetzner’s historic positioning: high-volume, low-margin infrastructure where prolonged component inflation cannot be absorbed without repricing. If your business model depends on Hetzner as the baseline low-cost supplier, this shift matters because it suggests less room for “temporary” discounts going forward.
Another nuance: Hetzner’s pricing in regions where it leases capacity (notably the US and Singapore) can move more aggressively because you are effectively stacking cost drivers: hardware inflation plus facility, power, and regional commercial terms.
What happens next: a realistic 2026 to 2027 outlook
For hosting providers, the most important question is whether this is a one-time bump or the start of a new normal. The evidence points to a multi-year adjustment cycle rather than a quick reversal.
- Memory tightness likely persists into 2027 and possibly 2028: as long as HBM and AI accelerator demand keeps priority access to wafer capacity, conventional DRAM volatility remains. Even if pricing stabilizes, it may stabilize at a higher plateau.
- More segmentation, not uniform cheapness: expect “old fleet” offerings (Eco, auctions, older generations) to remain relatively attractive, while new-generation instances track component markets more closely. Pricing becomes less like a smooth curve and more like tiers separated by hardware lifecycle.
- Networking line items keep growing: IPv4 rent, paid load balancers, paid firewalls, managed NAT, and “small” per-GB fees add up. Many teams will discover that optimizing networking architecture yields savings comparable to moving providers.
- Capacity becomes a differentiator: some providers may limit expansion to avoid buying peak-priced hardware, which can lead to longer provisioning times or reduced availability of popular SKUs.
How to mitigate the impact (practical steps for hosting providers)
If you resell hosting or run customer workloads on OVHcloud or Hetzner, you typically have two goals: protect margins and reduce the risk of more sudden increases. The best approach is usually a mix of architectural changes, procurement tactics, and product packaging.
- Audit RAM-per-service first: because RAM is the cost center under the current AI-driven squeeze, right-sizing memory often saves more than cutting CPU. Identify VMs that carry oversized buffers, language runtimes with high baseline usage, or caching layers that can be moved to dedicated nodes.
- Use lifecycle arbitrage: move steady, predictable workloads to older generation servers or “eco/auction” style offerings where the hardware cost is already amortized. Reserve newest generations for workloads that truly benefit from the performance per watt and higher I/O ceilings.
- Reduce IPv4 dependency: adopt an IPv6-first posture wherever your customer base allows it. If you run multi-tenant platforms, consider IPv6 as default and IPv4 as paid add-on. Also consolidate public endpoints: fewer public IPs, more routing behind them.
- Re-check storage strategy: separate “hot” and “cold” data. Keep high-IOPS datasets on NVMe, but move backups, archives, and logs to cheaper tiers and use retention policies aggressively. Storage is often over-provisioned because it feels cheap until it doesn’t.
- Renegotiate commitments and terms: OVHcloud’s notice indicates commitments can delay the effect until renewal. For stable workloads, commitments can buy time and predictability. For volatile workloads, avoid long lock-ins and keep portability high.
- Build a migration-ready baseline: even if you do not want multi-cloud complexity, keep portability: Infrastructure as Code, immutable images, and standard observability. The goal is not to move every week; it’s to have leverage if pricing or availability changes again.
- Update your own pricing model: if you resell VPS or managed hosting, consider indexing your plans to measurable resources (RAM, storage, IP count, bandwidth) and explicitly treating IPv4 as a billable scarce resource. Many providers absorb these costs until a market event forces a sudden and painful repricing.
Conclusion
The OVHcloud and Hetzner 2026 price increases are not random: they reflect AI-driven pressure on memory and storage supply chains, higher data center power constraints, and growing “hidden taxes” like IPv4 rent. OVHcloud is buffering parts of its catalog via hardware lifecycle segmentation, while Hetzner is executing a broader reset. Expect higher baseline infrastructure prices through 2027, and mitigate by right-sizing RAM, reducing IPv4 reliance, and keeping your stack portable across providers.