A carrier-neutral data center lets the customer choose connectivity independently of hosting. Its value lies not in accumulating logos but in creating a market around the rack: several carriers can compete on price, capacity, destination, latency and terms, and the company can combine or replace networks without moving its servers. That freedom alone does not guarantee lower cost, higher speed or physical diversity; each circuit must be contracted and its routes verified. But it does remove a structural constraint: that whoever controls the space also controls the only way out. Internet Society sums up the effect of direct network interconnection as a potential reduction in latency and cost and an improvement in resilience.[1] By contrast, hyperscalers offer multiple partners and private connections, but keep the interfaces, the pricing and the platform's economic perimeter under their control.
Key findings
- Useful neutrality turns network and hosting into separable decisions. The customer can contract more than one carrier and change a connection without relocating hardware; the existence of simultaneous alternatives is the material basis for negotiating and avoiding forced dependency.[8]
- Direct interconnection can improve economics and performance, but not as an automatic property of every neutral facility. Internet Society notes that peering and exchange points enable more efficient routes, with lower latency and cost, when the right traffic and participants are actually available.[1]
- Cloud economics are asymmetric. Google does not charge for data ingress, but its public Premium pricing from Madrid to Europe charges up to 0.12 dollars per GiB; 10 TiB per month works out at 1,136.52 dollars before discounts and additional services.[2]
- A private connection to a hyperscaler does not remove the platform's economics. AWS Direct Connect charges port hours and outbound transfer; Google Cloud Interconnect charges circuits, attachments and egress data. A Google circuit of 10 Gbps costs 2.328 dollars per hour before traffic and provider costs.[3][4]
- European regulation recognizes that exit can restrict competition. From January 12, 2027 the Data Act removes the charges required to switch providers, including migration egress; it does not make all ordinary transfer free.[5]
What does carrier-neutral mean in operational terms?
It means that contracting space, power and cooling does not force you to contract a network owned by the facility operator. The customer keeps a separate relationship with one or more carriers, or reaches them through the available interconnection infrastructure. Commercial neutrality must translate into concrete procedures: knowing which networks are present, how each circuit is delivered, who signs the contract, what the cross-connect costs, what lead time installation requires and how it is cancelled.
A facility can call itself neutral and offer few real alternatives; it can also host many networks but deliver some of them through third parties or shared routes. The useful question is not “how many logos are there?” but “which options can I contract today, over which path do they arrive, and what would I have to do to change them tomorrow?”.
Neutrality creates a market around the rack
Hosting stays stable while connectivity can be compared, combined or replaced.
Why can freedom of choice reduce cost?
Competition works when practicable substitutes exist. If the customer can request quotes from several carriers, separate transit, private access and backup, and replace one of them without dismantling the rack, every provider has to defend its price and its service. The advantage is not that everything carrier-neutral is cheap; it is that the cost of comparing and replacing is lower than when the network is tied to the building or to a platform.
Negotiation also improves when functions can be split. A company can use one carrier for primary transit, another for backup, a dedicated link to a cloud and peering for specific destinations. That composition makes it possible to pay for each capacity according to its function. A single bundle can be convenient, but it hides what each dependency costs and reduces competitive pressure at renewal time.
Does carrier-neutral automatically mean more speed?
No. Perceived speed depends on capacity, congestion, distance, hop count, routing policy, destination and equipment quality. Neutrality provides options to optimize those variables. Internet Society explains that local peering can avoid intermediaries and unnecessary detours, keep traffic close to its destination and improve performance and resilience.[1] The result must be measured for the company's actual destinations, not inferred from the number of providers.
A direct route to an exchange point or a content network can reduce latency; another application may perform better over international transit. That is why “more speed” must translate into tests: traceroute, packet loss, jitter, latency per destination, port utilization and sustained capacity. Network freedom for a business means being able to change the answer when the data changes.
Does commercial neutrality equal physical diversity?
No. Two contracts can end up sharing the same duct, manhole, intermediate building, wholesale provider or building entry. Commercial neutrality lets you choose brands; physical diversity requires documenting routes and failure points. A resilient design checks at least building entries, metropolitan route, termination equipment, node power and third-party dependency.
The distinction avoids two mistakes. The first is calling an architecture redundant because two invoices show up. The second is ruling out a neutral facility because it does not promise absolute diversity: routes are designed circuit by circuit, and the information may require non-disclosure agreements or confirmation from the carrier.
Do hyperscalers offer a single network provider?
Not literally. AWS Direct Connect, Google Cloud Interconnect and Azure ExpressRoute work with locations, partners and connectivity providers. Microsoft describes ExpressRoute as a private extension of the customer's network into its cloud through a provider and supports virtual connections from colocation facilities.[6] Plurality of access exists.
The difference is the point of control. Once inside, the platform defines regions, gateways, interfaces, price tables, transfer between zones, egress, compatible products and terms of service. AWS Direct Connect has two billing components — port hours and transfer — and still charges for outbound data.[3] Google Cloud Interconnect charges for the circuit, the VLAN attachments and egress; responses to inbound traffic also count as billable egress.[4] A private connection improves performance or predictability, but it does not make the cloud carrier-neutral from the customer's point of view.
Getting in and getting out are not priced the same in the cloud
Example using Google Cloud public pricing from Madrid to Europe, excluding discounts, taxes and other services.
Can a platform change the price of connectivity?
Yes, within the applicable terms. Google announced that on May 1, 2026 it would raise the European list price of CDN Interconnect, Direct Peering and Carrier Peering from 0.05 to 0.08 dollars per GiB, a 60% increase.[7] The announcement was public and recommended reviewing budgets or exploring alternatives. The point is not that a provider cannot revise prices; it is that the customer company does not control that decision and that a tightly integrated architecture can make it harder to respond.
In a neutral environment the carrier can also raise prices, but the customer keeps another lever: comparing a different network without relocating the physical layer. That capability does not eliminate setup costs, lead times or commitments; it reduces the unit of change. Business freedom becomes operational when a commercial decision does not force rebuilding the entire infrastructure.
What happens in a carrier-owned facility?
It can be a valid option. A bundle of hosting and network simplifies contracting, support and accountability, especially when the company has already chosen that carrier and needs no alternatives. The risk appears when changing networks also means changing location, or when competitors can only enter under terms controlled by the owner.
Before signing, it is worth asking whether connectivity is included, whether third parties can come in, who bills the cross-connects, what costs appear on switching and whether the customer keeps its addresses, ASN or contracts. The answer may justify the bundle; what matters is that the dependency is conscious and assessed, not a consequence discovered during an incident or a renewal.
How does this thesis apply to MAD-NE?
MAD-NE is carrier-neutral and lists nine connectivity providers: Adamo, Aire Networks, Cogent, Colt, Correos Telecom, GTD, Lyntia, NearIP and Vodafone. It has two fiber entries and can evaluate Meet-Me Room, cross-connects, dark fiber, DCI and links to ESpanix and DE-CIX depending on design, availability and contracting.[8] These facts prove commercial and interconnection options; they do not authorize claiming that two specific circuits have full physical diversity without reviewing their routes.
The benefit for the customer is being able to design connectivity around its destinations rather than accepting an ipcore-owned network along with the rack. ipcore coordinates physical delivery inside MAD-NE; the carrier remains responsible for its circuit and the customer decides which combination to contract.
Which questions prove that neutrality is real?
| Question | What the answer should clarify |
|---|---|
| Which networks are present? | Direct presence, indirect access and current availability. |
| Who contracts and who bills? | The customer's relationship with carrier, facility and third parties. |
| How is it delivered? | Port, cross-connect, Meet-Me Room, termination equipment and lead time. |
| Can I combine carriers? | Capacity, responsibilities and possible shared dependencies. |
| Which route does each circuit follow? | Entries, ducts, intermediate buildings and failure points. |
| What does switching cost? | Setup, cancellation, minimum term, new cabling, addresses and change window. |
| Can I leave without moving servers? | Availability of the alternative and the replacement procedure. |
Related reading
Methodology and limitations
This piece uses official provider documentation, public prices and materials from Internet Society and the European Commission. The traffic calculations are arithmetic examples using list prices in force in August 2026; they exclude discounts, taxes, commitments, carrier costs and associated products. No private offers are compared, and it is not claimed that every neutral connection is cheaper or faster. MAD-NE's capabilities come exclusively from the facts approved by ipcore.
Conclusion
What does carrier-neutral mean for network cost and freedom? It means that hosting does not decide who connects the customer. That separation turns connectivity into a competitive market, makes route optimization possible and preserves the ability to switch without relocating servers.
A hyperscaler or a carrier-tied facility can offer an excellent solution, but the network remains integrated into its technical and economic perimeter. In a neutral facility, the customer can keep choosing after moving in. The conclusion is a business one before an ideological one: when deliverable alternatives exist, the provider has to compete; when they do not, price, route and continuity become terms imposed by whoever controls the only way out.
Frequently asked questions
What exactly is a carrier-neutral data center?
It is a facility where contracting colocation does not force you to use a network owned by the facility operator. It must provide access to several providers and explain how each connection is delivered, contracted and replaced. Neutrality is proven through practicable options, not just a list of brands.[8]
Does carrier-neutral guarantee a lower price?
No. It guarantees the ability to compare and switch when alternatives are available. That competition can reduce costs or improve terms, but each offer depends on capacity, route, lead time, support and commitment. Peering can reduce costs when it avoids intermediaries and keeps traffic local.[1]
Does carrier-neutral guarantee lower latency?
No. It makes it possible to choose routes and providers that can reduce it. Latency must be measured toward actual destinations and can vary with congestion, routing policy and distance. Exchange points and direct peering usually help when the right traffic is present.[1]
Do two different carriers constitute redundancy?
Not necessarily. They can share an entry, a duct, wholesale fiber or an intermediate building. Diversity requires checking routes and failure points circuit by circuit. Neutrality provides commercial alternatives; technical design decides whether those alternatives are physically independent.[1][8]
Do Direct Connect, Interconnect or ExpressRoute make the cloud neutral?
No. They offer several private ways to reach the platform and can improve performance and predictability, but the hyperscaler keeps the interfaces, regions and pricing. AWS and Google charge port and egress components; ExpressRoute terminates at Microsoft's perimeter.[3][4][6]
What does the Data Act change for egress?
From January 12, 2027 it removes the charges required to switch providers, including the egress transfer associated with migration. It does not require all ordinary traffic or permanent replication to be free. Its existence confirms that exit cost was a relevant competitive barrier.[5]
Sources
- Internet Society, “Peering and IXPs: Building an Affordable and Reliable Internet”.
- Google Cloud, “VPC network pricing”, Premium ingress and egress from Madrid to Europe.
- Amazon Web Services, “AWS Direct Connect pricing” and billing components.
- Google Cloud, “Cloud Interconnect pricing”, circuits, VLAN attachments and egress transfer.
- European Commission, “Data Act explained”, switching rules and egress charges.
- Microsoft Learn, “Azure ExpressRoute overview”.
- Google Cloud, pricing update for CDN Interconnect, Direct Peering and Carrier Peering, May 1, 2026.
- ipcore, canonical MAD-NE connectivity facts, revision of August 23, 2026.
En la preparación de este artículo se han utilizado herramientas de inteligencia artificial para edición ligera y correcciones. La investigación, las afirmaciones y la responsabilidad editorial corresponden al autor firmante.