Below Hyperscale, Above the Island: Data Services Pacific and the Econ

Below Hyperscale, Above the Island: Data Services Pacific and the Economics of Local Cloud Survival in New Caledonia

Thesis

Data Services Pacific is best understood not as a miniature hyperscaler, and not merely as a local web host, but as a New Caledonian trust-and-continuity intermediary built around scarce local infrastructure. Its value is produced by three mechanisms that are unusually visible in small island markets: control of local hosting space, publicly verifiable Internet-number resources, and the ability to convert proximity into operational trust. Its constraint is equally clear: it sits below the scale at which cloud economics are driven by global purchasing power, yet above the scale at which a customer can cheaply self-host. That middle position creates a business in which survival depends less on raw compute scale than on local credibility, route visibility, supplier relationships, jurisdictional comfort, and customer switching costs.

The public record presents Data Services Pacific, usually shortened to DSP, as a Nouméa-based hosting and cloud-services operator associated with CIPAC Group. DSP appears in APNIC records as organisation handle ORG-DSP1-AP, a Local Internet Registry in New Caledonia, with the same Nouméa address used across its website, legal notices, PeeringDB records and customer hosting disclosures. Its current routed network identity is AS134405, with three IPv4 /24s and three IPv6 /48s visible in public routing and RPKI-valid route objects. Public routing databases identify one upstream, Office des Postes et Télécommunications de Nouvelle-Calédonie, or OPT-NC, and DSP is also present at the local CAN’L IX exchange in Nouméa with two 10G ports. DSP markets itself as the only local actor with two data centers and describes a third site for backup replication; PeeringDB independently records its DC1 facility in Nouméa, while Data Center Map and DSP’s own materials identify both DC1 and DC2.

The economic significance is not that DSP owns a large Internet footprint. It does not. Public IP intelligence sources count only 768 IPv4 addresses originated by AS134405, and PeeringDB places its traffic level in the 100–1000 Mbps band with a heavy outbound ratio. This is small by global hosting standards. But smallness is not the same as economic irrelevance. In New Caledonia, where international connectivity, wholesale access and local infrastructure are shaped by island geography and by OPT-NC’s historically central role, the ability to host applications locally, provide disaster recovery within the territory, publish a clean network identity, and offer a named local engineering team can be economically valuable even at modest scale.

DSP therefore reveals a broader rule of small-market infrastructure economics: below hyperscale, trust can substitute for scale only when customers face real frictions in moving elsewhere. Those frictions are legal, geographic, operational and psychological. A local enterprise may be able to buy remote cloud in Sydney, Singapore, France or elsewhere, but that does not eliminate the need for local support, predictable latency to local users, data-location comfort, disaster-recovery planning, and a counterparty that can be reached by phone in Nouméa. DSP monetizes those frictions. Its risk is that the same frictions are not fully under its control: upstream transit, power, imported hardware, software licensing, submarine-cable economics and the broader financial condition of the local market remain external constraints.

The target identity: DSP as legal company, operating label and network resource holder

The canonical target is Data Services Pacific, a Nouméa-based company using the operating label DSP. APNIC records list “Data Services Pacific” as ORG-DSP1-AP, an APNIC Local Internet Registry in New Caledonia, with address 210 Rue Gervolino, Nouméa, and administrative contacts at dsp.nc. The AS134405 APNIC record uses the as-name DATASERVICESPACIFIC-AS-AP, describes the holder as Data Services Pacific, and links the network to the same organisation record. The abuse and incident-response records are also at dsp.nc, and the abuse contact was validated in February 2026. This establishes that Data Services Pacific is not merely a brand appearing on a website; it is the public registry entity associated with routable Internet-number resources.

The legal identity visible in New Caledonian registry notices is “DATA SERVICES PACIFIC,” R.C.S. Nouméa 2005 B 759 779, a SARL with head office at 210 rue Gervolino. A 2011 registry notice records the company name, trade name, SARL form, capital of 1,000,000 XPF, Nouméa address and a change of managers. A 2014 notice records the same RCS number and SARL form with different managers. A 2018 registry notice records a formal decision not to dissolve the company after equity had fallen below half of share capital. The 2024 Chamber of Commerce electoral list still includes DATA SERVICES PACIFIC with the same RCS/RIDET number in the services category. These records indicate continuity of the legal entity from at least 2005 through 2024, but they do not by themselves reveal current shareholding or full financing arrangements.

The ownership and operating-control context is more ambiguous. DSP’s own website describes it as “100% private,” based in Nouméa, and dedicated to hosting IT solutions. The downloadable DSP brochure, however, describes DSP as a subsidiary of Groupe CIPAC. CIPAC’s own 2026 article also calls DSP a CIPAC subsidiary, and DSP’s legal notice says the dsp.nc site is administered by CIPAC SA, a Nouméa company with its own RCS and RIDET identifiers. The most conservative reading is that Data Services Pacific remains the legal/network-services company and DSP is the operating label, while CIPAC is the group or parent context. The public evidence does not prove the exact current shareholder chain, but it does prove that the DSP operating identity, the APNIC resource identity and the CIPAC group context are connected.

This ambiguity matters economically. If DSP is a standalone SARL with only limited parent support, its survival depends heavily on its own cash flows and debt capacity. If it is meaningfully supported by CIPAC Group, DSP may be able to carry data-center fixed costs, hardware procurement cycles and delayed customer collections more safely than a pure independent host. The 2018 non-dissolution notice is therefore a significant survival signal: at some point, the company had suffered enough accounting loss or balance-sheet deterioration to trigger formal equity-loss procedures. The later public evidence of active services, current APNIC validation, data-center expansion and CIPAC articles suggests continued operation, but it does not erase the earlier financial-stress marker.

What DSP sells: local continuity, not generic compute

DSP’s website frames the offer around hosting, cloud services, security, reliability, proximity and performance. It describes rack rental, electrical supply, redundant Internet connectivity, cooling and physical security as core colocation functions, while the customer retains ownership of its hardware and software. The same site emphasizes local hosting benefits: outsourcing infrastructure, proximity, security, simpler relationships, a French and European legal framework, bandwidth cost, network performance, resilience, and an ecosystem of local partners and Internet service providers. In economic terms, this is a bundle of physical hosting, local network access, managed operations and jurisdictional comfort.

The service catalogue extends beyond colocation. DSP’s brochure and CIPAC materials describe IaaS, BaaS, DRaaS, private cloud, application hosting, backup, supervision, platform operations, disaster-recovery planning, business-continuity planning, Microsoft licensing, and support. CIPAC’s 2026 GPUaaS article adds a newer layer: DSP had acquired two AI-dedicated servers and planned to offer compute capacity by monthly rental, positioned as a New Caledonian sovereign cloud extension. That GPUaaS move is economically revealing. It is not a hyperscale AI build-out; it is a local-capacity experiment. The question is whether there is enough local demand for GPU hours from public institutions, health establishments, local authorities, digital actors and enterprises to justify the stranded-capital risk of specialized servers.

The brochure describes DSP’s value chain as “end-to-end” and emphasizes a single interlocutor, 24/7 support, intervention within 30 minutes, monitoring and supervision, backup operations, relations with hardware vendors, and relationships with providers. Its SLA material states 99.98% annual availability for the data-center hosting environment, 99.8% annual availability for hosted services, a 30-minute response commitment, 4-hour restoration target, and disaster-recovery objectives of RTO two hours and RPO 24 hours. These are self-reported commercial commitments, not third-party certification results, but they show the shape of the product: DSP is selling operational assurance more than raw compute.

The brochure also claims approximately 100 clients on a proven resilient platform. The public customer sample is narrower but economically meaningful. Nespresso New Caledonia’s legal notice identifies DSP as host. MDF, a local medical and dental services group, identifies Data Services Pacific as the host of its website. ISEE, New Caledonia’s official statistical institute, lists DSP as the host of its website. These are not proof of the whole client base, and website hosting does not necessarily mean full IaaS or disaster recovery. But the visible sample spans retail/e-commerce, health-related services and public/statistical presence, which are exactly the categories where local trust, continuity and data-location comfort can matter.

DSP’s LinkedIn profile gives an additional channel signal. It describes the company as a private New Caledonian firm specialized in outsourcing and hosting, with more than ten years of experience, two interconnected data centers, fallback rooms for business continuity, and specialties including datacenter, housing, VPS, IaaS, PaaS, SaaS, networking, Microsoft SPLA, Veeam, backup, replication, storage, disaster recovery, business continuity, BaaS, DRaaS, IXP, private cloud and sovereign cloud. LinkedIn is not an audited source, and several terms, including ISO27001 and HDS, appear there as profile keywords rather than verified certifications in the public record reviewed here. Still, it is useful as market-channel material: DSP wants buyers to perceive it as a local sovereign-cloud and continuity provider, not as a low-end shared host.

Geography and facilities: scarcity of local hosting space

The physical footprint is central to DSP’s economics. DSP’s website and brochure place the company at 210 rue Gervolino in Nouméa. PeeringDB records the DSP DC1 facility at the same address and lists Data Services Pacific as the organization behind it. PeeringDB also records two networks at the facility: Data Services Pacific AS134405 and THEMIS AS149520, and identifies CAN’L IX as present there. This matters because facility recognition in PeeringDB is a form of industry-legible infrastructure evidence. It tells network operators that DSP DC1 is not only a building with servers but a location where networks can interconnect.

DSP’s own material claims a second data center, DC2-GAL, used for disaster recovery and business continuity, and says backups are copied to a third site. Its infrastructure diagram shows production, disaster recovery and off-site backup sites, with links to OPT, CAN’L, Lagoon and the local exchange. Data Center Map independently lists two DSP facilities in Nouméa: DSP-DC1 at 210 rue Gervolino with 20 racks and DSP-DC2 at 34 rue du général Gallieni with 6 racks. This supports the existence of two DSP data-center locations, while also showing their small absolute scale. A 20-rack and 6-rack footprint is not hyperscale infrastructure; it is a local resilience platform.

CIPAC’s March 2025 article says DSP was expanding its second data center to support economic activity and growing digital needs in New Caledonia, with emphasis on resilience amid sociopolitical issues and cybersecurity risk. The expansion included additional hosting space, access control, a new airlock and preparation space. This is useful because it is a recent group-level signal that DSP was still investing in physical infrastructure after the 2018 financial-stress marker. It also implies that the local demand thesis had not disappeared: CIPAC and DSP saw enough market need to justify capacity and security work at the second site.

The unresolved point is the independent status of the third site. DSP’s brochure says backups are copied to a third site and that data are secured in three first-rank data centers in New Caledonia. Public facility directories and PeeringDB records reviewed here do not independently identify a third DSP-operated facility. The economic interpretation should therefore distinguish between “two named DSP data centers” and “a third backup site claimed in DSP material.” If the third site is a controlled, contractually robust and geographically separated data center, DSP’s disaster-recovery value is stronger. If it is a smaller backup location or partner site, the product may still be valuable, but the resilience claim is more dependent on contract details and operational execution than on owned physical infrastructure.

Small local data centers have a different cost curve from hyperscale sites. They do not gain the same purchasing leverage on energy systems, servers, cooling equipment, optical transport, security systems or software licenses. Their advantage is not lower unit cost of compute; it is lower transaction cost for local continuity. For a New Caledonian buyer, moving a workload to DSP may avoid capital expenditure on an on-premise server room, reduce local operational burden, and keep support relationships in Nouméa. DSP’s own brochure makes this explicit by selling outsourcing, support, backup, supervision and energy savings, including an estimate of up to 118,000 XPF excluding tax per 42U rack per month in energy savings versus customer premises. That number is a marketing claim, but it illustrates the commercial logic: DSP turns the hidden cost of small enterprise server rooms into recurring infrastructure revenue.

ASN, addresses and route visibility as economic assets

AS134405 is the current public network identity of Data Services Pacific. BGP.tools identifies AS134405 as registered in October 2018, allocated under APNIC, and originating three IPv4 and three IPv6 prefixes: 103.123.232.0/24, 103.123.233.0/24, 203.34.36.0/24, 2404:e9c0::/48, 2404:e9c0:1::/48 and 2404:e9c0:2::/48. BGP.tools marks the listed routes as RPKI-valid. IPinfo similarly counts 768 IPv4 addresses and identifies the three IPv4 /24s as RPKI-valid, while WhatIsMyIP lists the same six IP ranges in New Caledonia.

The address block 203.34.36.0/24 is particularly interesting because APNIC lists it as “ASSIGNED PORTABLE,” with netname DATASERVICESPACIFIC-NC. Portable address space can be more valuable than provider-assigned space in a small hosting economy because it gives the operator greater continuity across upstream changes and supports a more independent routing identity. DSP’s 103.123.232.0/24 block is listed as DATA-SERVICES-PACIFIC-DC1, with description “Data Services Pacific DC 1 Noumea,” while APNIC also records 2404:e9c0::/32 IPv6 allocation to Data Services Pacific. The blocks are not large, but they are enough to support a local hosting and cloud platform with publicly visible routing and reputation.

DSP also has a historical ASN, AS24053. APNIC and BGP.tools identify AS24053 as Data Services Pacific, registered in 2005, but BGP.tools says it is not currently in the global routing table and originates no IPv4 or IPv6 prefixes. The likely economic interpretation is a network-identity transition: AS24053 reflects an older DSP Internet-resource history, while AS134405 is the current routed identity. The public record does not explain why AS24053 remains allocated but unrouted. That unresolved fact is not trivial. A retained but inactive ASN can be administrative residue, a reserve asset, a historical migration artifact, or a sign of earlier routing arrangements that were later replaced.

Route visibility is commercially important because it turns an otherwise opaque local infrastructure provider into a verifiable network counterparty. Customers, suppliers and operators can observe that DSP has APNIC registration, validated abuse contacts, routable prefixes, RPKI-valid announcements, PeeringDB presence and visible interconnection at CAN’L IX. In a small market, this is a reputational asset. It reduces diligence costs for buyers who need evidence that the provider is not simply reselling anonymous hosting, and it creates accountability through abuse-mailbox, route-object and peering records.

The same visibility also exposes DSP’s limits. IPinfo labels AS134405 as a stub AS, lists one peer and one upstream, and shows no downstreams. Its observed traceroute from Nouméa moves through AS18200 before reaching AS134405. PeeringDB gives DSP’s traffic level as 100–1000 Mbps and its traffic ratio as heavy outbound. These signals are consistent with a hosting/cloud operator that serves local customers and sends content or hosted-application traffic outward, not with a carrier-scale network or transit provider. Public visibility therefore creates trust, but it also prevents DSP from pretending to be larger than it is.

Upstream dependence: the invisible price setter

The central external dependency in DSP’s economics is OPT-NC. BGP.tools identifies AS18200, Office des Postes et Télécommunications de Nouvelle-Calédonie, as DSP’s upstream for both IPv4 and IPv6. IPinfo also lists AS18200 as DSP’s sole upstream. PeeringDB records DSP at CAN’L IX, but BGP.tools and IPinfo still show the upstream graph as OPT-dependent. This combination is economically important: local peering can reduce cost and latency for local traffic, but it does not eliminate dependence on the territory’s dominant infrastructure provider for broader reachability.

The regulatory backdrop reinforces this. The New Caledonian Competition Authority states that OPT-NC holds a monopoly over public telecommunications services, while the Internet access market is open to competition. The same summary warns about the relationship between OPT-NC’s monopoly missions and competitive markets, including risks of cross-subsidies, and recommends a declarative regime rather than a restrictive authorization regime for Internet access providers and service operators. A separate market summary describes OPT-NC as dominating the telecom sector, including fixed and mobile voice, mobile Internet, fixed broadband and wholesale services to ISPs.

The submarine-cable layer is also relevant. OPT-NC’s own material describes Gondwana-1 as the cable linking Nouméa to Sydney since 2008 and says a second cable project was launched to provide additional capacity and resilience. SubmarineNetworks identifies Gondwana-1 as a 2,151 km system connecting New Caledonia and Australia, ready for service in September 2008 and owned/operated by government-owned OPT. For DSP, this means the cost and resilience of international access are not simply a procurement line item; they are structural determinants of the local hosting market. If international bandwidth is expensive or fragile, local hosting has more value. If international capacity becomes cheaper, more resilient and more competitive, remote cloud substitutes become stronger.

DSP’s brochure shows this dependency indirectly. Its network diagram refers to OPT links at multiple sites and also references CAN’L, Lagoon and local interconnection. It argues that local hosting can improve performance and resilience and reduce local and international bandwidth cost. The economic mechanism is clear: DSP does not need to own international cables to benefit from their scarcity. It needs to sit at the point where local customers experience the scarcity as risk, latency, price or complexity, and then sell a locally managed alternative for workloads that can remain in New Caledonia.

This also defines DSP’s bargaining problem. A small hosting operator with one publicly visible upstream does not have strong supplier diversification. Its margins can be compressed by upstream transit prices, power costs, equipment import costs, maintenance contracts and software licensing. Its buyer-facing pricing power must come from service differentiation, not from procurement scale. The two 10G CAN’L IX ports are useful, but they do not by themselves create wholesale bargaining parity with OPT-NC. They create local network legitimacy and local traffic efficiency.

CAN’L IX and the local interconnection layer

DSP’s interconnection position is visible at CAN’L IX. PeeringDB records CAN’L IX in Nouméa with five peers, seven connections and total capacity of 60G. DSP appears there with two 10G connections, IPv4 addresses 103.23.55.9 and 103.23.55.10, IPv6 addresses 2401:c00:1:4::9 and 2401:c00:1:4::10, and an open peering policy. PeeringDB also records DSP’s network profile as open peering, with no contract requirement, no traffic-ratio requirement and no requirement for multiple locations.

The IX footprint has two economic effects. First, it supports local performance. If local ISPs, hosting providers or public-service networks exchange traffic at CAN’L IX, customer applications hosted at DSP can be reached through shorter local paths instead of hairpinning through remote transit. Second, it supports market trust. In small markets, the ability to say “we are present at the local IX” signals that a hosting provider is part of the local operator ecosystem rather than a pure reseller or office-based IT shop.

The limitation is that CAN’L IX is itself small. Five peers and seven connections are useful but not transformative in a global routing sense. The peer list includes local actors, and DSP’s own PeeringDB profile puts its scope in Asia Pacific, not global. This is the classic island-interconnection tradeoff: the local IX can economize on local traffic and make local hosting more attractive, while the island still depends on upstream international capacity for external reach.

The location of the IX inside DSP DC1 or associated with the DSP DC1 facility increases DSP’s strategic value. PeeringDB records CAN’L IX at DSP DC1, and the DSP facility record lists both CAN’L IX and networks present at the site. That means DSP is not only a user of interconnection; it is also a physical host of part of the local interconnection fabric. This is an important distinction. A small data-center operator with an IX presence can monetize colocation, cross-connects, network adjacency and reputational centrality even when raw compute demand is limited.

Address scarcity, reputation and the trust premium

DSP’s public address base is small: 768 IPv4 addresses across three /24s. In a large cloud, that would be trivial. In a small island hosting market, it is enough to create a scarce, reputation-bearing asset. IPv4 addresses can be allocated to hosted services, customer virtual machines, NAT pools, DNS, VPN, mail, monitoring and management systems. Because address reputation is cumulative, every abuse event or misconfigured customer can impose a cost on the whole operator. A small operator therefore has less room for noisy tenants than a hyperscaler does.

IPinfo reports 35 hosted domains across AS134405 and tags at least one IP as associated with VPN usage. That VPN tag should not be overread: it is a third-party classification signal, not proof of abuse or misconduct. But it illustrates why resource visibility matters. Hosting companies are judged by the cleanliness, responsiveness and traceability of their IP space. APNIC’s abuse contact validation in 2026 and DSP’s visible RPKI-valid routes are therefore part of the trust product. They tell counterparties that the operator can be identified, contacted and held accountable in the normal Internet-governance stack.

For enterprise and public-sector customers, this matters more than it may appear. A small local cloud provider cannot compete with hyperscalers on the breadth of services or the price of commodity compute. It can compete on the reduction of operational uncertainty. A buyer can verify the provider’s address resources, ask about local routing, inspect legal notices of other local customers, visit the facility or meet the team, and escalate issues through local relationships. That bundle of verification is a trust premium. It is also hard to scale beyond the territory. The same local embeddedness that makes DSP credible in Nouméa may not travel well to a larger regional market.

Address-resource visibility also influences switching costs. If a customer uses DSP-provided IP addresses, DNS, backups, VPNs or hosted firewall rules, migration is not just a matter of copying virtual machines. It involves re-addressing, DNS cutovers, firewall changes, backup-chain migration, compliance review, operational testing and user retraining. For a customer with disaster recovery or managed backups at DSP, the switching cost is higher still because the historical backup set and recovery procedures are part of the service. DSP’s brochure emphasizes backup, replication, supervision and PRA/PCA precisely because these are sticky products.

Revenue logic: recurring services over scarce local assurance

DSP’s revenue model is not disclosed in detailed public pricing. The website says offers are à la carte according to customer requirements and directs buyers to contact the company for availability, conditions and pricing. That lack of posted pricing is itself informative. In small enterprise infrastructure markets, pricing is often configured around rack space, power density, bandwidth commitments, VM sizing, backup retention, recovery objectives, Microsoft or other software licensing, managed-service hours and support commitments. Public price menus are less useful when the product is a risk bundle rather than a commodity server.

The likely recurring revenue streams are colocation or rack rental, virtual infrastructure, hosted applications, private cloud, backup, disaster recovery, monitoring, support, licensing pass-through and, more recently, GPU capacity by monthly rental. CIPAC’s GPUaaS article explicitly says the AI servers would be offered by monthly rental, and describes DSP’s broader service lines as IaaS, BaaS, DRaaS, hosting/private cloud and GPUaaS. These are recurring or quasi-recurring services, which is economically attractive because fixed data-center costs are high and utilization matters.

Pricing power comes from five sources. The first is local scarcity: DSP claims to be the only actor with two data centers, and facility directories show only a small number of listed data centers in New Caledonia. The second is operational proximity: DSP advertises local support, simplified relationships and a local team. The third is legal and data-location comfort: its own materials sell a French and European legal framework and local data security. The fourth is interconnection: DSP has visible ASN, RPKI and CAN’L IX presence. The fifth is switching cost: backup, DR and managed infrastructure are harder to migrate than a static website.

Gross-margin pressure comes from the other side of the ledger. DSP must pay for power, cooling, facility maintenance, security systems, battery or generator resilience, imported hardware, vendor support, software licensing, staff and upstream connectivity. Its brochure displays technology-partner logos including APC, VMware, Microsoft SPLA Partner, Veeam, Dell EMC, Synology, Cisco and related infrastructure vendors. Those relationships help DSP deliver enterprise-grade services, but they also expose it to foreign-currency, licensing, support and procurement constraints. A small operator cannot dictate terms to Microsoft, VMware, Veeam, Cisco or Dell EMC.

The most delicate margin question is utilization. A local data center must carry high fixed costs whether racks are full or empty. Adding GPU servers increases the utilization risk because specialized compute depreciates quickly and depends on workload adoption. If local enterprises, public institutions, healthcare entities and digital actors consume the GPUs steadily, DSP can create a differentiated sovereign-compute niche. If demand is episodic, GPUaaS becomes a symbolically useful but financially weak capital expenditure. This is the same below-hyperscale problem in sharper form: local control has value, but hardware economics still punish underuse.

Customer trust: why local evidence can beat global scale for some workloads

The public evidence suggests DSP serves customers for whom local trust matters. ISEE’s website identifies DSP as host; MDF’s legal notice identifies Data Services Pacific as host; Nespresso New Caledonia’s legal notice identifies DSP as host. These are visible examples rather than a full account list, but they cover public information, health-adjacent services and consumer-facing retail. In each case, the commercial meaning is less about the complexity of website hosting and more about the reputational signal: local institutions and recognizable brands are willing to disclose DSP as an infrastructure counterparty.

Trust in local infrastructure has several layers. The first is physical trust: customers can know where the servers are and, in some cases, visit or audit the facility. The second is legal trust: DSP sells local hosting under a French and European legal framework. The third is operational trust: DSP names support commitments, monitoring, restoration targets and local intervention. The fourth is social trust: in a small market, reputation travels through business networks, integrators, ISPs and group relationships. DSP’s own website says it is positioned at the center of the local digital ecosystem and relies on a partner network.

This trust premium is economically rational when the alternative is not simply “AWS versus DSP,” but a more complex choice among on-premise server rooms, local ISP services, regional cloud reachable over international links, and managed service providers. A Nouméa customer may prefer remote cloud for elasticity, breadth of services or global tooling. But if the workload is local, latency-sensitive, politically or legally sensitive, or operationally dependent on local technicians, DSP’s smaller scale can be offset by proximity and accountability.

The tradeoff is concentration. A customer who outsources primary hosting, backup and disaster recovery to the same local provider may reduce internal complexity while increasing counterparty dependence. DSP’s “single interlocutor” promise is commercially attractive because it lowers coordination cost. It is also a lock-in mechanism. The more DSP manages, the more costly it becomes for the customer to separate colocation, backup, licensing, monitoring, disaster recovery and support across multiple suppliers.

Competitive field: local ISPs, on-premise rooms and remote cloud

New Caledonia’s connectivity market is shaped by the interaction of OPT-NC, local ISPs and service operators. The APNIC blog notes that despite OPT-NC’s historical monopoly over infrastructure, CAN’L was established as an ISP in 1995 with CIPAC Group involvement, followed by MLS in 1997 and later Telenet and Nautile. HE.net’s New Caledonia network list shows local ASNs including OPT-NC, Micro Logic Systems, OFFRATEL, Nautile, TeleNet, CAN’L and DSP, with OPT-NC much larger by adjacencies and routes. DB-IP’s country allocation list likewise shows OPT, OFFRATEL, CAN’L, Micro Logic, Nautile and TeleNet with larger IPv4 allocations than DSP.

DSP’s competition should therefore be separated by layer. At the access layer, ISPs and telecom operators control customer connectivity and can bundle adjacent services. At the hosting layer, on-premise server rooms remain a substitute for small enterprises, especially when budgets are tight or legacy systems are hard to move. At the cloud layer, remote hyperscalers and regional data centers can offer lower unit compute costs and broader service catalogues. At the managed-services layer, local integrators can compete for customer relationships even if they do not own data centers. DSP’s defensible zone is where local hosting, network presence and managed continuity must be bought together.

Buyer power is likely mixed. Large public institutions, telecom-related buyers, banks, health entities and major enterprises can demand service guarantees and negotiate price because their contracts matter to a small provider. Smaller businesses have less negotiating leverage but can still choose on-premise equipment, ISP bundles or remote cloud. DSP’s ability to maintain price depends on whether customers treat local resilience and support as mission-critical or as optional insurance.

Supplier power is structurally high. OPT-NC’s wholesale and infrastructure role gives it influence over local and international connectivity economics. Power and facility inputs are local constraints. Hardware and software vendors are external. In this setting, DSP’s best procurement advantage may be group affiliation with CIPAC and relationships with local ISPs and partners, rather than its own scale. CIPAC’s broader digital ecosystem references CAN’L, DSP, Satnet, Le Cube and related services, indicating a channel environment where group relationships can matter commercially.

Regulatory change could cut both ways. A more open and competitive wholesale environment could lower DSP’s upstream costs and improve redundancy options. It could also make it easier for other service operators to enter hosting, interconnection or cloud-adjacent services. The Competition Authority’s preference for a declarative regime over administrative authorization would generally reduce entry barriers for service operators, which is good for market dynamism but not automatically good for DSP’s pricing power.

Corporate-control context and survival below scale

The most important corporate-control fact is the CIPAC connection. DSP’s brochure calls it a subsidiary of Groupe CIPAC, CIPAC articles call it a subsidiary, and the dsp.nc legal notice identifies CIPAC SA as the website administrator. CIPAC’s public communications position DSP as part of a broader local digital-infrastructure offer, with services for enterprises, public institutions, local authorities, healthcare establishments and digital actors. This parent or group context likely gives DSP commercial reach that a standalone technical host would lack.

CIPAC’s 2026 article describes DSP’s infrastructure as “invisible” but essential for local organizations, with IaaS providing hosted, backed-up, supervised and locally operated virtualized servers. That phrasing is commercially useful: it recasts hosting from a commodity into a continuity layer for the New Caledonian economy. CIPAC’s 2025 article on the second data-center extension similarly places DSP in the context of economic continuity, local digital needs, sociopolitical risk and cybersecurity. These are not neutral third-party assessments, but they reveal how the group wants to sell DSP internally to the market: as a resilience platform, not a hosting shop.

The 2018 non-dissolution notice complicates the narrative. A company whose equity fell below half of share capital had experienced material accumulated losses or balance-sheet impairment. For a small infrastructure operator, this could reflect the heavy fixed costs of data-center investment, underutilization, price pressure, slow customer adoption, or group-level restructuring. The public record does not reveal which. The later evidence of AS134405 allocation in 2018, active APNIC records, data-center expansion and new GPU services suggests that DSP survived and continued investing. But survival may have required group support, recapitalization, creditor patience or improved utilization.

This is the core below-hyperscale survival problem. Data-center economics reward high utilization and procurement scale. Small-market trust economics reward proximity, redundancy and customized support. DSP sits between the two. If utilization is high enough and customers pay a premium for local continuity, the business can survive despite limited scale. If customers treat local hosting as a commodity, DSP is exposed to gross-margin compression from suppliers above and buyers below. The public evidence points to an operator that has chosen to deepen differentiation—two data centers, backup replication, DRaaS, sovereign cloud, GPUaaS—rather than compete on generic server price.

The economic meaning of two data centers

DSP’s claim to be the only local actor with two data centers is not just a marketing line; it is a pricing argument. A single-site host sells availability within one facility. A dual-site host can sell continuity, replication, failover, backup separation and disaster-recovery planning. DSP’s brochure explicitly describes DC1-MGTA as the first production site and DC2-GAL as the second site used for disaster recovery and business continuity, with backups copied to a third site. Its SLA material includes RTO and RPO targets, which are only commercially credible when the provider can point to geographically separate infrastructure.

The small physical scale does not invalidate the strategy. In a territory with limited listed data-center capacity, a 20-rack production facility and 6-rack secondary site can be meaningful. The unit economics, however, are fragile. A second site increases fixed costs before it increases revenue. It requires space, power, cooling, access control, networking, replication equipment, monitoring and staff procedures. CIPAC’s 2025 expansion article shows precisely these investments: hosting space, access control, a new airlock and preparation space.

The second-site strategy creates an insurance-like revenue model. Customers pay recurring fees for an event they hope not to experience: outage, cyberattack, hardware failure, disaster or site loss. The provider must maintain readiness even when there is no incident. This makes trust central. A customer does not buy DRaaS only because it is cheap; it buys because it believes the provider will be there during failure. DSP’s local address, named team, facility visibility, APNIC identity and public customer disclosures all help support that belief.

The unresolved economic question is whether DSP’s two-site capability is sold mostly as high-margin assurance to many customers, or as bespoke infrastructure for a smaller number of larger accounts. The brochure’s claim of roughly 100 clients suggests a relatively broad base, but public evidence does not reveal revenue concentration. If a few major customers anchor the data centers, DSP’s economics are contract-renewal sensitive. If many mid-sized customers buy backup and hosted services, DSP has more diversified recurring revenue but may face higher support complexity.

DNS, hosted domains and the visible edge of the customer base

Public IP intelligence records only a partial view of DSP’s customer base. IPinfo reports 35 hosted domains across 11 IPs in AS134405. That is not a complete account count because many customers may use private addressing, customer-owned domains, remote DNS, VPN-only services, colocation without public IP disclosure, or IPs not easily attributed through reverse-domain collection. Still, the domain count supports the view that DSP is a modest-size hosting operator rather than a mass shared-hosting platform.

The visible legal notices matter because they are deliberate disclosures. ISEE, MDF and Nespresso New Caledonia identify DSP as host in their own public materials. In European and French-influenced legal notice culture, hosting-provider disclosure is a compliance and accountability practice. That makes DSP part of the public trust chain for those sites. The commercial implication is that DSP’s name appears where end users, regulators, vendors or litigants could find it. A host willing to be named by public-facing organizations must maintain basic reputational hygiene.

The customer evidence also suggests a channel model. DSP’s website emphasizes partners and a local digital ecosystem, while ISEE’s legal notice names SKAZY as the site creator and maintainer but DSP as the host. This is a common infrastructure split: web agencies or integrators own the application relationship, while DSP supplies hosting, backup, network and facility services. For DSP, channel relationships with agencies, ISPs, integrators and CIPAC affiliates may be as important as direct sales.

This channel model affects margins and bargaining. If DSP reaches customers through integrators, it may gain volume without large sales overhead, but it may also share margin or lose direct control of customer perception. If it sells directly to larger accounts, it can price risk more fully but must carry enterprise sales and support costs. The evidence suggests both motions: DSP has named sales and support contacts, while the ecosystem language and third-party site builders indicate indirect channels.

Security, abuse and reputation signals

Public records located for this report do not show a major DSP-specific outage, breach disclosure, litigation, procurement dispute, license sanction or public abuse controversy. That absence should be treated carefully. Small-market incidents may be handled privately, reported in French local channels not indexed broadly, or disclosed only to customers. Lack of public evidence is not proof of lack of incidents. It does, however, mean there is no obvious public scandal discount visible in the records reviewed here.

The positive security signals are operational rather than forensic. APNIC records show a validated abuse contact in 2026. RPKI status appears valid across the currently visible prefixes. DSP’s brochure promises monitoring, backup, supervision, hardware-vendor relationships, provider coordination, 24/7 support, GTI and GTR commitments, and disaster-recovery targets. CIPAC’s recent articles explicitly frame DSP in relation to cybersecurity risk and continuity. These are not independent security audits, but they show that security and continuity are central to the commercial proposition.

The weaker signal is certification ambiguity. DSP’s LinkedIn profile includes ISO27001 and HDS among specialties or keywords, but the public evidence reviewed here does not verify certification status. This matters because verified ISO 27001 or French HDS health-data hosting certification would materially expand DSP’s credibility for public-sector, healthcare and regulated workloads. A keyword on a profile is not enough. Buyers should distinguish between “security-oriented services,” “certification-aligned practices” and “audited certification.”

IP reputation appears manageable in the public sources reviewed, but not invisible. IPinfo identifies at least one IP associated with VPN usage, while also showing a small set of hosted domains, pingable infrastructure and visible traceroute paths. In a small /24-based hosting network, even minor reputation issues can have disproportionate effects because address pools are limited. The economically important point is not whether the VPN tag is harmful; it is that DSP’s address space is small enough that reputation management must be operationally active.

Alternative hypotheses and what unresolved facts would change

One hypothesis is that DSP is primarily a local sovereign-cloud and business-continuity platform. The evidence supporting this is strong: the website and brochure emphasize local hosting, legal comfort, proximity, two data centers, backup, DRaaS, BaaS and SLA commitments; CIPAC’s articles frame DSP around sovereignty, continuity and local organizations; APNIC and PeeringDB records show independent network identity and IX presence. Under this hypothesis, DSP’s value comes from local trust and infrastructure control, and the central financial question is utilization of fixed assets.

A second hypothesis is that DSP is partly a CIPAC-group infrastructure utility. The evidence is suggestive: DSP is described as a CIPAC subsidiary, CIPAC administers the website, and CIPAC’s broader digital ecosystem includes related connectivity and digital services. If this is true, DSP may not need to maximize standalone hosting margins in every service line; it can support group offerings, strengthen customer retention, and provide infrastructure depth to affiliated channels. The economic implication would be lower standalone fragility but higher dependence on group strategy.

A third hypothesis is that DSP is a technically credible but financially constrained niche operator. The 2018 non-dissolution notice supports the possibility of prior financial strain. The small IP footprint, small staff signal on LinkedIn, limited traffic scale and single visible upstream all point to a below-scale operator. Later expansion and GPUaaS investment suggest resilience, but they do not prove strong profitability. Under this hypothesis, DSP survives because it has a defensible local niche, but its margins remain exposed to utilization swings and supplier pricing.

A fourth hypothesis is that DSP’s strategic value lies more in interconnection and local facility control than in cloud services. The evidence for this is partial. PeeringDB records DSP DC1 as a facility, CAN’L IX is present there, and DSP has two 10G ports at the IX. This gives DSP a role in the local network graph. But the service materials emphasize hosting, cloud, backup and DR more than neutral colocation or carrier hotel economics. The likely reality is hybrid: facility and IX presence support the cloud product rather than replace it.

A fifth hypothesis is that DSP’s future depends on whether local demand for sovereign compute expands faster than remote-cloud substitution. The GPUaaS announcement is the test case. If New Caledonian public institutions, health entities and enterprises need local AI or data-processing capacity for sovereignty, latency or compliance reasons, DSP can capture a premium. If workloads are bursty or better served by remote clouds, specialized local compute may depress returns. The public evidence proves the investment intent; it does not prove demand depth.

What DSP reveals about small hosting economics in New Caledonia

DSP shows that in small island markets, address-resource visibility is not a technical footnote. It is part of the commercial product. An ASN, APNIC organisation record, portable IPv4 block, IPv6 allocation, RPKI-valid routes, abuse contact, PeeringDB profile and IX presence collectively make a local host legible to customers and counterparties. That legibility supports trust, and trust supports pricing power.

DSP also shows that upstream dependence is the hidden governor on local cloud economics. The company can own racks, servers and IP resources, but it cannot escape the economics of wholesale connectivity and international reach. Public routing data showing one upstream through OPT-NC, combined with regulatory evidence of OPT-NC’s monopoly role in public telecommunications services, places DSP inside a constrained supplier structure. Local IX participation improves the position; it does not remove the dependency.

The firm further shows how customer trust becomes a substitute for hyperscale breadth. DSP cannot offer the global service catalogue of a hyperscaler. It can offer proximity, local legal framing, a named support team, visible data-center locations, disaster recovery inside New Caledonia, and local customer references. For some workloads, those attributes are worth more than the marginal cost advantage of remote cloud. For other workloads, they are not. DSP’s survival depends on segmenting the market correctly.

Finally, DSP shows that below-hyperscale survival is path-dependent. The company’s early legal continuity, old ASN, newer AS134405 routing identity, CIPAC affiliation, data-center investments, 2018 balance-sheet stress signal and later expansion all matter. Small infrastructure firms survive not by achieving global scale, but by accumulating enough local proof points that customers hesitate to switch: known address, known engineers, known facility, known routes, known backups, known legal notices, known parent context. Each proof point is small. Together they create a local infrastructure moat.

The moat is not permanent. It can be eroded by cheaper international capacity, stronger remote-cloud adoption, regulatory liberalization, new local facilities, supplier price shocks, power instability, cyber incidents, loss of key staff, or customer concentration. But as of the public evidence reviewed here, DSP remains a visible and active local infrastructure operator whose economic significance exceeds its raw address count.

Evidence ledger

APNIC ORG-DSP1-AP organisation record: identifies Data Services Pacific as an APNIC Local Internet Registry in New Caledonia, with Nouméa address and dsp.nc contacts.
APNIC AS134405 record: identifies DATASERVICESPACIFIC-AS-AP, Data Services Pacific, country NC, organisation ORG-DSP1-AP, and validated abuse/IRT contact information.
APNIC 203.34.36.0/24 record: identifies DATASERVICESPACIFIC-NC as an assigned portable IPv4 block and names Glenn Penin as a network contact.
APNIC 103.123.232.0/24 record: identifies DATA-SERVICES-PACIFIC-DC1 and describes the block as Data Services Pacific DC1 Nouméa.
APNIC IPv6 allocation record: identifies 2404:e9c0::/32 as a Data Services Pacific IPv6 allocation.
BGP.tools AS134405: gives routed-prefix set, RPKI-valid status, registration date, upstream/peer data and CAN’L IX participation.
BGP.tools AS24053: shows historical Data Services Pacific ASN, registered in 2005 and not currently in the global routing table.
APNIC AS24053 record: confirms AS24053 as DATASERVICESPACIFIC-AS-AP with Data Services Pacific organisation reference.
IPinfo AS134405: provides address counts, RPKI status, hosted-domain count, upstream/peer summary, traceroute observations and stub-AS characterization.
WhatIsMyIP AS134405 record: corroborates six Data Services Pacific IP ranges in New Caledonia.
PeeringDB Data Services Pacific network profile: identifies ASN 134405, website, open peering policy, traffic band, heavy outbound ratio and CAN’L IX entries.
PeeringDB Data Services Pacific organisation profile: identifies DSP, address at 210 Rue Roger Gervolino, and associated facility/network.
PeeringDB DSP DC1 facility profile: records DSP DC1 in Nouméa, support and sales contacts, CAN’L IX presence and networks at the facility.
PeeringDB CAN’L IX profile: records the Nouméa exchange, five peers, seven connections, capacity, local facility and DSP’s two 10G connections.
Data Center Map Nouméa listing: lists DSP-DC1 and DSP-DC2 with addresses and rack counts.
DSP official website: describes DSP as Nouméa-based, 100% private, a hosting/cloud actor with two data centers, and lists services, local-hosting arguments, team and contact details.
DSP legal notice: identifies CIPAC SA as site administrator, gives CIPAC corporate details, and states the site is hosted by D.S.P. in a New Caledonian data center.
DSP brochure PDF: describes DSP as a CIPAC Group subsidiary, presents services, two data centers, third backup site, SLA commitments, client-count claim, support model and service lines.
DSP brochure infrastructure diagram screenshot: shows the production, disaster-recovery and off-site backup topology with OPT, CAN’L, Lagoon and IXP connectivity references.
DSP brochure technology-partner screenshot: shows APC, VMware, Microsoft SPLA, Veeam, Dell EMC, Synology, Cisco and related vendor ecosystem references.
CIPAC article, June 2026: describes DSP as a CIPAC subsidiary providing hosted, backed-up, supervised and locally operated IaaS for local companies, public services and organisations.
CIPAC article, March 2025: describes DSP’s second data-center extension and frames it around resilience, cybersecurity and local economic continuity.
CIPAC article on GPUaaS: states DSP acquired two AI-dedicated servers, planned monthly rental GPUaaS, and lists IaaS, BaaS, DRaaS and private-cloud services.
DSP LinkedIn company profile: gives market-channel description, company size signal, founded date, specialties, locations and employee names.
New Caledonia registry notice, 2011: records DATA SERVICES PACIFIC SARL, RCS Nouméa B 759 779, capital, address and managers.
New Caledonia registry notice, 2014: records DATA SERVICES PACIFIC, same RCS identity and management change.
New Caledonia registry notice, 2018: records non-dissolution after equity fell below half of share capital.
CCI 2024 electoral list: records DATA SERVICES PACIFIC as a current legal person in the services category.
Nespresso New Caledonia legal notice: identifies DSP as website host.
MDF legal notice: identifies Data Services Pacific as host.
ISEE legal notice: identifies Data Services Pacific as host of the official statistical institute website.
New Caledonia Competition Authority English summary: describes OPT-NC’s monopoly over public telecommunications services, open Internet access market and regulatory competition concerns.
Market summary on New Caledonia telecoms: describes OPT-NC’s dominant role in telecoms and wholesale ISP services.
OPT-NC submarine-cable article: describes Gondwana-1 and the second cable project for capacity and resilience.
SubmarineNetworks Gondwana-1 profile: identifies Gondwana-1 as a 2,151 km New Caledonia–Australia cable owned and operated by OPT.
APNIC blog on New Caledonia: provides local ISP history, including CAN’L, CIPAC Group involvement, MLS, Telenet and Nautile.
DB-IP New Caledonia allocation table: gives relative IPv4 allocation scale among OPT, OFFRATEL, CAN’L, Micro Logic, Nautile, TeleNet and DSP.
HE.net New Caledonia network list: gives comparative ASN adjacency and route counts for New Caledonian networks, including DSP and local competitors.

Watchpoints

A second upstream or non-OPT transit path. If AS134405 gains another upstream beyond AS18200, DSP’s supplier dependence and resilience profile would materially improve. If it remains single-upstream, local IX participation remains useful but not sufficient to change the international-connectivity bargaining structure.
Changes in OPT-NC wholesale pricing, cable capacity or regulatory obligations. Lower international bandwidth cost could reduce the local-hosting premium by making remote cloud more attractive. Better wholesale competition could also improve DSP’s margins and redundancy options. The direction depends on whether cheaper connectivity benefits DSP more as a buyer or hyperscalers more as substitutes.
Public verification of ISO 27001, HDS or comparable certifications. Verified certification would change DSP’s addressable market for public-sector, health and regulated workloads. Marketing keywords alone do not have the same economic effect.
DC2 and third-site transparency. Independent facility records, certifications, customer references or peering entries for DC2 would strengthen the disaster-recovery thesis. Evidence that the third site is only a limited backup location would narrow the resilience premium.
RPKI or APNIC contact drift. Invalid ROAs, expired abuse contacts or inconsistent registry data would damage the trust value created by DSP’s public resource identity. Continued validation supports counterparty confidence.
GPUaaS utilization. Sustained customers for local GPU compute would signal a new sovereign-compute niche. Low utilization would turn the servers into fast-depreciating capex and reveal the limits of local AI demand.
Customer concentration disclosures or major public-sector wins. A large anchor customer would improve utilization but increase renewal risk. A diversified set of mid-market backup, DRaaS and IaaS customers would create more resilient recurring revenue.
CIPAC ownership or financing changes. Formal confirmation of shareholder support, recapitalization or integration into a broader CIPAC digital platform would lower standalone survival risk. Separation from group support would make DSP’s fixed-cost structure more exposed.
New local data-center entry. A rival facility with carrier-neutral positioning, public-sector certification or stronger upstream diversity would pressure DSP’s scarcity premium. If no rival emerges, DSP’s two-site claim remains commercially powerful.
Hyperscaler edge, CDN or private-connect expansion near New Caledonia. Better regional cloud reachability could erode DSP’s generic compute demand. It could also increase demand for local hybrid hosting if DSP becomes the local continuity and interconnection partner.
Public incident record. A major outage, cyber incident or abuse event would be more damaging to DSP than to a hyperscaler because the trust premium is central to its economics and its address pool is small.
Post-crisis local enterprise IT spending. DSP’s business improves if New Caledonian organizations prioritize continuity, backup and outsourced infrastructure. It weakens if budget pressure pushes buyers toward deferred upgrades, low-cost remote SaaS or minimum-maintenance on-premise systems.