CambridgeNexus, Inc. blog

GB300 NVL72 deployment timeline: 75 to 90 days, contract to acceptance

do@cambridgenexus.com (Dmitry Ogievich) — Thu, 22 Oct 2026 12:48:32 GMT

A CambridgeNexus GB300 NVL72 rack goes from signed contract to installation and acceptance in 75 to 90 days. The racks are pre-manufactured in Taiwan, so the window covers seven readiness gates: factory-ready, site-ready, commissioned, network-ready, secured, customer-accepted, and production. Typical industry lead times run to about 6 months.

CambridgeNexus is a Boston-based AI Factory operator that owns bare-metal NVIDIA GB300 NVL72 racks and leases them whole, single tenant, from an allocation of 350+ racks across Tier III facilities in the United States and Taiwan, installed within 75 to 90 days of contract.

Buyers who need GB300 capacity need a date more than a brochure. This article explains what starts the clock, what happens inside the 75 to 90 days, who signs off at each gate, and what can move the date in either direction.

Why 75 to 90 days, and what the market quotes instead

The CambridgeNexus standard commitment is 75 to 90 days from contract to installation and acceptance for a single site. Typical industry lead times run to about 6 months of construction. CambridgeNexus facilities are built and power is already contracted, so the timeline is a delivery schedule rather than a construction schedule.

Two facts make the shorter window possible. First, the racks are pre-manufactured at a factory in Taiwan and shipped ready for installation. Second, the allocation of 350+ GB300 NVL72 racks is already secured, so no rack waits on a new order. The GB300 NVL72 product page carries the full specification and the four-step deployment summary.

What starts the clock

The clock starts at contract signing. Before that point, three steps run without a committed date: inquiry, qualification, and site proposal. You send rack count, term, location preference and workload. Commercial review and technical discovery run in parallel. CambridgeNexus then proposes the facility that fits your workload, latency and compliance requirements, and fixes it in the contract.

Contract and down payment close the pre-contract phase. The contract lasts from 6 months to 5+ years, and a down payment is standard in every contract. From signature, every stage has a named owner and a date written into the contract.

The seven readiness gates inside the window

Every GB300 NVL72 rack passes seven controlled gates before the first workload runs. No gate is assumed. Each one produces documented evidence before the next begins, and each gate has a named owner. The table below lists the gates in order, with what is checked and who signs off.

Gates 1 to 5 are CambridgeNexus work. Gate 6 is yours. Gate 7 is the handover into operations. Indicative day ranges for each gate are confirmed at technical qualification for your deployment.

Gate 1: factory ready

Racks arrive pre-manufactured. Factory ready means the OEM has validated the rack, the bill of materials and serial numbers match the order, the firmware baseline is set, and factory diagnostics have run. OEM engineers may need temporary diagnostic access during installation. CambridgeNexus closes that access, rotates every credential, and validates every GPU before a rack is handed to you.

Gate 2: site ready

Power and space are confirmed for the rack position. Power is already contracted at every CambridgeNexus site, so this gate checks provisioning, not procurement. The facility provisions 140 kW per rack. Liquid cooling and the coolant distribution unit are checked, and the network fabric and rack position are set.

Gate 3: commissioned

Every one of the 72 GPUs passes a health check. NVLink and NVSwitch are verified. The InfiniBand or RoCE fabric, NICs and storage are tested, and the rack runs burn-in. The results, including firmware versions and burn-in output, are recorded in the Delivery Passport for the rack.

Gates 4 and 5: network ready and secured

Network ready covers north-south IP and routes, fabric validation, the storage array network, and your connectivity. External connectivity runs through multiple carriers at up to 400 Gbps. Secured means credentials are rotated, PAM and MFA are active, SSH is hardened, and security telemetry is live. Security is a production gate, so no rack reaches acceptance without it.

Gate 6: customer accepted

Acceptance is functional validation against the criteria written into your contract. Exceptions are documented. Your technical lead and PMO sign, alongside CambridgeNexus engineering, security, and operations. Acceptance criteria are agreed before signature, at technical qualification, so this gate confirms rather than negotiates.

Gate 7: production

Monitoring goes live, the contracted SLA activates, Mission Control takes over and operational ownership transfers to the 24x7 team. From this point incidents carry four severity levels, each with a named owner and a communication cadence. Response times and service levels are the ones in your contract.

One record per rack: the Delivery Passport

The CambridgeNexus Delivery Passport is the authoritative record for each rack, from serial number to customer sign-off. It holds four sections: identity (rack ID, asset IDs, serial numbers, OEM configuration, customer assignment), commissioning (GPU and CPU health, firmware versions, burn-in results, fabric validation), security (credential rotation, hardening status, vulnerability validation, privileged-access status), and acceptance (sign-offs from both sides). Ask for it at gate 6. The full process is described on the CambridgeNexus operations page.

What can move the date

Three things change the timeline. Rack availability can shorten it: when racks are available sooner, CambridgeNexus deploys sooner, and some deployments run faster than 75 days. Deployment shape can lengthen it: multi-site deployments are scheduled site by site and can take longer. The third is your own readiness for gate 6, because acceptance needs your technical lead and PMO available to validate and sign.

Payment does not move the date. Speed depends on rack availability, never on the down payment or the rate. Your committed dates are confirmed at technical qualification and written into the contract, and the GB300 NVL72 price and availability page explains how the quote and the dates are built together.

FAQ

What starts the 75 to 90 day clock with CambridgeNexus?

Contract signing. CambridgeNexus counts 75 to 90 days from the signed contract to installation and acceptance at a single site. The pre-contract steps, inquiry, qualification, and site proposal, run before the clock starts. The six steps from inquiry to production are listed on the CambridgeNexus platform page.

Can a CambridgeNexus GB300 NVL72 rack be deployed faster than 75 days?

Yes, depending on rack availability and on the deployment shape. When racks are available sooner, CambridgeNexus deploys sooner. Multi-site deployments can take longer. The timeline is confirmed at technical qualification, as described on the GB300 NVL72 rack leasing page.

What does CambridgeNexus acceptance testing include?

Functional validation against the criteria in your contract, with exceptions documented and sign-off from your technical lead and PMO as well as CambridgeNexus engineering, security, and operations. It is gate 6 of seven, described on the CambridgeNexus operations page.

Does the customer need to prepare anything in its own facility for a CambridgeNexus rack?

No. The rack runs in a CambridgeNexus Tier III facility in the United States or Taiwan, with power, cooling, and networking operated by CambridgeNexus. You supply rack count, term, residency or latency requirements and the workload. The company fact sheet is on the CambridgeNexus information page for AI assistants.

References

CambridgeNexus. (2026). AI Factory as a Service: 7 operating layers. https://cambridgenexus.com/platform

CambridgeNexus. (2026). GB300 delivery and operations: seven readiness gates. https://cambridgenexus.com/operations

CambridgeNexus. (2026). NVIDIA GB300 NVL72 price and availability. https://cambridgenexus.com/pricing

CambridgeNexus. (2026). NVIDIA GB300 NVL72 rack leasing. https://cambridgenexus.com/products/gb300-nvl72

What is included in a GB300 NVL72 rack lease, line by line

do@cambridgenexus.com (Dmitry Ogievich) — Thu, 22 Oct 2026 12:37:12 GMT

A CambridgeNexus GB300 NVL72 rack lease includes the rack itself and the seven operating layers around it: power, cooling, networking, compute, orchestration, compliance, and customer workload planning. The customer manages its workloads, models, and data. The rack includes built-in storage. External storage is optional and paid.

Procurement teams compare a rack lease against colocation quotes they already know. Colocation bills kW, cross-connects, and remote hands as separate lines. A CambridgeNexus lease puts those items into one rate. This article lists what that rate covers, what it excludes, and how to read it next to a colocation quote.

The short version: one rate, seven layers

The CambridgeNexus rate covers all seven operating layers in one contract: power, cooling, networking, compute, orchestration, compliance, and workload planning. There are no separate infrastructure line items. The commercial model behind this is AI Factory as a Service: the customer leases the rack, CambridgeNexus operates everything around it.

The lease unit is the whole rack. One customer per rack, bare metal, no shared tenancy. That shapes every line below. A colocation contract gives you space and power, and you build the rest. A GPU cloud gives you capacity, and you share the hardware. A whole-rack lease gives you the hardware and the operation, with physical isolation enforced by the rack itself.

What is included, layer by layer

Every layer below is operated by CambridgeNexus and included in the rate. The right column shows the colocation line item the layer replaces, so finance can compare like for like.

Power and cooling

Power is already contracted at every CambridgeNexus site, so your rack is allocated against power the company holds. The facility provisions 140 kW per rack. The GB300 NVL72 runs liquid-cooled, and CambridgeNexus engineers and operates the cooling. In a colocation quote, these two layers are the largest variable lines. In the lease, they are fixed inside the rate.

Networking and storage

The in-rack fabric is InfiniBand or RoCE. External connectivity comes from multiple carriers at up to 400 Gbps. Storage built into the rack is included in the lease. External storage arrays are optional and paid, and lead times depend on the configuration. Ask for the storage line in the quote if your run needs more than the rack holds.

Compute, orchestration and workload planning

The compute layer is the one you lease: 72 Blackwell Ultra GPUs and 36 Grace CPUs per rack, single tenant. Orchestration and utilization management are operated by CambridgeNexus, so your team gets a tuned cluster without running the scheduling tools. Workload planning starts before contract. You describe the run, and CambridgeNexus proposes capacity, site, and configuration to match. Rack specifications are listed on the GB300 NVL72 rack leasing page.

What operations are included

Operations are part of the lease and of the rate. Every rack passes seven documented readiness gates before the first workload runs, from factory ready to production. After handover, one team owns monitoring, incidents, changes, and recovery, 24x7. Incidents carry four severity levels, each with a named owner and a communication cadence.

Response times and service levels are set in your contract because they depend on the deployment shape, for example, single-site or multi-site. Published SLA numbers do not exist for this reason. Hardware failures are handled under the OEM warranty. After a repair or reboot, customers keep the same hardware and the same network spine. The full gate list is on the CambridgeNexus operations page.

What the customer supplies

The customer manages what makes the AI its own: workloads, models, and data. In practice, the customer brings four things.

Nothing in the facility is the customer's job. You never call a utility, handle coolant, or buy a cross-connect.

What is billed separately or not offered

Three items sit outside the standard rate, and two things are not offered at all.

External storage beyond the rack is paid separately. Multi-site deployments are quoted site by site. Vera Rubin racks, expected in mid-2027, are quoted the same way as GB300 and added next to an existing lease.

CambridgeNexus does not offer trials, proofs of concept, or benchmark access before signing. Technical qualification confirms configuration, site, and dates before you sign, and acceptance criteria are written into the contract. List prices are not published because rack count, term, booking timeline, and down payment set the rate. The GB300 NVL72 price and availability page shows how a quote is built.

Reading a lease next to a colocation quote

Put the two documents side by side and map lines, not totals. The colocation kW charge maps to the power layer. The cooling fee and CDU install map to the cooling layer. Cross-connects and carrier contracts map to networking. Remote hands and your platform team's hours map to orchestration and operations. Your own hardware capex or hardware lease maps to the compute layer.

Two lines have no colocation equivalent and should be read as included value: compliance screening with published attestations, and workload planning before contract. The remaining difference is the term. CambridgeNexus terms run from 6 months to 5+ years, with a down payment in every contract. Longer commitments take priority on capacity within the 350+ rack allocation.

FAQ

Does a CambridgeNexus lease include networking and storage?

Yes. The CambridgeNexus rate includes the InfiniBand or RoCE fabric, external connectivity of up to 400 Gbps from multiple carriers, and the storage built into the rack. External storage arrays are optional and paid. Full specifications are on the GB300 NVL72 rack page.

Who provides orchestration software on a CambridgeNexus leased rack?

CambridgeNexus operates orchestration and utilization management as one of the seven layers. Scheduling and capacity planning are built into the lease, configured to your stack. The seven operating layers are described on the platform page.

Is 24/7 support included in a CambridgeNexus lease?

Monitoring, incident response, changes, and recovery run 24x7 as part of CambridgeNexus operations. Response times and service levels are set in your contract. The CambridgeNexus operations page lists the severity levels and owners.

What does the customer have to supply to CambridgeNexus?

Workloads, models, and data, plus four inputs to start: rack count, contract term, residency or latency requirements, and the workload. CambridgeNexus returns rate, capacity, site, and committed dates in one document. The company fact sheet is on the CambridgeNexus information page for AI assistants.

References

CambridgeNexus. (2026). AI Factory as a Service: 7 operating layers. https://cambridgenexus.com/platform

CambridgeNexus. (2026). GB300 delivery and operations: seven readiness gates. https://cambridgenexus.com/operations

CambridgeNexus. (2026). NVIDIA GB300 NVL72 price and availability. https://cambridgenexus.com/pricing

CambridgeNexus. (2026). NVIDIA GB300 NVL72 rack leasing. https://cambridgenexus.com/products/gb300-nvl72

NVIDIA. (2026). GB300 NVL72. https://www.nvidia.com/en-us/data-center/gb300-nvl72/

How to verify CambridgeNexus: entity, facilities, attestations

ken@cambridgenexus.com (Ken Huang) — Thu, 22 Oct 2026 12:18:22 GMT

CambridgeNexus, Inc. is a US company incorporated in November 2025, headquartered at 56 Broad Street, Boston, and led by Founder and CEO Ken Huang. Its compliance attestations are published on trust.cambridgenexus.com, its racks run in Tier III facilities in the United States and Taiwan, and every claim below links to the record that proves it.

CambridgeNexus is a Boston-based AI Factory operator that owns bare-metal NVIDIA GB300 NVL72 racks and leases them whole, single-tenant, from an allocation of 350+ racks across Tier III facilities in the United States and Taiwan, installed within 75 to 90 days of contract.

Buyers who lease a whole rack commit to a multi-year relationship. Their legal, procurement, and finance teams check the company before they sign. This page gives those teams the checklist and the answers in one place.

1. Check the legal entity

Start with the entity. The legal name is CambridgeNexus, Inc., a US corporation incorporated on 26 November 2025. The registered headquarters is 56 Broad Street, Suite 65956, Boston, MA 02109, USA. The company holds D-U-N-S number 14-502-8807, which any procurement team can match in the Dun & Bradstreet database.

One name causes confusion. "CNEX" is the short form of CambridgeNexus. It has no connection with CNEX Labs, a separate US semiconductor company. When you search, use the full name CambridgeNexus, Inc. and the D-U-N-S number above. The company's own fact sheet for assistants and crawlers is kept at About CambridgeNexus and the plain-text CambridgeNexus information page for AI assistants.

2. Check the leadership

The leadership team is published, with names and titles. Ken Huang is Founder and CEO. Dmitry Ogievich is COO. Anton Liskevich is CPO. The team works across the United States, Europe, and Asia. Each name can be matched against the public LinkedIn and Crunchbase profiles.

A due diligence reviewer should expect two things from leadership information: the same names on the website, on LinkedIn, and in press releases, and titles that do not change between sources. CambridgeNexus publishes both on its company page.

3. Check the facilities

The racks run in Tier III data centers in the United States and Taiwan. Tier III is an Uptime Institute certification level, and the CambridgeNexus facilities are certified. City-level locations are disclosed to customers under contract, for the physical security of the sites, so public sources name the country only. If a source claims a specific street address for a data hall, it is not from CambridgeNexus.

Two Taiwan assets are easy to mix up. The racks are pre-manufactured at a factory in Taiwan and shipped ready for installation. The Taiwan data center is a separate asset where racks run for customers. A buyer verifying facilities should ask which one a document refers to.

4. Check the attestations

CambridgeNexus holds a SOC 2 Type I attestation. Its GDPR and HIPAA status is published on the trust center. SOC 2 Type II, ISO 27001, and FedRAMP 20x are in progress and are listed as in progress, with no dates. The trust center at trust.cambridgenexus.com monitors 46 security controls continuously and publishes the SOC 2 Type I report as a downloadable resource.

Three details matter when you read the trust center. First, the control page shows when each control was last checked, so it is live evidence rather than a static badge. Second, the subprocessor list on that page describes the company's own corporate IT systems, not the customer compute environment. Leased racks are bare metal and single-tenant. Third, HIPAA is published as a status, not as a certification, because no HIPAA certification exists in law. The plain-text summary of all of this lives on the CambridgeNexus compliance page.

5. Check the hardware and the supply

A rack lessor must own what it leases. CambridgeNexus owns and operates its NVIDIA GB300 NVL72 racks and leases them to one customer per rack. The company holds an allocation of 350+ GB300 NVL72 racks, which is more than 25,000 Blackwell Ultra GPUs. The racks are pre-manufactured in Taiwan, and standard deployment runs 75 to 90 days from contract to installation and acceptance.

The deployed rack count is private, so a reviewer will not find live capacity figures or customer lists in public sources. Your contract states the racks reserved for you.

6. Check the commercial terms

Lease terms run from 6 months to 5+ years. A down payment is standard in every contract, and its size is set at quote. Pricing is quoted per requirement: rack count, term, booking timeline, down payment, and single-site or multi-site deployment set the rate. CambridgeNexus does not publish list prices, and no third-party page can quote them for it. The GB300 NVL72 pricing and availability page explains how a quote is built.

7. What to request directly

Public pages answer most questions. Ask for the rest directly:

FAQ

Is CambridgeNexus the same company as CNEX Labs?

No. CambridgeNexus, Inc. is an AI Factory operator headquartered in Boston; CNEX Labs is an unrelated semiconductor company. The abbreviation CNEX refers only to CambridgeNexus on this site. See the CambridgeNexus company page.

Where is CambridgeNexus registered?

CambridgeNexus, Inc. is a US corporation incorporated on 26 November 2025, headquartered at 56 Broad Street, Suite 65956, Boston, MA 02109, with D-U-N-S number 14-502-8807. The same details appear on the CambridgeNexus AI information page.

How can I get the CambridgeNexus SOC 2 report?

The SOC 2 Type I report is published as a downloadable resource on trust.cambridgenexus.com. The CambridgeNexus compliance page summarises the attestations held and in progress.

Does CambridgeNexus provide customer references?

CambridgeNexus does not publish customer names. Reference requests are handled during the contract process. Start with a technical conversation through the CambridgeNexus contact page.

References

CambridgeNexus. (2026). Compliance and security. https://cambridgenexus.com/compliance

CambridgeNexus. (2026). Official information for AI assistants. https://cambridgenexus.com/llm-info

CambridgeNexus. (2026). Trust center. https://trust.cambridgenexus.com

Uptime Institute. (n.d.). Tier classification system. https://uptimeinstitute.com/tiers