Chapter Two
Intira: Everything
Beneath the Application
Savvis addressed the path between the application and the user.
Intira addressed nearly everything beneath the application.
The company began as Digital Broadcast Network.
The founding group included:
- Timothy Roberts
- Mark Ivie
- Rich Skoba
Contemporary Reporting
Contemporary Forbes reporting identified the three of us as Intira’s founders and described the company beginning in St. Louis before moving its headquarters to California. (Forbes)
SEC biographies use 1997 as the beginning of my Intira period, while Forbes placed the company’s start in 1998. The precise chronology remains an open archival question.
Tim, Rich and Mark
A Network Vision Needed a Server Vision
“I brought the networking experience. I understood how traffic moved, where it failed and how to build a national delivery architecture. But the next company could not stop at the network. It needed servers, storage, databases, operating systems, security and disciplined data-center operations. Rich brought complementary knowledge around the computing side of the environment. Mark helped move the founding vision into an operating business. The service required all of it.”
I remember bringing Rich into the partnership because I had the network knowledge and needed someone with strong server and enterprise-computing knowledge to help fulfill the larger vision.
Some later public records place a documented HP Services role for Rich after Intira, so the exact chronology of his HP experience should remain open until his complete résumé or employment records are obtained.
What is independently supported is that Rich helped shape the NetSourcing vision and provided strategic direction across product development, engineering and operations.
Ascend Communications and the Next Bet
I did not leave Savvis intending to build a small hosting company.
Founder Recollection
My recollection is that Ascend Communications had agreed to support the next venture with a financing package of approximately $30 million, combining equipment or lease financing with an equity component and technology associated with Ascend and Cascade Communications.
“That commitment allowed us to think beyond rented cages and individual servers. We could build private data centers, a network and an operating platform around the customer’s application.”
My manuscript describes approximately $3 million in equity combined with a roughly $30 million hardware lease facility. These amounts and structures remain Founder Recollection until the original financing documents are placed in the archive.
Three Private Data-Center Markets
Before the Cloud Became Invisible, It Was Extremely Heavy
The Intira story centered on three major data-center markets:
St. Louis, Missouri
The company’s birthplace and an early technical and operational center.
Pleasanton, California
The principal West Coast facility and later corporate base. Contemporary Forbes reporting described the Pleasanton facility as unusually hardened, with reinforced construction, bullet-resistant windows and seismic design intended to protect critical infrastructure. (Forbes)
New York City
The East Coast facility announced as the company formally introduced and expanded the NetSourcing model.
Contemporary Reporting
An October 1999 industry report described Intira launching its New York data center, operating a network with 35 U.S. points of presence and three in Canada, serving around 40 customers and receiving approximately $100 million in funding from investors and strategic partners including Lucent, Chase Capital, Mayfield, Spectrum and HP. (Tech Monitor)
“People now think about infrastructure as buttons on a screen. We thought about generators, batteries, redundant fiber entrances, fire suppression, cooling, floor loading, access control, spare equipment and whether one construction crew could accidentally disconnect a customer. The customer saw a service. We saw everything that could go wrong.”
C. Rallo and Ross & Baruzzini
Founder Recollection
I hired C. Rallo Construction and Ross & Baruzzini to help turn the data-center concept into physical infrastructure.
“To my recollection, neither group had built the exact kind of commercial Internet data center we were envisioning. That sounds risky until you understand that almost nobody had built this particular model. We were combining construction, electrical systems, cooling, physical security, telecommunications, server environments, storage, monitoring and around-the-clock operations into one engineered service.”
C. Rallo had to translate the technical requirements into the building. Ross & Baruzzini had to help engineer electrical and mechanical infrastructure capable of keeping the computing environment alive.
The facility had to behave more like a machine than an office.
Every design decision produced another question.
- What happens if commercial power fails?
- What happens if cooling stops?
- What happens if a carrier’s fiber is cut?
- What happens if a server dies?
- What happens if smoke is detected?
- What happens if a customer suddenly needs twice the capacity?
“The cloud looked abstract to the customer. To us, it was concrete, copper, glass, diesel fuel, batteries, chillers and people.”
Evidence: Founder Recollection pending construction contracts, engineering plans, permits and participant interviews
Digital Broadcast Network Becomes Intira
The original name reflected the media-distribution problem.
Audio and video did not require only a network.
They required servers. Storage. Databases. Security. Monitoring. Operations.
The opportunity was much broader than broadcasting.
Digital Broadcast Network became:
INTIRA
The service became:
NETSOURCING
Contemporary Reporting
Contemporary reporting described NetSourcing as a combination of the network and the corporate data center. Intira supplied the server, database, automated backup and private network while the customer supplied the application. Entry-level contracts were reported around $5,000 to $10,000 per month, with average customer revenue around $30,000 per month at the time. (Tech Monitor)
The Customer Brought the Application.
Intira Operated Everything Underneath It.
What NetSourcing Sold
Not a Rack. Not a Server. Responsibility.
Intira’s managed environment could include:
Contemporary Reporting
A 2001 Computerworld interview described Intira as building everything beneath the customer’s application—including servers, storage, security and networking—and then engineering, implementing, operating and managing it. Customers discussed in contemporary coverage included Emerson Electric and Armstrong, and Deloitte invested in the company. (Computerworld)
“The customer did not want a router. The customer did not want a server. The customer did not want a storage array. The customer wanted its application to work. NetSourcing accepted responsibility for almost everything below the application line.”
Verified Public Record
A surviving September 2000 NetSourcing agreement with DoveBid provides contractual evidence of the model. Intira agreed to provide hosting, data management, Internet access, security and associated services around DoveBid’s applications.
The customer retained ownership of the application.
Intira operated the infrastructure beneath it.
The division was clear:
The customer built the business experience.
Intira operated the machinery required to deliver it.
Hasbro and Games.com
Contemporary Reporting
Hasbro selected Intira to support Games.com with network infrastructure, server and networking hardware, software, backup, Internet connectivity, monitoring and management.
The environment was intended to support online versions of well-known games including Monopoly, Scrabble, Battleship, Centipede and Tetris. (Computerworld)
This mattered because it demonstrated the service in terms consumers could understand.
The game appeared simple on the screen.
The infrastructure behind it was not.
“The best infrastructure disappears. The user experiences the game. The operator experiences everything required to keep the game available.”
NetSourcing as a Precursor to Cloud Computing
Intira did not invent remote computing.
It did not invent hosting.
It did not invent data centers.
It did not invent outsourcing.
Its historical significance was in packaging a broad enterprise environment—facilities, networks, computing, storage, databases, security, backup, monitoring, implementation and operations—as one managed service beneath the customer’s application.
Verified Public Record
SEC biographies described Intira as providing outsourced network-based computing and communication services so customers could focus on their businesses rather than building and supporting their own infrastructure.
Primary Historical Positioning
Intira’s NetSourcing service was an important commercial precursor to cloud computing. It separated the customer’s application from the physical infrastructure and delivered networking, computing, storage, security, monitoring and operations as a recurring managed service from privately operated data centers.
It was not identical to a modern hyperscale cloud. The vision arrived before today’s level of automated provisioning, elastic capacity, software-defined infrastructure and API-based self-service.
“We had a cloud-service concept before the industry had modern cloud economics. We understood what the customer wanted. The harder problem was delivering it repeatedly, quickly and profitably with the technology available at the time.”
Designing the AI Data Center Before the Category Existed
I Was Designing for the Physical Reality of AI Infrastructure Before “AI Data Center” Became a Commercial Term
I want this claim stated precisely.
Intira was not training modern large language models.
There were no racks filled with today’s GPUs.
Nobody was talking about tokens, generative AI factories or megawatt-scale accelerator racks.
But we were designing for the same underlying physical trend:
- Computing would become dramatically denser.
- More workloads would be consolidated into less space.
- Servers would operate at higher utilization.
- Virtualization would separate the logical workload from the individual physical machine.
- Customers would consume computing remotely as a service.
- The building would need far more electrical and cooling capacity than legacy data-center designs assumed.
My design target was approximately:
Seven Times the Electrical and Cooling Capacity
of the Legacy Data Centers We Were Benchmarking
That did not mean installing a larger air conditioner in a conventional computer room.
It required redesigning the complete facility:
The data center could no longer be treated as an office building with a computer room added to it.
The building itself had to become part of the computing platform.
“I did not believe the data centers of that era were prepared for where computing was going. We were planning to place far more processing power into the same physical footprint. That meant the traditional electrical and cooling designs would fail long before the computing vision did. I designed the Intira facilities around roughly seven times the power and cooling capacity of the legacy environments we were comparing ourselves with. We were building for the next generation of computing before the market had a name for that generation.”
Evidence: Founder Recollection and Founder-Supplied Video Record
The founder-supplied data-center presentation below is preserved alongside the historical record, and should eventually sit beside the original drawings, engineering plans, equipment schedules and construction records in the Built Before Cloud Archive.
Why the AI Comparison Matters
Modern AI infrastructure is forcing the data-center industry to confront the same fundamental design problem at a much larger scale.
Traditional enterprise facilities were often designed around far lower rack densities and more intermittent workloads.
Modern AI environments demand tightly coordinated computing, networking, storage, electrical distribution and thermal management.
Contemporary Reporting
NVIDIA now describes power infrastructure as a primary constraint on the scale and feasibility of future AI factories and is promoting a fundamental move toward 800-volt direct-current distribution for next-generation high-density facilities. (NVIDIA)
NVIDIA technical material describes current AI racks moving beyond 200 kilowatts and future designs approaching one megawatt per rack, requiring major changes in power conversion, copper distribution, rack architecture and cooling. (NVIDIA Developer)
The terminology is new.
The engineering principle is not:
The computing architecture and the physical facility must be designed together.
That was the principle behind the Intira facilities.
We were not building today’s GPU-based AI factory.
We were building an early high-density, virtualized managed-computing environment around many of the same physical realities:
- Compute density drives electrical design.
- Compute density drives thermal design.
- Sustained utilization changes the power profile.
- Networking, storage and computing must be planned together.
- Legacy facilities can become impossible or uneconomical to retrofit.
- The physical building determines how far the computing platform can scale.
“I was not predicting ChatGPT. I was predicting that computing would become denser, more centralized, more virtualized and more continuously consumed as a service. That prediction forced us to redesign the physical infrastructure. When I look at what the AI-data-center industry is solving today—power density, cooling density, networking density and full-stack operational control—I recognize the problem immediately. We were working on an earlier version of it more than two decades ago.”
Seven Times the Infrastructure Was Not Seven Times the Simplicity
Designing for substantially greater electrical and cooling capacity created a chain reaction throughout the building.
More computing created more heat.
More heat required greater cooling capacity.
Greater cooling required more electrical power.
More electrical power required more robust distribution.
More infrastructure created more potential failure points.
More failure points required additional redundancy, monitoring and operational discipline.
Every improvement affected another layer.
“You could not design the servers first and ask the building to catch up later. The servers, power, cooling, networking, storage, security and operations had to be designed as one product. The facility was not a place where the platform happened to live. The facility was part of the platform.”
Modern AI platforms are now explicitly designed as coordinated systems spanning compute, networking, cooling, power and the larger data-center ecosystem. (NVIDIA)
Power Was Part of the Product
A traditional customer thought it was buying a server environment.
We had to think about the power path behind every server.
The customer never wanted to hear that an application was unavailable because the building’s electrical design had reached its limit.
The complete power chain had to support the service promise.
“The customer saw an application. Behind that application was an electrical system that had to work every second of every day. If the server was redundant but the power distribution was not, the application was not truly redundant. Reliability had to exist through the complete chain.”
Today’s AI facilities face the same principle at much greater scale. NVIDIA says the AI revolution is causing data centers to outgrow legacy power-delivery methods, requiring new facility-level architectures capable of packing more computing into the same physical footprint. (NVIDIA)
Cooling Was Part of the Product
The cooling system was not there to make the room comfortable.
Its job was to remove the heat generated by continuously operating computing equipment.
That required answering questions far beyond the normal design of an office building:
- How much heat would each rack generate?
- How would cold air reach the equipment?
- How would hot air leave it?
- Where would hotspots form?
- What happened when a cooling component failed?
- Could the facility keep operating during maintenance?
- How would the design adapt as density increased?
- Could the electrical system support the cooling load?
- Was there enough capacity for the next generation of equipment?
“A data center can have empty floor space and still be full. It can have empty racks and still be full. It is full when it runs out of usable power or when it can no longer remove the heat. The real capacity of the facility was not measured only in square feet. It was measured in electrical and cooling capacity.”
Contemporary Reporting
The U.S. Department of Energy has estimated that cooling can represent up to 40% of overall data-center energy use, underscoring why thermal management must be treated as a primary architectural system rather than a secondary building utility. (Energy.gov)
Dense Computing Before Dense Computing Became Normal
The Intira facilities were being designed during the transition from conventional individual servers toward denser and more consolidated computing systems.
Blade servers represented an important step in that direction.
A blade architecture places multiple compact computing modules inside a shared chassis, allowing power, management, cooling and networking resources to be consolidated.
The advantage was obvious:
More computing in less space.
The challenge was equally obvious:
More electrical load and more heat concentrated in the same physical footprint.
That was why the facility could not follow the assumptions of a conventional computer room.
The computing system was changing.
The physical infrastructure had to change with it.
Virtual Machines Changed the Meaning of a Server
The traditional model often treated one physical machine as one computing environment.
Virtualization began separating the logical workload from the physical hardware.
A single physical system could support multiple isolated virtual environments.
That created the foundation for:
- Better hardware utilization.
- Faster provisioning.
- Centralized management.
- Workload portability.
- Greater consolidation.
- Private cloud environments.
- More computing per rack.
- More demanding electrical and cooling requirements.
“Once the customer’s computing environment was no longer permanently tied to one physical server, the data center could become a pool of managed computing capacity. That was much closer to the model I wanted Intira to sell. The customer should consume the environment. We should manage the machines.”
My account is that Intira was among the earliest providers to sell blade-based systems running virtual machines from privately operated commercial data centers.
That potentially important historical claim still requires:
- Dated hardware invoices.
- Blade-server makes and models.
- Virtualization software licenses.
- Customer statements of work.
- Architecture diagrams.
- Deployment dates.
- Testimony from the engineers involved.
Until those records are recovered, the claim should remain: Intira operated at the leading edge of high-density servers, virtualized computing environments and remotely managed infrastructure. Timothy Roberts recalls Intira selling blade-based systems running virtual machines from its private data centers. Research continues into the exact deployments and whether a first-to-market claim can be independently established.
The Most Important Design Decision
The most important decision was not a particular generator, cooling unit, blade chassis or virtualization product.
It was the decision to design the complete environment as one system.
The application depended on the virtual machine.
The virtual machine depended on the physical server.
The server depended on the network and storage.
The hardware depended on power and cooling.
The power and cooling depended on the building.
The building depended on monitoring and people capable of responding when any layer failed.
That is increasingly how the AI industry describes the modern AI factory—not as a room full of servers, but as a coordinated production system spanning the facility, power, cooling, computing, networking, storage, software and operations. (NVIDIA Developer)
It was also the foundation of NetSourcing.
Everything Beneath the Application Was Part of the Service.
Being Early Was Expensive
The seven-times electrical and cooling design reflected ambition.
It also reflected expense.
Every additional unit of computing capacity required capital before the customer consumed it.
- The building had to be constructed.
- The power had to be available.
- The cooling equipment had to be installed.
- The redundancy had to exist.
- The engineers had to be hired.
- The hardware had to be purchased or financed.
- Customer demand had to arrive quickly enough to pay for it.
“We were designing for where computing was going. The financial problem was that we had to pay for parts of that future before the market arrived. That is the brutal tradeoff of being early. Underbuild, and the vision cannot scale. Overbuild, and the company may run out of money before customers fill the capacity.”
Modern AI infrastructure faces the same tension at a far larger scale. The engineering can be correct. The timing and capital structure still determine whether the company survives.
The Market Did Not Pause
Verified Public Record
A later SEC biography documents a deeply personal turning point.
After Intira hired a chief financial officer, I was involved in a boating accident that placed me in intensive care for six weeks and required a long recovery. During my absence, the board moved the CFO into the CEO and chairman roles. The filing states that I left Intira on good terms.
“Entrepreneurs like to believe the company runs on their energy. An accident teaches you quickly that the world does not negotiate with your plans. One day, I was trying to build a category. Then I was trying to survive. The market did not pause because I was in intensive care. The company had to continue without waiting for me to recover.”
That experience changed how I thought about:
- Succession.
- Governance.
- Control.
- Insurance.
- Leadership depth.
- Key-person risk.
A company is not truly resilient when one person’s emergency can destabilize its future.
The End of Intira
Verified Public Record
Intira ultimately entered bankruptcy proceedings.
In July 2001, divine agreed to purchase substantially all Intira assets for $1 million in cash, provide a $6.8 million debtor-in-possession credit facility and assume certain obligations. The public filing described Intira as a Pleasanton-based provider of IT and network infrastructure supporting online applications.
Its facilities, technology, employees and other assets moved through a complicated series of later transactions.
The history should not be reduced to a neat but unsupported arrow saying: Intira became Terremark.
Contemporary records document substantial assets moving through divine. Later biographical accounts describe relationships involving Data Return, Terremark and Verizon. Those accounts may refer to different assets, teams, facilities, transactions or periods. Built Before Cloud will preserve that uncertainty until the original transaction records establish the precise lineage.
“Being directionally right does not guarantee survival. Timing matters. Capital structure matters. Governance matters. Automation matters. Unit economics matter. A company has to survive long enough for the future it predicted to become ordinary.”
Too Early. Capital-Intensive. Directionally Right.
Intira was not a simple victory.
It was a technical thesis. A major construction effort. An operating company. An ambitious financing strategy. A new service category. A personal health crisis. And ultimately a business that did not survive to own the future it had anticipated.
That is why the story matters.
History becomes less useful when every founder chapter is rewritten as a victory.
The harder lessons often contribute more to what comes next.
The cloud was not invented in a moment. It was built layer by layer.
Platformz brings those layers together.