By Russell Brandom
Enriched and Expanded Edition
Main Facts
As the artificial intelligence revolution forces a massive, unprecedented physical expansion of global digital infrastructure, data center developers are bracing for staggering capital expenditures. Industry forecasts indicate that total spending on data center construction and modernization could reach as high as $4 trillion by the end of the decade.
However, this unprecedented gold rush faces severe physical constraints. Developers are no longer limited merely by capital or silicon availability; they are heavily restricted by geopolitical hurdles, stringent zoning regulations, and, most critically, the crippling limitations of local power grids. Finding parcels of land that feature both the necessary electrical capacity and regulatory clearance has turned the site-selection process into a high-stakes bottleneck.
While most industry stakeholders treat the speed of fiber-optic communication as an immutable law of physics, a bold new bet is emerging that could fundamentally alter the geographical mathematics of data center buildouts.
Enter Relativity Networks, a pioneering connectivity firm that announced $22 million in funding on Tuesday via Simple Agreements for Future Equity (SAFE notes). The round was spearheaded by prominent venture capital entities, including Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC. Simultaneously, the company revealed a massive $40 million follow-on commercial order from a leading hyperscale cloud provider that has chosen to remain anonymous.
At the core of Relativity Networks’ value proposition is hollow-core fiber, a rarely deployed, highly specialized technology that enables data transmission at speeds roughly 30% faster than conventional fiber-optic cables. While traditional fiber channels light through solid glass strands, hollow-core fiber transmits the exact same light signals through an engineered vacuum chamber located at the center of the line. By removing the physical glass medium from the light path, the technology brings data transmission vastly closer to the absolute theoretical limit of the speed of light.
This breakthrough addresses a creeping architectural crisis in modern AI infrastructure. As computing clusters scale from a single rack of GPUs to sprawling campuses spanning hundreds of acres and dozens of distinct buildings, physical latency—the microscopic delay inherent in data transit—has re-emerged as a major performance bottleneck. By shaving crucial microseconds off transit times, Relativity Networks believes it can decouple massive compute installations from localized power constraints, allowing separate data center campuses to link up and operate as a single, unified, synchronized machine.
Chronology
To understand how hollow-core fiber moved from a scientific curiosity into a commercially viable venture-backed enterprise, it is necessary to examine the timeline of infrastructure scaling and telecommunications development over the past decade.
-
Pre-2020: The Era of Pure Compute Optimization
During the dawn of the modern deep learning boom, artificial intelligence infrastructure was overwhelmingly concentrated on silicon optimization. The industry ethos was strictly focused on raw processing power: "GPU, GPU, GPU." Data centers were compact enough that inter-rack latency via standard copper or short-run fiber was negligible, and little attention was paid to wide-area or campus-scale fiber networking efficiency. -
2020–2023: The Networking Bottleneck Within the Four Walls
As Large Language Models (LLMs) scaled from billions to hundreds of billions of parameters, single-rack architectures became obsolete. Training clusters expanded to occupy entire rooms and floors within data centers. This phase forced the industry to aggressively optimize internal networking—pioneering high-speed InfiniBand switches, advanced optical transceivers, and low-latency internal topologies to keep thousands of GPUs synchronized without stalling. -
Late 2023–Early 2024: The Power Grid Crisis and Spatial Constraints
As generative AI adoption exploded, power requirements skyrocketed. Individual data center footprints grew from tens of megawatts to hundreds of megawatts—and increasingly, gigawatt-scale campuses were demanded. Utilities across the United States and Europe began buckling under the strain. Developers found themselves sitting on billions of dollars in capital with no local substations available to feed their server farms. The industry hit a geographical wall: they needed to build where power existed, even if that meant fragmenting operations across non-contiguous real estate. -
Tuesday’s Funding and Commercial Milestones (Current Event)
Relativity Networks formally announced its $22 million SAFE note financing round, backed by Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC. Alongside the capital infusion, the company disclosed a $40 million follow-on order from an unnamed hyperscaler, signaling that major cloud providers are moving past proof-of-concept testing and aggressively procuring hollow-core infrastructure for deployment.
Supporting Data and Technical Architecture
The physics governing optical telecommunications have remained largely unchanged for decades, but hollow-core fiber represents a radical paradigm shift in material science and photonics.
Traditional Fiber vs. Hollow-Core Fiber
In standard fiber-optic cables, data is transmitted by pulsing laser light through a core made of ultra-pure silica glass. While glass is remarkably transparent, it possesses a refractive index greater than one. This means light travels through glass at roughly two-thirds of its maximum speed in a vacuum.
- Standard Fiber Latency: According to Relativity Networks CEO Jason Eisenholz, a digital signal traveling through conventional glass fiber takes approximately five microseconds per kilometer.
- Hollow-Core Fiber Latency: By replacing the solid glass core with an air- or vacuum-filled channel surrounded by a microscopic photonic crystal cladding, light is guided without interacting with dense matter. This reduces transit time to approximately three and a half microseconds per kilometer—a performance gain of roughly 30%.
The Mathematics of Sprawl
While a savings of 1.5 microseconds per kilometer may sound negligible to the casual observer, it is epochal in high-performance computing and distributed AI training.
Distributed training algorithms require constant, synchronous communication between massive clusters of accelerators. If the physical distance between two clusters introduces too much latency, the processors spend idle cycles waiting for gradient updates, severely degrading training efficiency.
By increasing transmission speed by 30%, Relativity Networks effectively expands the geographical radius within which data centers can communicate without performance degradation. In latency terms, developers can now span 30% larger distances before latency crosses critical performance thresholds.
Financial and Market Indicators
- $4 Trillion: Projected cumulative global spending on data center construction and modernization through the end of the decade.
- $22 Million: Total capital secured by Relativity Networks in its latest SAFE note investment round.
- $40 Million: Value of the follow-on commercial procurement order secured from a leading, undisclosed hyperscale cloud provider.
- 30%: The performance margin—both in speed enhancement and expanded geographical reach—that hollow-core fiber holds over legacy glass infrastructure.
Official Responses and Industry Perspectives
The commercial validation of hollow-core fiber is reflected not just in the venture capital backing, but in the strategic decisions being made by the world’s largest cloud operators.
Reflecting on the evolution of industry bottlenecks, CEO Jason Eisenholz breaks down the historical progression of AI scaling into three distinct eras in an interview with TechCrunch:
"The first era of AI optimized for compute," Eisenholz explained. "It was GPU, GPU, GPU. The second era optimized the networking inside the data center to take advantage of that compute. The third era that we see coming is optimizing the geography."
This shift in perspective is driven entirely by the pragmatic realities of site selection and energy procurement. As power availability dictates where massive digital infrastructure can physically be established, cloud operators are frequently forced to split massive workloads across disparate plots of land.
"The largest systems are distributing the compute across multiple campuses to reach the power that exists," Eisenholz noted. "They’re moving to where the warm shell is, but they still need to operate as one synchronized machine."
Investors share this vision. Rhapsody Venture Partners and Bell Ventures, known for backing deep-tech and foundational physical infrastructure, view ultra-fast networking not merely as a telecommunications play, but as an indispensable enabler of next-generation artificial intelligence. By bridging the physical gap between fragmented real estate assets, investors believe hollow-core technology will unlock stranded power capacity worldwide.
Implications
The commercialization and deployment of hollow-core fiber carry profound implications for the future of commercial real estate, energy markets, geopolitics, and artificial intelligence development.
1. Easing the Grid Bottleneck
The most immediate impact will be felt in energy markets. By enabling non-contiguous data center campuses to act as a single, unified hyper-cluster, developers are no longer forced to secure single, monolithic plots of land adjacent to multi-gigawatt power sources. Instead, they can site smaller server installations across multiple rural or suburban locations—wherever localized pockets of renewable energy or surplus grid capacity exist—and stitch them together with hollow-core pipelines.
2. Redefining Real Estate Dynamics
For decades, data center real estate has been hyper-concentrated in established, power-constrained "Flap-D" markets (Frankfurt, London, Amsterdam, Paris, Dublin) and primary U.S. hubs like Northern Virginia (Data Center Alley). If multi-campus deployment enabled by ultra-fast fiber becomes standard practice, the geographic definition of a "prime data center market" will expand outward. Secondary and tertiary markets with abundant green energy access could suddenly become viable extensions of primary metro computing hubs.
3. Unlocking Next-Generation AI Scaling
As frontier model parameters push past tens of trillions, the sheer physical size of training clusters must expand. Without breakthroughs in transmission speed, the speed-of-light delay within copper and glass networks would eventually place a hard physical ceiling on how large a single training run could grow. Hollow-core fiber pushes that horizon further out, buying the AI industry critical time as researchers work toward more efficient architectures.
4. Telecommunications and Infrastructure Disruption
For legacy fiber providers and telecom giants, the rise of hollow-core technology poses both a threat and an acquisition opportunity. While traditional glass fiber networks represent trillions of dollars in sunk capital, the performance advantages of air-core transmission for enterprise, financial, and hyperscale AI traffic are undeniable. As manufacturing processes mature and production costs decline, hollow-core fiber is poised to transition from a niche ultra-low-latency tool for high-frequency trading into the high-speed backbone of the global artificial intelligence economy.
For further coverage on emerging enterprise technologies, platform policy, and artificial intelligence infrastructure, contact Russell Brandom at [email protected] or via Signal at 412-401-5489.