A Munich startup's €1.6 million bet on glass storage arrives just as energy-starved data centers search for alternatives to spinning disks and tape migrations
There's something almost absurd about the pitch: write your data into glass with lasers, unplug everything, and walk away for a millennium or two. No cooling systems humming through the night. No drives spinning down and back up. No migration anxiety every five years when tape formats evolve.
ewigbyte, a seven-month-old startup based near Munich, closed €1.6 million in pre-seed funding last month betting that absurdity is actually inevitability. The round, led by Vanagon Ventures and Bayern Kapital, attracted a roster of finance and energy veterans—Johann von Wersebe from Morgan Stanley, Oliver Engels formerly at Deutsche Börse, Christopher Delbrück who navigated CFO roles at Uniper and the now-defunct air taxi venture Lilium. Germany's federal innovation agency SPRIND, which typically backs technologies verging on science fiction, also validated the concept.
The timing is less whimsical than it appears. Data centers are headed toward a reckoning with physics and power grids alike. The International Energy Agency projects electricity consumption will hit roughly 945 terawatt-hours by 2030—more than double the 2024 baseline—with the U.S. and China driving 80 percent of that surge. Archival storage sits at the uncomfortable center of that equation: mountains of cold data that enterprises and cloud giants must preserve for decades, spinning on disks or migrating across tape generations, each approach bleeding energy and operational costs year after year.
Traditional architectures weren't built for this. CERN's particle physics archive crossed one exabyte in 2025, predominantly stored on IBM tape libraries that ingested about 50 petabytes last August alone—1.7 petabytes daily. Tape shipments keep shattering records despite decades of "tape is dead" predictions; the LTO consortium reported 176.5 exabytes shipped in 2024 (compressed), up 15.4 percent from the prior year and marking the fourth straight year of growth. Those volumes jumped from 152.9 exabytes in 2023 and 148.3 the year before.
But tape demands periodic migrations to newer formats. Even offline media means powering libraries for writes, reads, and integrity checks that never quite stop. Hard drives carry a 1.36 percent annualized failure rate, according to Backblaze's dataset spanning 344,000 drives last year—modest individually, catastrophic at scale without redundancy and refresh cycles. Cloud archive tiers from AWS, Azure, and Google promise tape-like economics but ultimately rely on spinning rust or tape in distant hyperscale facilities, with retrieval times stretching into hours and egress fees that create quiet lock-in.
Brad Johns, an analyst tracking archival economics, figures stored data will breach 40 zettabytes by 2035. His projections suggest up to 62 percent of archival shipments that year could shift toward tape or emerging passive technologies just to control costs and energy. The IEA's electricity forecasts—coupled with grid constraints already surfacing in Northern Virginia and Frankfurt—are pushing "zero-power-at-rest" higher on procurement checklists, perhaps faster than vendors expected.
When Glass Stops Being a Metaphor
ewigbyte enters a market suddenly crowded after years of dormancy. The company, founded July 31, 2025, brings together Dr. Steffen Klewitz (CEO, previously COO at Cerabyte with a photonics background), Dr. Ina Dorothee von Haeften handling operations, and Philipp Wittwer on technology. Their approach centers on direct laser ablation—etching data patterns straight into uncoated glass surfaces rather than relying on ceramic coatings or photosensitive layers. The pitch: electromagnetic pulse resistance, heat and water immunity, radiation tolerance, and multi-millennial durability that makes decades-old tape look fragile.
The same day ewigbyte announced its round, Microsoft Research published a peer-reviewed breakthrough in Nature detailing Project Silica's ability to write 4.8 terabytes onto a 12 cm × 12 cm × 2 mm borosilicate glass tile using femtosecond lasers. The paper—"Laser writing in glass for dense, fast and efficient archival data storage"—demonstrated 10,000-year extrapolated lifetimes through accelerated aging and simplified readout using a single camera instead of complex polarization optics. Microsoft had shown the concept back in 2019 by archiving the original "Superman" film on quartz glass, but the shift to cheaper borosilicate represents a step closer to commercial viability, assuming commercial means anything in archival storage economics.
The technologies diverge in implementation but converge on promise: passive media requiring zero power between write and read events, durability measured in millennia rather than migration cycles, and resistance to threats that corrupt magnetic or semiconductor storage. ewigbyte's system design emphasizes modular automation—robotic handling paired with optical read/write heads—to scale from petabytes to exabytes. Early specifications suggest roughly 10 gigabytes per glass tablet (writing both sides), with per-head throughput around 500 megabytes per second and system-level ingest rates reaching 4 gigabytes per second through parallelism.
Those are projections, naturally.
The Crowded Race Nobody Noticed
ewigbyte faces competition that arrived quietly and is now accelerating. Cerabyte, another German startup expanding into the U.S., uses ultraviolet femtosecond lasers and digital micromirror arrays to write data into ceramic nanolayers on glass tablets. The company claims 5,000-year durability and targets 100 petabytes per rack by 2030, with media costs potentially dropping below $1 per terabyte—aspirational figures that would upend storage economics if remotely achievable. Cerabyte's pilots report one petabyte per rack and approximately 100 megabytes per second ingest, positioning the technology for hyperscale cold storage if the physics and manufacturing scale.
In the UK, Sphotonix—a University of Southampton spin-out—pursues "5D memory crystals" in fused silica discs with theoretical capacities reaching 360 terabytes per disc and durability claims extending to billions of years under lab conditions. Current prototypes write at roughly 4 megabytes per second and read at 30 megabytes per second, however, which sits far below enterprise adoption thresholds. Sphotonix raised about $4.5 million and targets technology readiness level 6 demonstrations with a licensing model rather than hardware sales, suggesting commercialization remains distant.
DNA storage represents an entirely different trajectory. Twist Bioscience spun out Atlas Data Storage last May with a $155 million seed round, launching the "Eon 100" service claiming 60 petabytes per liter density and thousands-year durability. Synthesis and sequencing costs remain prohibitive for most use cases, but the density advantage is undeniable—Atlas markets the solution as 1,000 times denser than LTO-10 tape. Early-access customers are exploring high-value archival applications where floor space constraints justify premium pricing, assuming those customers exist beyond press releases.
Sony's Optical Disc Archive Gen3 represents the mature baseline in this landscape: 5.5-terabyte cartridges with approximately 100-year media life, write speeds around 187 megabytes per second, read speeds at 375 megabytes per second, and established enterprise integration. The technology is write-once-read-many, EMP-resistant, and actively supported with software updates through January 2026. It occupies a niche between tape's raw capacity and glass storage's durability claims, though it still requires powered libraries and periodic migration—an operational reality that glass vendors promise to eliminate.
The Throughput Problem
Enthusiasm around glass storage confronts hard questions about speed, economics, and ecosystem maturity. Microsoft's Project Silica breakthrough achieved peak write speeds around 65.9 megabits per second per system in some configurations—roughly 8 megabytes per second—with demonstrations showing approximately 3.13 megabytes per second sustained. Those figures, impressive for research, sit orders of magnitude below the multi-gigabyte-per-second ingest rates that hyperscalers achieve with modern tape libraries or disk arrays. Massive parallelism—deploying dozens or hundreds of write heads simultaneously—could theoretically close that gap, but at what capital cost and reliability profile?
Dr. Klewitz has framed the challenge differently in press interviews, suggesting cold data archival is fundamentally "a time and throughput problem more than a density problem." Fair enough. The company's positioning emphasizes European data sovereignty and resilience against electromagnetic events, themes that resonate with regulations like the EU Data Act (applicable since September 2025) and eIDAS 2.0's introduction of "Qualified Electronic Archiving" as a trust service. These regulatory tailwinds may create niches where premium archival solutions justify higher acquisition costs through reduced operational expenditure and compliance benefits—or they may simply add complexity without shifting procurement dollars.
Tape's continued strength complicates the narrative for glass entrants. LTO-10 cartridges shipping in January 2026 from Fujifilm deliver 40 terabytes native capacity (100 terabytes compressed) with mature library automation and decades of operational expertise baked in. CERN, the National Energy Research Scientific Computing Center, and hyperscalers continue expanding tape infrastructure precisely because it works at exabyte scale today, not tomorrow. The value proposition for glass storage must overcome not just technical specifications but switching costs, operator training, and ecosystem lock-in that favors incumbents who've spent decades getting boring things right.
Early adoption will likely concentrate in verticals where longevity and immutability carry outsized value: national archives preserving cultural heritage, financial institutions meeting multi-decade retention mandates, media companies safeguarding master recordings, and scientific repositories storing irreplaceable datasets. GitHub's Arctic Code Vault—which archived a February 2020 snapshot on Piql film in a Svalbard permafrost mine—demonstrates demand for millennium-scale preservation, even if that particular use case represents symbolism more than operational necessity.
The 2026 Inflection
ewigbyte's €1.6 million pre-seed positions the company to refine prototypes and secure pilot customers, but commercialization will require follow-on capital and proof points at scale—ideally customers willing to discuss deployments publicly rather than under NDA. The startup's technical differentiation—direct ablation into uncoated glass rather than ceramic layers or complex 5D encoding—may simplify manufacturing and reduce per-bit costs if the physics holds at production volumes. That's a considerable if.
The involvement of Bayern Kapital and SPRIND validation signals official German interest in cultivating storage technology sovereignty, a strategic consideration as data residency requirements proliferate and hyperscaler dominance raises questions in Brussels and Berlin. Whether that translates into procurement preferences or simply conference presentations remains unclear.
The broader glass storage cohort faces what feels like an inflection point this year. Microsoft's Nature publication lends academic credibility and demonstrates that borosilicate—far cheaper than fused silica—can achieve archival-grade durability. Cerabyte's U.S. expansion and 100-petabyte-per-rack roadmap will test whether ceramic-on-glass can meet hyperscale performance thresholds. Sphotonix's licensing strategy and Atlas Data Storage's early-access DNA service will clarify whether exotic densities find market fit despite throughput and cost constraints.
The energy argument remains compelling, perhaps more compelling than vendors realize. If data center operators face power-capped expansion and carbon reduction mandates—and they increasingly do—passive media that writes once and consumes zero watts for decades could reshape total cost of ownership calculations in ways spreadsheets haven't captured yet. Whether that potential translates into displaced tape shipments or creates a new tier above existing cold storage depends on execution: write speeds, media reliability at scale, automation robustness, and supply chain scalability that doesn't break when order volumes jump from hundreds to hundreds of thousands of units.
ewigbyte and its peers have secured the capital to attempt those proofs. The question is whether the physics and economics cooperate before the next round of investor skepticism arrives—or before data centers find less exotic ways to solve the power problem.
