Three layers under the compute curve: per-die thermal personalization shipping today, fiber-photonic acceleration on the roadmap, and a carbon nanomaterial supply chain manufactured on U.S. soil.
Agentic AI demands roughly 100× more compute than the industry expected a year ago. The chips that meet that demand burn through their thermal budget before they hit their compute budget, move data through copper traces that can’t keep up with photonic-grade bandwidth, and ship through supply chains the U.S. doesn’t control.
Three walls. Three answers. One platform.
“The amount of computation we need at this point as a result of agentic AI, as a result of reasoning, is easily 100× more than we thought we needed this time last year.”
JENSEN HUANG, NVIDIA
Each pillar attacks a different wall: thermal answers the cooling crisis, photonic answers the electron ceiling, and a domestic supply chain answers the sovereignty exposure. They share one feedstock and one manufacturing footprint.
Per-die personalization of GPU thermal stacks. Proprietary TIM and targeted micro-fin arrays placed exactly where each chip overheats. Shipping today.
Carbon-based fiber and photonic components for chip-to-chip interconnect and on-die optical compute. Higher bandwidth, lower energy-per-bit than copper.
U.S.-onshored carbon nanomaterial supply chain: feedstock, growth, and device fabrication on American soil. CHIPS-aligned, defense-eligible.
The wedge product: GPU-Personalized Thermal. We image each die’s heat signature, lay down a proprietary thermal interface material exactly where it overheats, and bond targeted micro-fin arrays on top, recovering 8–15°C of junction headroom without changing the rack, the cold plate, or the workload.
Two commercial tracks running in parallel: a hands-on thermal personalization service for existing GPU fleets, and a product line of carbon components for advanced chip designs.
For: Data centers · hyperscalers · OEM reference designs
Per-die FLIR profiling, proprietary TIM, and custom micro-fin arrays applied to each GPU exactly where it overheats. White-glove engagement; zero rack-architecture change.
You commit: Cluster access (or shipped silicon) and a named technical owner.
We commit: 5-day per-die turnaround at volume; 8–15°C junction drop; co-published operational data.
Engagement: NRE + per-unit, or managed-service contract.
For: Chip designers · OEMs · ASIC programs · photonic integrators
Custom CNT transistors, special-compound interconnects, photonic primitives, and integration support for leading-edge chips. Catalog parts for prototyping; custom design-ins for production.
You commit: Spec sheet, design intent, or BOM target.
We commit: Sample turnaround on catalog parts; NRE quote on bespoke designs; full integration support through qualification.
Engagement: Catalog pricing, custom NRE, or recurring supply.
Copper interconnects are the next thermal and bandwidth bottleneck in AI accelerators. Every chip-to-chip hop costs energy as heat, and the wires can’t move data fast enough to keep tensor cores fed. Photonic compute swaps electrons for photons in the dense data paths. Same silicon, fundamentally different physics.
Our development track focuses on three carbon-enabled photonic primitives: low-loss optical fiber and waveguides, carbon-nanotube photodetectors, and chip-edge optical interfaces designed to drop into existing co-packaged-optics roadmaps from NVIDIA, Broadcom, and Marvell.
Maturity: early R&D. Carbon nanomaterials demonstrate measured advantages in photonic switching speed and on-chip absorption versus conventional photonic platforms. Productization follows our thermal revenue.
Carbon is one of the rare advanced-compute substrates where the United States can build a supply chain from scratch rather than negotiate with overseas incumbents. Every step of our pipeline runs on American soil, from raw material sourcing through manufacturing to final integration. The domestic stack is the asset; the products are what it enables.
Every customer tier starts U.S.-based and DoD-eligible. We climb from data-center thermal contracts into hyperscaler co-development, then into OEM reference designs, with photonic components opening a parallel design-in track as the technology matures.
Operators and inventors who’ve shipped at Amazon, VeriFone, Space Dynamics Laboratory, and Wavetronix, paired with capital strategists who’ve raised across the U.S., U.K., India, and Asia-Pacific.
Actively recruiting 3 advisor profiles: published CNT physicist, retired semiconductor industry executive, and DARPA / government-contracts veteran.