Wilson

@WilsonSignals

Joined September 2026

29 Following

9 Followers

29 Posts

WilsonSignals retweeted

Future Markets Research

@FutureMarketsRT

5 days ago

$CLGN just posted the first peer-reviewed external validation that photonic computing can outrun the GPU by five orders of magnitude. That claim is no longer a pitch deck thesis. It is ACM-published science, co-authored with one of Europe's premier supercomputing institutions. ⚡ LPU architecture projects 40x to 80,000x speedup vs. GPU on sparse linear solvers. 🔬 Peer-reviewed paper published in 23rd ACM Computing Frontiers proceedings, Sep 24. 🤝 Joint research with HLRS Stuttgart, Germany's first federal HPC center, est. 1996. 🌐 Vision: CPU + GPU + LPU hybrid stack, each handling its optimal workload class. 📅 LightSolver commercial availability target: LPU platform ready in 2028. https://t.co/TkWvqDirkT The LightSolver story inside CLGN is a structural inflection, not a product release. When CollPlant completed the LightSolver acquisition on September 3, it did not simply add a new subsidiary. It repositioned itself at the intersection of deep-tech computing and advanced biotechnology. The two verticals share almost no customers, no go-to-market infrastructure, and no underlying science. What they share is a Nasdaq ticker, a leadership team, and a thesis: that the same company can build world-class IP in biological fabrication and computational physics simultaneously. The ACM paper published September 24 is the first piece of external scientific evidence that the computing side of that thesis has independent credibility. The specific technical claim requires careful reading. The projections, 40x to more than 80,000x acceleration over GPU algorithms, come from an emulator of LightSolver's LPU architecture, not from measurements on physical LPU hardware. The variance in that range is real: the speedup depends heavily on the algorithm and problem structure. For highly structured sparse systems, optical computing can exploit the physical properties of light in ways that silicon processors cannot. For less structured problems, the advantage narrows. This is standard photonic computing physics, not a caveat that diminishes the result. The paper validates the architectural concept, establishes a peer-reviewed methodology, and provides the scientific foundation for the next stage: validation on physical LPU hardware at HLRS-class supercomputing environments. The HLRS partnership carries strategic weight beyond the benchmark numbers. HLRS is not a startup lab or a friendly collaborator. It was established in 1996 as Germany's first federal supercomputing center, operates infrastructure used by academic researchers and industrial partners across aerospace, energy, and engineering, and publishes in top-tier venues. The HLRS Director framing the LPU as a potential component in future hybrid computing architectures is not a marketing quote. It is an academic endorsement with downstream procurement implications. LightSolver's existing commercial validation program includes Boeing, a top-tier global financial institution, leading defense companies, and oil and gas organizations. Moving those relationships from evaluation pilots to embedded deployment is the commercial arc the ACM paper is designed to accelerate, and CollPlant's next series of announcements will show whether that arc is tracking. https://t.co/TkWvqDirkT _ Disclosure: This content is not financial or investment advice. CollPlant Biotechnologies Ltd. is a paid subscriber to Wall Street Wire's coverage and distribution platform. Full disclosure: https://t.co/XfvLkFMKXu

FutureMarketsRT's tweet photo. $CLGN just posted the first peer-reviewed external validation that photonic computing can outrun the GPU by five orders of magnitude. That claim is no longer a pitch deck thesis. It is ACM-published science, co-authored with one of Europe's premier supercomputing institutions.

⚡ LPU architecture projects 40x to 80,000x speedup vs. GPU on sparse linear solvers.
🔬 Peer-reviewed paper published in 23rd ACM Computing Frontiers proceedings, Sep 24.
🤝 Joint research with HLRS Stuttgart, Germany's first federal HPC center, est. 1996.
🌐 Vision: CPU + GPU + LPU hybrid stack, each handling its optimal workload class.
📅 LightSolver commercial availability target: LPU platform ready in 2028.

https://t.co/TkWvqDirkT

The LightSolver story inside CLGN is a structural inflection, not a product release. When CollPlant completed the LightSolver acquisition on September 3, it did not simply add a new subsidiary. It repositioned itself at the intersection of deep-tech computing and advanced biotechnology. The two verticals share almost no customers, no go-to-market infrastructure, and no underlying science. What they share is a Nasdaq ticker, a leadership team, and a thesis: that the same company can build world-class IP in biological fabrication and computational physics simultaneously. The ACM paper published September 24 is the first piece of external scientific evidence that the computing side of that thesis has independent credibility.

The specific technical claim requires careful reading. The projections, 40x to more than 80,000x acceleration over GPU algorithms, come from an emulator of LightSolver's LPU architecture, not from measurements on physical LPU hardware. The variance in that range is real: the speedup depends heavily on the algorithm and problem structure. For highly structured sparse systems, optical computing can exploit the physical properties of light in ways that silicon processors cannot. For less structured problems, the advantage narrows. This is standard photonic computing physics, not a caveat that diminishes the result. The paper validates the architectural concept, establishes a peer-reviewed methodology, and provides the scientific foundation for the next stage: validation on physical LPU hardware at HLRS-class supercomputing environments.

The HLRS partnership carries strategic weight beyond the benchmark numbers. HLRS is not a startup lab or a friendly collaborator. It was established in 1996 as Germany's first federal supercomputing center, operates infrastructure used by academic researchers and industrial partners across aerospace, energy, and engineering, and publishes in top-tier venues. The HLRS Director framing the LPU as a potential component in future hybrid computing architectures is not a marketing quote. It is an academic endorsement with downstream procurement implications. LightSolver's existing commercial validation program includes Boeing, a top-tier global financial institution, leading defense companies, and oil and gas organizations. Moving those relationships from evaluation pilots to embedded deployment is the commercial arc the ACM paper is designed to accelerate, and CollPlant's next series of announcements will show whether that arc is tracking.

https://t.co/TkWvqDirkT

_
Disclosure: This content is not financial or investment advice. CollPlant Biotechnologies Ltd. is a paid subscriber to Wall Street Wire's coverage and distribution platform. Full disclosure: https://t.co/XfvLkFMKXu

WilsonSignals retweeted

The Market Observer

@TheMarkObserver

5 days ago

$CLGN: 40x to 80,000x. That is the projected performance gap between LightSolver's Laser Processing Unit and a GPU, for the workload class that dominates computational fluid dynamics, structural mechanics, and materials science simulations. https://t.co/YaGDOwcY5T Those numbers come from a peer-reviewed ACM study published September 24, 2026, conducted jointly by LightSolver, a CollPlant subsidiary, and HLRS, the High-Performance Computing Center Stuttgart, which was established in 1996 as Germany's first federal supercomputing center. Researchers benchmarked an emulator of LightSolver's LPU architecture against established iterative GPU algorithms across large-scale sparse linear equation problems. The workload is not niche. Linear systems underlie some of the world's most computationally intensive science: molecular dynamics, electromagnetics, finite element analysis. These problems dominate HPC runtime and energy cost. LightSolver's vision is a hybrid stack where CPUs handle orchestration, GPUs handle parallel tensor compute, and LPUs handle the linear algebra bottlenecks optically. The next update will reveal whether a physical LPU system can replicate these emulator-projected speedups at the scale HLRS operates, the test that determines whether this architecture moves from scientific validation to procurement. _ Disclosure: This content is not financial or investment advice. CollPlant Biotechnologies Ltd. is a paid subscriber to Wall Street Wire's coverage and distribution platform. Full disclosure: https://t.co/c2JcMjJfRE

Wilson

@WilsonSignals

Last Seen Users on Sotwe

Trends for you

Most Popular Users