Compute & Chips: New Engines for Computing Download as image Share on X
Semiconductors, computing hardware, quantum systems and the software that supports them. A Cosign selection of startups, scaleups and selected public or acquired companies, listed alphabetically.
Quantum computing with cat qubits
Helium turns cat-qubit research into a complete machine, moving a different approach to quantum errors closer to practical use.
Cosign Quantum algorithm development and compilation software
BMW and Citi use its tools to build quantum programs without hand-designing every circuit.
Cosign Frontier inference clusters, co-designed from chips to software.
Jane Street received its first rack, putting an alternative AI chip architecture against real computing demand.
Cosign Laser-based computing for optimization and engineering simulation.
Boeing-backed simulation work tests whether a different kind of computing can make hard engineering searches more practical.
Cosign Low-power processors and software for edge AI.
With 300 million devices shipped, its next push brings AI to hardware that cannot afford a big power bill.
Cosign AI infrastructure built from energy up, connecting power, data centers, and cloud.
Nearly five gigawatts are contracted. The hard part now is turning land, power and hardware into working AI capacity before demand outruns the grid.
Cosign Inference hardware and software designed for practical data-center power limits.
OpenAI’s 120-billion-parameter model has run on its chip. The bigger test is serving real AI demand without letting power consumption dictate the economics.
Cosign Full-stack enterprise networks across internet, wired, Wi-Fi, cellular, hardware, software, and support.
Lyft and Reddit run on its networks, backing a bet that companies should buy working connectivity rather than assemble it themselves.
Cosign Optical connections that let AI clusters scale beyond the rack.
Working with chip partners to move AI data with light, tackling the traffic jams between increasingly powerful processors.
Cosign Memory-centric compute for AI inference.
Corsair entered production in June, aiming to make coding and voice agents respond faster without relying on GPUs alone.
Cosign Software infrastructure for portable quantum applications.
Early Beryllium users test a less hardware-bound way to write quantum software, making programs easier to carry between machines.
Cosign AI infrastructure for training, inference, and agents, scaling to thousands of GPUs.
Reported fivefold growth reflects a bigger shift: AI teams want to ship models without becoming experts in running GPU fleets.
Cosign Flexible circuits combining power, data, and sensing.
Its flexible wiring is already in production EVs, changing how power and signals move through increasingly complex vehicles.
Cosign Quantum computing with electrons on helium and CMOS chips
An autonomous quantum-lab demonstration tests whether AI can accelerate the experiments needed to build better computers.
Cosign Superconducting quantum computers, on premises and through the cloud.
Seventeen quantum computers delivered means more researchers can run experiments on machines they actually have access to.
Cosign Quantum networking hardware for connecting processors at scale.
A bigger lab backs its networking hardware, pursuing quantum computers that grow by connecting processors rather than enlarging one machine.
Cosign Semiconductor skills and engineering services
Verification work for Trilinear tests chip designs before fabrication, when finding an error is far less costly.
Cosign Software for designing and deploying quantum error correction.
Its real-time testbed tackles quantum errors as they happen, a prerequisite for running longer, useful calculations.
Cosign Rethinking the limits of AI, Lightmatter merges photonics and computing to build a future where speed, efficiency, and intelligence converge
Customers have its optical hardware, aimed at removing the data bottlenecks that keep powerful AI chips waiting.
Cosign Wireless power architectures and reference designs for connector-free devices.
WHOOP and Logitech already use its technology. The broader bet is devices that can shed charging connectors without sacrificing usefulness.
Cosign Show 13 more
Cosign
