LDA Technologies launches high-density 25G Layer 1 fabric for trading

LDA Maze delivers 160 links at sub-4ns latency, targeting capital markets, HPC and AI workloads in a single chassis fabric.

By Disrupts Editorial Team Mon, 24 August 2026

A bright modern office space features a curved white desk with blue LED strip lighting, supporting three monitors on black articulated arms displaying dark-themed data visualizations, with large windows and blurred glass partition walls in

LDA Technologies, the Mississauga-based ultra-low latency networking specialist, has launched LDA Maze, a non-blocking, full-mesh Layer 1 fabric promising 160 x 25G links at port-to-port latency of 3.6 to 4.1 nanoseconds. The company says the link density is the highest available at 25G speeds, and the product is positioned to underpin trading architectures, high-performance computing clusters and AI inference workloads within a single rack.

Layer 1 fabrics sit beneath the software stack entirely, managing raw physical-layer signal routing before any protocol processing occurs. That positioning gives them a latency floor that no higher-layer switch or router can match, which is why electronic trading venues, market makers and quantitative hedge funds treat sub-5ns interconnect as a hard infrastructure requirement rather than a preference. LDA says LDA Maze addresses a specific gap: previous high-density offerings at 25G either sacrificed configurability or required firms to over-provision rack space and cabling to compensate.

Capital markets compute meets AI infrastructure

The product's dual billing, capital markets and AI/HPC, reflects a structural shift in how trading infrastructure vendors are positioning themselves. Demand for ultra-low latency compute is no longer confined to the trading floor. GPU clusters running inference workloads share the same physical rack constraints (power budget, cabling density, heat dissipation) that high-frequency trading firms have wrestled with for two decades. By aggregating 160 x 10/25G links to 40 x 100G, LDA Maze can serve both use cases on the same hardware, a convergence that reduces the total number of specialised chassis a data centre operator must procure and manage.

Vahan Sardaryan, CEO and Co-founder of LDA, framed it directly: "One fabric can power devices for finance, HPC and AI workflows."

The physical footprint argument carries weight beyond marketing. Colocation costs in the key financial data centre hubs, Equinix NY4/NY5 in Mahwah, the LD4 complex in Slough, TY3 in Tokyo, are denominated by rack unit and power draw, not by port count. A single LDA Maze unit replacing multiple patch-panel tiers and discrete switches can lower monthly colocation spend materially, particularly for mid-tier trading firms that lack the negotiating leverage of Tier 1 banks.

Convergence implications for cross-sector infrastructure spend

The broader significance for Disrupts readers is what this product class signals about infrastructure investment patterns at the intersection of fintech and AI compute. Specialist low-latency vendors historically served a narrow capital markets customer base. The expansion of that addressable market into AI and HPC is being driven not by the vendors themselves but by the physical convergence of workloads inside enterprise and colocation data centres.

Sovereign wealth and infrastructure funds that have been deploying capital into hyperscale data centre capacity are beginning to encounter the same rack-density economics that trading firms resolved years ago. LDA Maze's dual positioning suggests that the tooling developed for nanosecond-class finance applications is now migrating upward into the broader compute stack, a pattern consistent with how FPGA technology moved from specialised trading accelerators into more general ML inference pipelines over the previous decade.

LDA says LDA Maze is currently integrated into its Neo X platform, with availability across two additional platforms expected in the coming weeks. The company has not disclosed pricing. It is headquartered in Mississauga and describes itself as independent, without naming external investors or funding rounds.

For capital allocators watching the AI infrastructure build-out, the LDA Maze launch is a small but instructive data point: the latency and density demands of AI workloads are pulling specialised financial-infrastructure technology into the mainstream compute stack faster than most data centre architects had modelled.