AttoTude is a Menlo Park–based technology company building terahertz (THz) “radio over wire” electronic interconnects aimed at solving bandwidth, power and reliability limits in hyperscale data centers and AI infrastructure[1][4].
High‑Level Overview
AttoTude develops THz-frequency wired transceivers and interconnect technology intended to replace or augment conventional electrical and optical links inside and between racks to support extreme parallelism for AI and Exascale computing[4]. The company says its approach uses standard ASIC manufacturing processes to achieve terahertz electronic communications over wiring lengths up to ~40 meters, targeting hyperscalers and AI infrastructure providers as customers[1][4]. AttoTude has raised multiple venture rounds (seed, Series A and a $50M Series B led by Mayfield in 2024), bringing total financing to roughly $91M to scale manufacturing, expand customer engagements and prepare for volume production[1][2].
Origin Story
AttoTude was co‑founded by industry veterans including Dave Welch (co‑founder of Infinera and former CTO of SDL) and Joy Laskar (former Georgia Tech professor and serial entrepreneur), with a founding team and board composed of seasoned networking and semiconductor executives and investors[1][2]. The company emerged to address a clear infrastructure pain point: AI workloads are driving unprecedented GPU counts and interconnect demand inside data centers, and AttoTude’s founders pursued a terahertz electronic approach that could be manufactured in standard ASIC flows and deployed at scale[1][4]. Early traction included design wins and pilot engagements with leading hyperscale and AI infrastructure companies as AttoTude moved from research prototypes toward productization and manufacturing scale[1].
Core Differentiators
- Terahertz radio‑over‑wire approach: Operates at THz frequencies over copper/other wiring to deliver higher bandwidth per pair over datacenter distances (claims up to ~40m), differentiating from traditional electrical signaling and some optical link approaches[1][4].
- ASIC‑friendly manufacturing: Designed to be compatible with standard ASIC processes, which is positioned to lower cost and accelerate volume production versus exotic packaging or custom III‑V RF hardware[1].
- Targeted for AI/hyperscale: Engineering focus and go‑to‑market aimed specifically at hyperscalers and AI infrastructure vendors facing exponential interconnect demand[1][4].
- Experienced founding and investor base: Leadership with prior exits and deep optical/networking expertise and backing from Sutter Hill, Canaan, Wing VC, Mayfield and others[1][2].
Role in the Broader Tech Landscape
AttoTude is riding two converging trends: explosive growth in AI compute (projected large increases in high‑end GPU deployment) and the resulting need for greater intra‑system and intra‑rack bandwidth, lower link power, and higher reliability[4]. The timing aligns with hyperscalers’ shift toward disaggregated, highly parallel architectures where wiring reach and cost matter; a viable THz wired solution could change system designs by enabling higher‑density interconnect without the full cost/complexity of optical transceivers[4]. If AttoTude’s claims on performance, manufacturability and reliability hold at scale, the company could influence switch, server and rack‑level interconnect architectures and create competitive pressure on short‑reach optics and SerDes roadmap choices.
Quick Take & Future Outlook
What’s next: move from customer pilots to volume production and broader deployments with hyperscalers and AI infrastructure partners, supported by scaled manufacturing funded by the 2024 Series B[1]. Key trends that will shape success include continued AI compute growth, data‑center economics favoring lower total cost of ownership for interconnects, and how short‑reach optics vendors respond. Risks and open questions include demonstration of long‑term reliability and error rates at THz over practical wiring, integration complexity with existing switching/serdes ecosystems, and whether manufacturing yields and costs meet hyperscaler requirements. If AttoTude converts pilots into production contracts and demonstrates cost/power advantages at scale, it could become a influential enabler of next‑generation data‑center topologies and a significant supplier to the AI infrastructure stack, validating the company’s founding thesis[1][4].
