<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk" -->

---
title: Robot Videos: Morphing Robots, Dancing Humanoids, More
description: AI inference is now limited by power, driving a wave of optical-computing startups aiming to replace parts of GPU workloads with light-based matrix-vector...
canonical: https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Robot Videos: Morphing Robots, Dancing Humanoids, More | daily.dev
og:description: AI inference is now limited by power, driving a wave of optical-computing startups aiming to replace parts of GPU workloads with light-based matrix-vector...
og:url: https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk
og:image: https://api.daily.dev/og/posts/Dl8vcudbK.png
og:image:alt: Robot Videos: Morphing Robots, Dancing Humanoids, More
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Robot Videos: Morphing Robots, Dancing Humanoids, More

**[IEEE Spectrum](https://daily.dev/sources/ieeespectrum)** · 18 min read · 0 upvotes · 0 comments

## Summary

AI inference is now limited by power, driving a wave of optical-computing startups aiming to replace parts of GPU workloads with light-based matrix-vector multiplication. Companies like Lumai, Neurophos, Q.ANT, Lightelligence, LightSolver and Optalysys are each targeting different pieces of the AI pipeline. Neurophos targets 50 times Nvidia Blackwell B200's speed and 35 times its energy efficiency, with a roadmap to 250 times the speed and roughly 90 times the energy efficiency in later generations. Lumai says its next-generation Iris Aura server will do prefill using roughly one-tenth of the energy a GPU would use. Researchers caution that encoding costs, Amdahl's law and nonlinear operations limit real-world gains, and optical-computing companies are aiming for a 10 to 80 fold improvement in energy efficiency over today's GPUs.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://spectrum.ieee.org/video-friday-morphing-robots-quadrotor>

## Questions this post answers

### How much faster does Neurophos claim its optical chip will be compared to Nvidia's Blackwell B200 GPU?

Neurophos is targeting 50 times the speed and 35 times the energy efficiency of Nvidia's Blackwell B200 GPU, with a roadmap to reach 250 times the speed and roughly 90 times the energy efficiency in subsequent generations. The company uses metamaterial-based modulators packed into a 1,000 by 1,000 array on a 5-millimeter chip, switching over 10,000 times faster than standard liquid-crystal spatial light modulators.

_Engineers evaluating next-gen AI accelerators can track optical computing benchmarks like this on daily.dev._

### What energy efficiency gains are optical computing companies targeting compared to today's GPUs for AI inference?

Optical-computing companies today are aiming for a 10 to 80 fold improvement in energy efficiency over today's GPUs, after accounting for encoding costs and Amdahl's law limitations on the nonlinear operations optics can't accelerate. Lumai's upcoming Iris Aura server specifically targets one-tenth the energy a GPU uses for the prefill phase of inference.

_Teams weighing future inference hardware options can follow efficiency claims like these on daily.dev._

### Why hasn't optical computing replaced GPUs for AI despite its theoretical energy advantages?

Encoding data into light and reading it back out still costs energy, so electronics win when vector length is under roughly 1,000 to 10,000 elements, per Cornell's Peter McMahon. Amdahl's law also limits gains since matrix-vector multiplication is only part of inference workloads, and Yale's Logan Wright notes AI algorithm development tends to favor GPUs rather than optics.

_Developers deciding whether emerging optical hardware is worth tracking can follow this debate on daily.dev._

## Similar posts on daily.dev

- [This Light-Powered AI Chip Is 100x Faster Than a Top Nvidia GPU](https://daily.dev/posts/this-light-powered-ai-chip-is-100x-faster-than-a-top-nvidia-gpu-cmurblonw) · Singularity Hub · 1 upvotes · 0 comments
- [From invisibility cloaks to AI chips: Neurophos raises $110M to build tiny optical processors for inferencing](https://daily.dev/posts/from-invisibility-cloaks-to-ai-chips-neurophos-raises-110m-to-build-tiny-optical-processors-for-in-7oo0cb60y) · TechCrunch · 0 upvotes · 0 comments
- [UK’s ARIA will test Lumai’s light-based AI servers in a data centre](https://daily.dev/posts/uk-s-aria-will-test-lumai-s-light-based-ai-servers-in-a-data-centre-evaz2ameo) · The Next Web · 0 upvotes · 0 comments

---

Tags: [#hardware](https://daily.dev/tags/hardware), [#gpu](https://daily.dev/tags/gpu), [#ai-inference](https://daily.dev/tags/ai-inference)

[View this post on daily.dev](https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Robot Videos: Morphing Robots, Dancing Humanoids, More","url":"https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk"},"datePublished":"2026-11-20T17:51:39.545Z","dateModified":"2026-11-20T17:52:10.451Z","description":"AI inference is now limited by power, driving a wave of optical-computing startups aiming to replace parts of GPU workloads with light-based matrix-vector...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8f27585253caa5f17a7ba025e44490ba?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8f27585253caa5f17a7ba025e44490ba?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"IEEE Spectrum","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"IEEE Spectrum","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/b123f5e8420e492ab36f6adb30c1793a","url":"https://daily.dev/sources/ieeespectrum"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"hardware,gpu,ai-inference","timeRequired":"PT18M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"IEEE Spectrum","item":"https://daily.dev/sources/ieeespectrum"},{"@type":"ListItem","position":3,"name":"Robot Videos: Morphing Robots, Dancing Humanoids, More"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/robot-videos-morphing-robots-dancing-humanoids-more-dl8vcudbk#faq","mainEntity":[{"@type":"Question","name":"How much faster does Neurophos claim its optical chip will be compared to Nvidia's Blackwell B200 GPU?","acceptedAnswer":{"@type":"Answer","text":"Neurophos is targeting 50 times the speed and 35 times the energy efficiency of Nvidia's Blackwell B200 GPU, with a roadmap to reach 250 times the speed and roughly 90 times the energy efficiency in subsequent generations. The company uses metamaterial-based modulators packed into a 1,000 by 1,000 array on a 5-millimeter chip, switching over 10,000 times faster than standard liquid-crystal spatial light modulators. Engineers evaluating next-gen AI accelerators can track optical computing benchmarks like this on daily.dev."}},{"@type":"Question","name":"What energy efficiency gains are optical computing companies targeting compared to today's GPUs for AI inference?","acceptedAnswer":{"@type":"Answer","text":"Optical-computing companies today are aiming for a 10 to 80 fold improvement in energy efficiency over today's GPUs, after accounting for encoding costs and Amdahl's law limitations on the nonlinear operations optics can't accelerate. Lumai's upcoming Iris Aura server specifically targets one-tenth the energy a GPU uses for the prefill phase of inference. Teams weighing future inference hardware options can follow efficiency claims like these on daily.dev."}},{"@type":"Question","name":"Why hasn't optical computing replaced GPUs for AI despite its theoretical energy advantages?","acceptedAnswer":{"@type":"Answer","text":"Encoding data into light and reading it back out still costs energy, so electronics win when vector length is under roughly 1,000 to 10,000 elements, per Cornell's Peter McMahon. Amdahl's law also limits gains since matrix-vector multiplication is only part of inference workloads, and Yale's Logan Wright notes AI algorithm development tends to favor GPUs rather than optics. Developers deciding whether emerging optical hardware is worth tracking can follow this debate on daily.dev."}}]}
```

