Volantis raised $88 million Series A to address AI memory bottlenecks with optical chips

Posts cited

1. [23] JACK · @jackrider69 · 06:59

   The Binance Wallet campaign with Axis Robotics is officially closed And the numbers show how much activity happened in just a few weeks 2,521 contributors 477,461 task completions 989 tasks covered What I like here is that this was not just another campaign where people click around for points Axis is using browser based simulation to turn human actions into robot training data and build a compounding data engine for Physical AI No expensive robot No robotics lab Just a browser and a task Every completed trajectory can become part of a much bigger data pipeline and accepted trajectories can be verified with onchain provenance on Base The Binance campaign may be over but the bigger experiment is still running Turning collective human activity into better data for the next generation of robots is the interesting part of Axis Robotics to me @axisrobotics https://t.co/xnx7dVAbkC

   Open on X

2. [34] Whale Insider · @whaleinsider · 05:16

   JUST IN: Volantis raises $88M Series A to tackle AI’s memory bottleneck with optics, enabling chips with massive fast & cheap memory for ultra-fast inference on models over 10T parameters.

   Open on X

3. [43] CBS News · @cbsnews · 03:46

   BREAKING: President Trump is likely to pick director of national intelligence Jay Clayton to be his AI czar, sources say.

   Open on X

4. [48] Andrej Karpathy · @karpathy · 02:37

   We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better: Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better: Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better: Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work! In summary: - As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding. - Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

   Open on X

5. [50] Not Jerome Powell · @alifarhat79 · 02:31

   Volantis just raised $88M to solve AI’s memory problem with literal light. 10,000 tokens/sec. 10T+ models. Agents that finish in seconds, not hours. Apparently the next AI scaling law is: turn the RAM into a laser beam.

   Open on X

6. [51] Shay Boloor · @stocksavvyshay · 02:28

   Volantis just raised $88M from investors including $GOOGL Jeff Dean to build optical AI chips that ease the memory bottleneck in inference. Interesting timing right after $MU’s record quarter with Volantis betting light can make agent workloads dramatically faster.

   Open on X

7. [55] Wall St Engine · @wallstengine · 02:04

   $GOOGL JUST PUT ITS AI CHIPS INTO ORBIT FOR THE FIRST TIME Google and Planet Labs $PL have successfully launched Project Suncatcher M1 aboard SpaceX’s Transporter-18 mission. Planet says it has already made contact with the spacecraft and begun commissioning. This is the first-ever in-orbit test of Google’s Tensor Processing Units, or TPUs. The refrigerator-sized prototype carries four Trillium TPUs, Google’s custom AI accelerators. This is not yet an orbital data center. The goal is to see whether the same type of AI hardware Google uses on Earth can survive launch and reliably run machine-learning workloads in low Earth orbit. There are three major engineering problems Google is testing: RADIATION: cosmic radiation can corrupt calculations or damage electronics. Google previously tested Trillium TPUs with proton radiation and found they could withstand more total radiation than expected over a five-year mission. COOLING: there is no air in space, so fans cannot cool the chips. Heat has to be moved through heat pipes into radiators and then emitted away from the spacecraft. The current prototype is expected to run AI workloads in roughly 15-minute bursts before allowing the system to cool. NETWORKING: Google’s long-term plan is not one satellite, but clusters of satellites carrying dozens of TPUs each and connected through high-speed laser links. Google estimates future AI workloads could require tens of terabits per second between satellites. Its ground prototype has already demonstrated 800 Gbps in each direction. Why put AI compute in space at all? In the right low-Earth orbit, solar panels can generate up to 8x more power than on Earth because satellites can receive near-continuous sunlight, reducing one of the biggest constraints facing AI data centers: electricity. The economics: Google estimates that if launch costs eventually fall below roughly $200 per kilogram by the mid-2030s, the cost of operating compute in space could begin approaching the energy cost of comparable terrestrial data centers.

   Open on X