←
Lightspeed10,072 posts
Lightspeed avatar

Lightspeed

@lightspeedvp

Serving bold builders of the future. Not an offer to buy/sell securities or investment advisory services. Disclosures: lsvp.com/legal/

◷ Joined July 2009▧ 1,573 Photos and videos
Lightspeed avatar
Lightspeed@lightspeedvp·15 Dec 2025

Today we're announcing over $9B in capital for a new family of funds — our largest raise in 25 years. AI is rewriting the rules. We're backing the founders building what comes next.

Post attachment
Justin M. Overdorff avatar
⟳ Lightspeed retweeted
Justin M. Overdorff@jmover·14h

Kat and I sat down with @andrawesbahou to unpack how compute went from a line item to the underlying collateral that supports a multi-billion-dollar capital market. Historically, much of the value in commodity markets and around capital markets has accrued to the financiers, speculators, asset managers, hedge funds, etc. Andy and @SenorScience believe that for the first time in many asset classes, a new entrant is here: AI infra companies that can take well-managed risks. Thank you to Andy for joining us! @katkzhang @lightspeedvp

Compute is becoming a financial asset. @ComputeDesk co-founder and CEO @andrawesbahou joins Lightspeed partners @jmover and @katkzhang on The Investment Memo to discuss how that market is taking shape. Andy argues that reliable delivery of compute brings buyers and sellers into the market. Their transactions help establish the price benchmark others can use to trade and manage risk. Tune in to the full conversation below. 0:00 How Compute Became A Financial Asset 3:08 Historical Parallels to Oil 5:57 What Compute Desk Builds 7:37 Why Capital Markets Have Not Coalesced Around Compute 9:07 Who Could Benefit From The Financialization of Compute? 11:44 How Compute Volume Can Grow 14:46 The Three Products: Compute Signal, Trader And Clear 19:25 Why The Reference Rate Is "The Ultimate Prize" 20:49 Building A Team With Dual DNA 27:53 Why Physical Settlement Matters 34:18 The Paper Versus Physical Market 37:54 The Paradox Of Market Efficiency 39:29 The Fundraising Process And The Ten Year Vision

43:06
Kat Zhang avatar
⟳ Lightspeed retweeted
Kat Zhang@katkzhang·15h

Before we met @andrawesbahou, our team at @Lightspeedvp spent the past 12 months studying the emerging compute market. What we found were early signs of compute being traded like a commodity, but the infrastructure to support those trades still missing. @jmover and I sat down with Andy to unpack what we’re seeing on the financialization of compute. We get into why owning the reference rate is so valuable, why physical settlement is so complex, and how Compute Desk’s approach to building the clearing platform for compute solves for both.

Compute is becoming a financial asset. @ComputeDesk co-founder and CEO @andrawesbahou joins Lightspeed partners @jmover and @katkzhang on The Investment Memo to discuss how that market is taking shape. Andy argues that reliable delivery of compute brings buyers and sellers into the market. Their transactions help establish the price benchmark others can use to trade and manage risk. Tune in to the full conversation below. 0:00 How Compute Became A Financial Asset 3:08 Historical Parallels to Oil 5:57 What Compute Desk Builds 7:37 Why Capital Markets Have Not Coalesced Around Compute 9:07 Who Could Benefit From The Financialization of Compute? 11:44 How Compute Volume Can Grow 14:46 The Three Products: Compute Signal, Trader And Clear 19:25 Why The Reference Rate Is "The Ultimate Prize" 20:49 Building A Team With Dual DNA 27:53 Why Physical Settlement Matters 34:18 The Paper Versus Physical Market 37:54 The Paradox Of Market Efficiency 39:29 The Fundraising Process And The Ten Year Vision

43:06
Lightspeed avatar
Lightspeed@lightspeedvp·16h

Compute is becoming a financial asset. @ComputeDesk co-founder and CEO @andrawesbahou joins Lightspeed partners @jmover and @katkzhang on The Investment Memo to discuss how that market is taking shape. Andy argues that reliable delivery of compute brings buyers and sellers into the market. Their transactions help establish the price benchmark others can use to trade and manage risk. Tune in to the full conversation below. 0:00 How Compute Became A Financial Asset 3:08 Historical Parallels to Oil 5:57 What Compute Desk Builds 7:37 Why Capital Markets Have Not Coalesced Around Compute 9:07 Who Could Benefit From The Financialization of Compute? 11:44 How Compute Volume Can Grow 14:46 The Three Products: Compute Signal, Trader And Clear 19:25 Why The Reference Rate Is "The Ultimate Prize" 20:49 Building A Team With Dual DNA 27:53 Why Physical Settlement Matters 34:18 The Paper Versus Physical Market 37:54 The Paradox Of Market Efficiency 39:29 The Fundraising Process And The Ten Year Vision

43:06
Justin M. Overdorff avatar
⟳ Lightspeed retweeted
Justin M. Overdorff@jmover·16h

Excited to lead @ComputeDesk's $45M raise! Long-term customer contracts can help providers finance GPU infrastructure, but tie buyers to years of capacity they may not need. More ways to hedge future rental prices could help providers offer shorter commitments. Congrats @andrawesbahou, @SenorScience and the team! @katkzhang @Lightspeedvp

Post attachment

Announcing @ComputeDesk raised $45M, led by @lightspeedvp, to solve compute risk. I left Meta last year to build a team of quants and GPU infrastructure engineers to build the clearing infrastructure for compute.

0:56
Lightspeed avatar
Lightspeed@lightspeedvp·17h

Compute is moving beyond IT budgets and into capital markets. We’re leading @ComputeDesk’s $45M Seed round as the team builds infrastructure to support that shift. Co-founders Andrawes (Andy) Bahou (@andrawesbahou) and David López Mateos (@SenorScience) bring experience in the infrastructure and financial markets behind this opportunity. Andy led GPU build-outs at Meta and worked in derivatives pricing; David’s background spans commodities and pricing algorithms. That experience matters when compute commitments can stretch years into the future. Buyers and sellers need ways to price those commitments, manage risk and verify that contracted capacity is delivered. Compute Desk is building across the whole stack, from pricing benchmarks and trade execution to physical settlement. Congratulations to Andy, David and the entire team! CC: @jmover, @katkzhang

Post attachment

Announcing @ComputeDesk raised $45M, led by @lightspeedvp, to solve compute risk. I left Meta last year to build a team of quants and GPU infrastructure engineers to build the clearing infrastructure for compute.

0:56
Lightspeed avatar
Lightspeed@lightspeedvp·Nov 16

“Text is at best almost like a highly lossy compression” of human communication. Tone, facial expressions, personality, and the subtle back-and-forth of a conversation can all be lost when we reduce interaction to text. @fangchangma, co-founder and CEO of @nuance_ai, on why the future of AI interaction is audiovisual. From the Investment Memo with @whoisnnamdi

1:53

AI understands language, but can it master the art of conversation? Some AI avatars today still operate like walkie-talkies. You speak, they process, then they respond. They can miss the subtle cues that make conversations feel human. @nuance_ai is rethinking that problem from the ground up. Co-founders @Fangchangma, @Kyzyx, and @KarrenYangz2 joined @whoisnnamdi on The Investment Memo to discuss their full-duplex foundation model, the science of emotional intelligence, and why the future of AI interaction extends beyond text. “LLMs are trained on the task of next-token prediction… We’re doing almost a parallel task where we’re focusing on next-behavior prediction,” Fangchang explains. We recently led their $50M Series A because we believe the future of AI interaction hinges on building systems that can better understand human communication. Tune into the full episode below. 0:00 Why LLMs Fake Conversation And Introducing Nuance Labs 2:06 Why Audiovisual And The Limits Of Text 5:32 Cascaded Systems Latency And Broken Causality 10:04 The Full-Duplex Engine 14:24 Feeling Human And The Data Problem 18:04 Founder Origins, Team, Hiring, And Culture 29:40 Seattle Talent 31:44 Real Time IQ And EQ 38:09 Lessons From ElevenLabs Use Cases And Competition 45:10 Fundraising What The Capital Unlocks And The Vision

52:48
Lightspeed avatar
Lightspeed@lightspeedvp·Nov 16

“Either the agent is talking or the user is talking at a time. Information is only ever traveling one way.” That's how most turn-taking AI systems work today. “Active listening is such a core part of real human interaction, and it's just completely impossible with turn-taking systems.” The team at @nuance_ai explains how full-duplex models enable AI to listen, react, and respond simultaneously, much like humans do in real conversations. From the Investment Memo with @whoisnnamdi

2:08

AI understands language, but can it master the art of conversation? Some AI avatars today still operate like walkie-talkies. You speak, they process, then they respond. They can miss the subtle cues that make conversations feel human. @nuance_ai is rethinking that problem from the ground up. Co-founders @Fangchangma, @Kyzyx, and @KarrenYangz2 joined @whoisnnamdi on The Investment Memo to discuss their full-duplex foundation model, the science of emotional intelligence, and why the future of AI interaction extends beyond text. “LLMs are trained on the task of next-token prediction… We’re doing almost a parallel task where we’re focusing on next-behavior prediction,” Fangchang explains. We recently led their $50M Series A because we believe the future of AI interaction hinges on building systems that can better understand human communication. Tune into the full episode below. 0:00 Why LLMs Fake Conversation And Introducing Nuance Labs 2:06 Why Audiovisual And The Limits Of Text 5:32 Cascaded Systems Latency And Broken Causality 10:04 The Full-Duplex Engine 14:24 Feeling Human And The Data Problem 18:04 Founder Origins, Team, Hiring, And Culture 29:40 Seattle Talent 31:44 Real Time IQ And EQ 38:09 Lessons From ElevenLabs Use Cases And Competition 45:10 Fundraising What The Capital Unlocks And The Vision

52:48
Lightspeed avatar
Lightspeed@lightspeedvp·Nov 14

“The amount you can juice out of your finite data center capacity goes down the faster you try to run it.” Fractile founder and CEO Walter Goodwin on the cost of faster AI: “You have to end up running your chips at kind of like really low utilization. Your joules per token shoots up.” On Lightwork, he shares what the team has spent years developing in a “secret skunkworks”: packaging DRAM directly on top of the logic doing the math, giving models more memory bandwidth to run faster. @goodwin_ml

1:31

BREAKING: FRACTILE RAISES A $720M SERIES C CO-LED BY LIGHTSPEED I got an inside look at the technical breakthroughs @fractile_ai has been quietly building for years, with Founder and CEO @goodwin_ml, alongside @guruchahal and @jonahcader 🧠 Why memory, not compute, is AI’s real bottleneck - and why today’s chips force a choice between big but slow (HBM) or fast but small (SRAM) 🧱 Stacking DRAM directly on logic gives ~20x the bandwidth per chip of HBM and 10x+ the capacity of SRAM 🔁 Why they chose not to bake the transformer into silicon when models change faster than a chip’s 3-5 year life in the data center 🌍 How a team spanning the UK, US and Taiwan pulled off a “secret skunkworks” in advanced packaging 🧑‍💻 Hiring across silicon, systems and software in London, Bristol and Palo Alto (My favorite part was hearing how the team flew to London, drove three hours to Bristol for one day with the Fractile team, and even got to catch an eclipse! 🌘) FULL SEGMENT BELOW @lightspeedvp

23:37
Lightspeed avatar
Lightspeed@lightspeedvp·Nov 14

Could AI scribes become an early-warning system for outbreaks? Dr. Shiv Rao says Abridge may have spotted cyclosporiasis before public-health systems did: "I'm convinced that we knew about that outbreak before, I dare say, the CDC knew about it." "We didn't know because doctors were prescribing it or putting that ICD diagnosis into the EMR. We knew because patients were describing symptoms, and this was spiking." "So we reached out to health systems and said: is there some public good, non-commercial public good, that we could build together with this de-identified data set?" "It's not quite PCR-ing sewage lines, but it's upstream of the data that ends up in claims and powers a lot of population health dashboards today." @ShivdevRao

2:31

Abridge CEO @ShivdevRao believes healthcare still hasn’t grasped what an AI-native company looks like. With ambient clinical documentation as the wedge, @AbridgeHQ is now expanding to build for the processes that occur before and after the visit. Their goal is to craft a healthcare tool that can handle everything from risk adjustment and care gap closure to note-taking. The idea is to provide whatever the clinician or patient needs, with AI assistance. Shiv explains their mission to Lightspeed Partners Dr. @BrentonFargnoli and Dr. @GalymImanbayev. When he thinks about approaching problems in healthcare, Shiv zooms in on the end user’s perspective. He asks what an enterprise-grade assistant for a doctor would look like, and how Abridge could build one: “Some of these outcomes could be financial, but aspirationally, these outcomes should be clinical.” Watch the full interview below. 00:00 The Clinic Prep Vision: Abridge Working Behind the Scenes 01:43 From History Major and Skateboarder to Physician 07:12 Cardiology and the First Healthcare Software Project 11:57 What Abridge Is Today: Building an AI-Native Health System 15:21 Clinical Jobs the Assistant Can Do: The PET Scan Example 20:09 Pre-Charting Reimagined: Gaps, Prior Auths, and Clinical Trials 24:15 Computer Use Demo: From Generative AI to Actions 28:17 The North Star: A Cardiologist Who Retired Too Early 30:02 Clinician Scientists: Building for Doctors, by Doctors 33:03 Open Weights and Staying Adaptable 39:56 Scaling AI Across the VA 43:27 Life Sciences and Public Health: Spotting an Outbreak Early 46:06 Rearchitecting the Provider and Payer Relationship 52:44 Rapid Fire: Overrated, Underrated, and Investor Advice

56:17
Ravi Mhatre avatar
⟳ Lightspeed retweeted
Ravi Mhatre@ravi_lsvp·Nov 13

Faster inference can expand what’s practical to build with AI. For reasoning models and agents, that could mean working through more steps of a difficult problem within the same amount of time. Proud that @lightspeedvp is co-leading @fractile_ai $720M Series C. Congratulations Walter and the entire team! CC: @guruchahal, @jonahcader, @jshen99

We have raised a $720M Series C, led by @ThriveCapital, @lightspeedvp, @Redpoint and @foundersfund, to allow us to run faster in shipping the world's fastest chips for AI. We've seen over the past year that many of the hardest problems that humanity faces - in science, medicine, physics - may be solved in our lifetimes by AI. One hallmark of AI applied to the toughest problems is that extreme scaling of inference compute becomes vital. In tackling a Millenium Prize problem, for instance, @OpenAI ran 10,000 parallel agents for 80 hours. Costing over $10M, this level of capability is far from democratised today. When we study agentic scaling, there are two paths: parallel, and sequential. In just the same way that most hard problems solved by humans don't parallelise efficiently, we see the same with agents: fewer agents, thinking for longer, scales better in every way except for the time that you wait for a response. Building radically faster inference hardware, which doesn't force a trade-off on throughput, model sizes or context lengths, is a key step in bringing down the cost and time we incur in solving our hardest problems. We're thrilled to be backed to bring this to the world!

Lightspeed avatar
Lightspeed@lightspeedvp·Nov 13

What if investing in U.S. housing didn’t require buying a property? “The ability to actually make that purchase and just get into the asset class, ignoring the fact that it’s a single asset in a single market, is increasingly hard.” @JReedRosenthal explains how @Index_inds is building technology to support an investable housing index. From The Investment Memo with @jmover

1:38

One of the largest asset classes in the US doesn’t have a liquid index. @Index_inds’ tech is powering a new investable index product for the US housing market. Co-founder and CEO @JReedRosenthal joined Lightspeed Partner @jmover on The Investment Memo to expand on what they’re building. Index Industries’ first offering will be the U.S. Home Price Index (USHP Index), a proprietary benchmark for U.S. home prices. Calculated daily, it integrates data from the major housing indices into a single national measure, and will serve as a reference for the USHP token, which is issued by USHP Limited. USHP is a first-of-its-kind liquid index product backed by housing-related assets. Josh explained why he feels USHP is so necessary in today’s market: Advanced blockchain capabilities, a more open regulatory environment, and a generation that has increasingly felt they couldn’t afford owning a home. His goal is to broaden access to the housing market as an asset class. “Every home is a unique snowflake. You need deep expertise in specific markets, specific home types, and so on and so forth,” Josh says. 00:00 Meet Josh Rosenthal, CEO of Index Industries 01:45 From Investor to Co-Founder and CEO 03:29 The Ticker for the US Housing Market 05:19 On Traditional ETF Infrastructures 06:44 Blockchains as a Sandbox for Financial Innovation 10:17 The $300T Real Estate Asset Class 12:36 Tokenization Meets Real Estate 18:58 Building the Team 19:54 Five-Year Vision for US Housing

20:50