Brian Schimpf

When Nvidia reports earnings, it’s not just one company on trial - it’s a pulse check for the entire AI economy. With hype fatigue setting in and budgets under scrutiny, this quarter felt like a referendum: Is AI demand peaking, pausing… or powering ahead? Nvidia’s answer: still accelerating. Beat on revenue. Beat on earnings. Guided ahead. And the stock dropped. Why? Not because the results were weak, but because they weren’t superlative. Data center revenue was a hair shy: $41.1B vs. ~$41.3B. Market consensus for next quarter was $53B, but the whisper number was $55B. Nvidia guided to $54B. If you're priced for perfection but merely crush expectations - you get a “beat and meh.” But the real story isn’t in the share price. It’s in what Nvidia actually said. Let’s lift the hood. 1. Rack = Computer Jensen Huang put it plainly: Each rack is a computer. With NVLink 72, Nvidia made the rack the new atomic unit of compute, turning GPUs into one coordinated machine. This pushes the moat from chips into system architecture. Once a customer builds around Nvidia’s rack-scale blueprint, switching vendors means rebuilding the factory. Good luck pitching a “drop-in alternative.” 2. Networking = Strategic Leverage Networking was the boring part of AI infra - wires, switches etc. Nvidia turned it into a $7.3B business this quarter (+98% YoY). Spectrum-X is running at a $10B+ run rate; InfiniBand nearly doubled. Huang argued that improved networking boosts utilization enough to “make networking effectively free.” It’s not a bundle item anymore, it’s a margin driver. (3) Performance-per-Watt = New ARPU AI isn't compute-constrained, it's power-constrained. Grid limits are real. As they put it: “Performance-per-watt drives the revenues of our customers.” The new KPI chain: perf-per-watt → tokens-per-megawatt → revenue. Blackwell (now) and Rubin (next) aren’t cosmetic - they’re built to compound output inside fixed power budgets. Chips now sell on ROI per watt over speed. Better chips won't cut it, competitors need a better energy-to-token pipeline. That’s a full-stack, full-system problem. (4) Software + Numerics > Silicon The leap didn’t come from the chip, it came from how it’s used. NVFP4 (4-bit math) enables up to 7x faster training without loss of accuracy. Combined with NVLink and Nvidia’s stack (CUDA, TensorRT-LLM, Dynamo), the result: 10-50x energy efficiency per token. An operating system level play. (5) Optionality > Guidance While the Street focused on $54B guide, Nvidia left upside on the table: H20 China sales: $2-$5B not in the guide. Sovereign AI: now a $20B+ business (2× YoY). RTX Pro: seeding enterprise AI in 90+ logos across pharma, auto, and more. Nvidia’s revenue streams are growing wider, not just taller. Labeling Nvidia a chip company understates what they’ve become. They’re not competing on silicon, rather on system-level economics: who can turn power into tokens most efficiently. This is an AI systems company.

283

31 Comments

Uber is pushing out Waymo Uber just announced a multibillion-dollar push into robotaxis in partnership with Lucid and Nuro. The era of self-driving at scale is arriving, and Uber wants to be the operating system of urban mobility. What struck me about this story wasn’t just the deal - but the evolution behind it. I had the chance to speak with Lesya Hendrix, a former Technical Program Manager at Nuro, who worked directly on the robotic fleet involved in this transition. Her story offers a rare behind-the-scenes look at what it takes to pivot a robotics company into a scalable AI platform. → From robots to revenue models “In 2023, Nuro pivoted from being a hardware company to a software company,” Lesya shared. “We had built a fleet of 30 delivery robots across #MountainView, #PaloAlto, and #Houston. But ultimately, the economics didn’t support scale. #Hardware was costly, and the unit economics of last-mile delivery didn’t justify the infrastructure.” Instead, Nuro focused on what it did best: its autonomy stack. They shifted from building robots to licensing their self-driving software - partnering with Lucid who had the vehicles, and Uber who could go to market. → A pivot born of pragmatism This isn’t uncommon in robotics. Many #hardware-heavy startups face the harsh math of scaling physical infrastructure. But what Nuro did well - according to Lesya - was preserve the integrity of the team and the culture during the shift. “In all my time working in #SiliconValley, Nuro had the best culture. The CTO became my mentor. And even during the pivot they kept open communication and offered healthy terms of transition” → Why Uber, not Waymo? “Because Uber is better positioned to tap into global markets,” Lesya told me. “And whenever they won’t have have enough autonomous rides in your area - they can offer you the good old human driven alternative and keep customers. This is a stronger business proposition and I am so glad that Nuro’s technology is a part of it” Uber’s strength lies in scale and logistics. What they needed was a cutting-edge autonomy stack that could plug into their massive rider base. Nuro and Lucid complete the trifecta: #AI brains, electric muscle, and consumer reach. → What this means for investors Nuro remains private. But public market participants are already reacting - Uber’s stock jumped on the news, and Lucid may follow. For those looking to get exposure to autonomy, Uber and Alphabet (which owns Waymo) remain key players. But the real story is in how traditional business models are being rebuilt around software, data, and partnerships. Lesya shared - "Today, I’m building Spacer Robotics - a company focused on autonomous, non-humanoid robots for construction on Earth and in Space. Because if robotaxis are transforming how we move within the cities, then construction is next. And someone has to build the builders."

81

54 Comments

SF was never dead. 3 years ago, people said the city was over. There were stories of empty towers and headlines about doom. Then AI hit escape velocity, and you could feel the energy reignite. This past year, AI companies have taken millions of square feet across the city (1). The marquee names keep expanding. Anthropic added ~100,000 square feet across from its SoMa HQ (2).  Rents are on the rise again. Hacker houses are back (3)(4). SF had what others didn’t.  What changed San Francisco stacked the right inputs for this cycle. Dense talent, fast capital, and tight feedback loops. In AI, those loops matter. Agents get good by shipping, breaking, and retraining. Things compound when researchers, product leads, and investors are within a stone’s throw. The numbers tell the same story. U.S. startup funding in the first half of 2025 jumped ~76% year over year, on pace for the second-best year ever, with AI responsible for a historic share of the dollars (5). By mid-year, GenAI funding had already surpassed all of 2024 (6). Globally, AI startups took ~53% of all VC dollars in H1 2025 (7), and in the U.S., AI accounted for ~64% of deal value (5). The Bay Area drew nearly $70B of the $134.6B in global AI funding in 2024, and it’s still the gravitational center in 2025 (8). What we’re seeing To understand why SF is back, look at where investors are leaning in. Recent a16z AI investments span agentic CX, coding, creative tools, and multimodality. Decagon, building autonomous customer-support agents, raised a $131M Series C co-led by a16z (9). Cursor, the AI code editor and “vibe coding” tool many engineers live in, raised a $900M round (10). Krea, the creative AI interface for images/video, just closed an $83M round (12). Risk capital The Bay Area’s edge has always been access to the “right money” at the “right stage.” Later-stage capital is returning, and big pools are forming to fund AI’s compute-heavy scale-up phase (16). The talent wars We’ve all read the headlines. Meta has reportedly offered top researchers pay packages up to ~$300M over four years, with first-year comp figures reportedly above $100M for a small number of leadership roles (18). Whatever the exact figures, the platforms are competing for a very small pool of elite talent, and much of that is happening in the Bay Area. At the same time, SignalFire’s 2025 report shows Anthropic near 80% two-year retention, with OpenAI ~67% and Meta ~64% (19). Labs with strong cultures tend to retain their people. They also tend to cluster around SF.  Cities have product-market fit, too. In 2025, San Francisco is a product that AI builders want. If you want to experience the energy for yourself, come build with us during SF Tech Week (Oct 6–12)! This year, the SF Tech Week calendar blew past 1,000 events, which tells you something about where builders are spending their time. https://lnkd.in/gdnjrpyT 

193

18 Comments

Lightspeed is thrilled to have invested in Periodic Labs’s $300M seed round. This company is led by an incredible team of ex-OpenAI and DeepMind researchers, who are turning AI into an engine for scientific discovery. When I first met Co-Founders William (Liam) Fedus and Ekin Dogus Cubuk, I knew they were building something that could change many industries. Most AI companies focus on software. Periodic Labs is building the AI and the actual labs. Their system doesn't just theorize, it creates hypotheses, designs experiments, and then actually runs them through robotic systems. Every experiment, whether it succeeds or fails, teaches the AI something new. We're talking about accelerating scientific discovery in semiconductors, fusion energy, and sustainable materials, which are the building blocks for everything from better electronics to space travel. It's rare to find founders tackling problems this ambitious with this level of technical depth. Congrats to the entire team at Periodic Labs. 

593

20 Comments

Hardware enabled software will be the most defensible companies in the AI ERA. Hardware-enabled SaaS turns sensors and devices into proprietary data engines, then converts that data into sticky, high value software with elite retention. The top companies check the boxes that Jeff Horing at Insight Partners highlighted on his recent interview on Invest Like The Best: Big ROI, Big ASP, Fast Time-to-Value. How it looks in the real world 1. Axon ($56B Market Cap) - Law Enforcement Tech What they sell: cameras and software that help police record what happened and finish paperwork faster. The Result: $2B+ Revenue and 124% Net Revenue Retention. Why it wins: devices create the evidence, Evidence.com and AI draft the reports, and multi-year bundles keep customers for years. 2. Badger Meter ($5.38B Market Cap) - Water Tech What they sell: smart water meters and software that help cities find leaks and bill accurately. The Result: $826.6M revenue in 2024. Why it wins: meters stream usage data, BEACON flags leaks and losses, and long replacement cycles plus data history raise switching costs. 3. Samsara ($22.62B Market Cap) - Safety for the Physical World -What they sell: trackers, dash cams, and software that help fleets and job sites run safely and on time. -The Result: $1.35B ARR growing mid-30s percent year over year. Gross logo retention cited at 98% plus. Net Revenue Retention 115-120%. -Why it wins: devices instrument vehicles and equipment, the Connected Operations Cloud runs daily ops, and multi-product expansion fuels retention. Why now Component costs keep falling, connectivity covers the edge, and AI is turning raw signals into decisions in seconds. This enables counter positioning. Give away or subsidize the device, price software on outcomes, and let the data flywheel deepen the moat. My simple HES playbook 1. Start with a wedge that instruments a high value, high frequency decision. 2. Remove capex friction with free or subsidized hardware that undercuts incumbents dramatically 3. Close the loop so alerts trigger real work orders in the system of record or complete the action that operators had to do historically by leveraging AI (e.g. report writing) 4. Compound moats as every deployed device grows your proprietary dataset that trains your models. 5. Sell outcomes - increased revenue, direct labor savings, insurance savings, loss reduction, etc. I'm going to share a deep dive in these businesses and my thesis in coming weeks. Let me know any other companies you admire that you think I should study!

149

26 Comments

Today on Training Data: Ping Wu CEO of Cresta and special guest Douglas Leone on the thorny last mile in AI applications. It's easy to make an AI demo of a feature like auto summarization, but deploying it to production turns into be a web of real-world complexity. Moats and durable enterprise value are built in that thorny last mile, which in the customer support space means: * Integration with on-premise systems (50% of customer support calls!) * Transfers to other agents and specialists * Calls that spill over past the contact window * Real time streaming and background noise reduction * PII handling etc Fun to hear Ping's perspectives plus Doug the goat

150

12 Comments

Reinforcement Learning allows agents to learn by doing through trial and error, feedback loops, and real interaction. It’s the process that agents use to learn, adapt, and improve. For RL to work at scale, two things are critical: immersive training environments and easy-to-use infrastructure. That’s exactly what we’re seeing emerge today. Platforms like Kaizen and Mechanize Inc. are building realistic digital playgrounds where agents can learn meaningful work like replicating software tasks, debugging code, and filling out forms. Meanwhile, RL-as-a-Service providers like Applied Compute, Veris AI, and Osmosis [Gulp (YC W25)] are making it possible for any company to train agents on custom goals, without needing a research team. I shared more about RL’s moment here: https://lnkd.in/dR2m87UR

52

5 Comments

LLMs are famously bad at understanding XLSX -- until now. Shortcut can now jump inside of attached excel files to understand them better. Here a hedge fund receives a huge model to extract data from to update their internal model. Turns hours of work into minutes.

260

33 Comments

Viam begins where silicon ends and motion begins. Founded in September 2020 by Eliot Horowitz, longtime CTO of MongoDB and Brown-trained computer scientist, Viam delivers on a vision he's chased since ShopWiki: to treat robotics software like modern web-app frameworks. He didn’t just launch a startup; he built an abstraction layer for the physical world, packing data, AI, and hardware into a unified engineering stack. Across that stack stands Simone Kalmakis, VP of Engineering, who stitched modular drivers, edge-cloud orchestration, and #ML inference into a system designed for real saying yes to unreasonable requests. Alongside her is Edouard Servan-Schreiber, VP of Solution Engineering, forging customer integrations across industries that actually work, not just pitch decks. Advisors like Carolyn Everson (Permira, ex-Instacart/Facebook) and Albert Wenger (Union Square Ventures) bring the kind of M&A and scaling fluency that turns engineering into category power. Viam's platform lets companies plug in #devicecontrol, #datacapture, #fleet-wide management, #predictivemaintenance, even #remotedebugging, without months of glue code. Their modular driver model and edge-cloud flexibility mean factory floors, arenas, yachts, QSR kitchens, even sonar fishfinders get full-stack intelligence with swap-in modules and reusable community code. Funding tells the story. Seed through Series C, Viam has raised $117M, including $45M in Series B (led by USV and Battery Ventures, Mar 2024) and $30M in Series C (led again by USV with Battery and Neurone, Mar 2025). That capital fuels real deployments: robotic sanding on Viking Yachts, AI-powered sonar for Kongsberg Discovery, fan-flow optimization and interactive demos at UBS Arena with the New York Islanders. Viam’s traction spans #robotics, #marine, #climatetech, #QSRautomation, #industrial manufacturing, and #sports. They’ve cut time-to-production from months to days. That’s the kind of shift that signals a new category: physical intelligence. Culture is engineered for impact: collaboration, curiosity, accountability, diversity. #Opensource isn’t lip service, it’s alive in the Viam Registry, driven by community-built modules. Hardware, software, ML engineers build shoulder-to-shoulder. Partnerships with WAVS and Billion Oyster Project reinforce the build-for-good DNA. Now hiring across #engineering, #solutionengineering, #productmarketing, #projectcoordination. If you thrive under ambiguity, this is your sandbox, and stakes are high. Viam is writing software in the real world, across #yachts, #arenas, #kitchens, #boats, where every inch matters. Follow Viam's journey, explore roles, book a demo, and let's see how much reality you can tame. #Startups #StartupSpotlight #StartupGrowth #FounderStory #TechEcosystem #StartupEcosystem #Data #DataDriven #AI #Hardware #AIHardware #Cloud #CloudComputing #Technology #Innovation #Hiring #TechHiring If engineering peace of mind is what you crave, Vention is your zen.

29

2 Comments

Flox raised a $25M Series B last week to rebuild the foundations of modern software development. On top of that, they just announced a partnership with NVIDIA and the Nix Foundation to make CUDA-accelerated stacks seamless - critical infrastructure as AI-driven code becomes the norm. Ron Efroni, Michael Brantley and the whole Flox team are leading the NixOS ecosystem forward and bringing Flox to developers everywhere. I’m excited to continue our work with Addition, New Enterprise Associates (NEA), The D. E. Shaw Group and more.

74

4 Comments

Waymo's Tekedra N. Mawakana recently joined the guys on The New York Times Hard Fork podcast and dropped that they're targeting one million rides per week by next year. This milestone isn't just about autonomous vehicles. It's arguably embodied AI's first major breakthrough in unstructured consumer environments. And for investors, that changes everything. Beyond software: Embodied AI manipulates the physical world (cars, robots, drones) not just digital data. McKinsey estimates automation of labor could add $9T to global GDP by 2030. Trust infrastructure: When ChatGPT makes a mistake, you regenerate the answer. When embodied AI makes a mistake, someone could get hurt. Companies that solve for public trust unlock adoption bottlenecks across the category. System effects: Every safe Waymo ride doesn't just strengthen Waymo. It builds societal comfort with AI agents in physical environments, accelerating adoption across logistics, healthcare, and beyond. Capital intensity: Unlike pure software, embodied AI requires infrastructure, manufacturing, and safety systems. This favors early winners who can raise and deploy billions but potentially also latecomers who can sprint fast without heavy cap tables and tech debt. Above all, Waymo's milestone validates the embodied AI investment thesis at scale. The next decade will be defined by AI agents acting autonomously in physical environments. If you're curious to dig in, I unpack this further in my latest Substack essay (link in comments). #AI #Mobility #VentureCapital #EmbodiedAI #SystemsThinking

32

4 Comments

Is your AI startup's first $10M ARR really just a tech echo chamber? Our latest analysis reveals that for Bay Area founders, 60-65% of early revenue comes from fellow tech companies—not the broader U.S. market in states like Ohio or Florida—based on data from CB Insights, PitchBook, and McKinsey. We've unpacked global variations, counterexamples like Veracyte and Kabbage, and actionable strategies to break free—check out the full breakdown in my new post! What’s your take on diversifying beyond the Valley bubble? Share your experiences or tips in the comments! https://lnkd.in/gCKTVyT4 #AIStartups #TechRevenue #Entrepreneurship #VentureCapital

37