Simon Chan

HUGE news from the Pioneer Fund portfolio: Kapital becomes the FIRST YC unicorn backed since 2022. They just raised $100M at a $1.3B valuation led by Tribe Capital.

114

5 Comments

A Founders Guide to The Long Build™ - Physical AI Thesis Part III This one is for the builders. In this final post, my partner Will Porteous and I along with insights from builders in the trenches and investors who play the long game share our guide for founders building intelligence in the real world. A few core takeaways: • Sim before steel: Runway is precious, model the system behavior • Capital intensity is not the enemy — misaligned capital is • Build complete solutions before you talk platform • Hardware opens the door, software wins the room • Trust is earned through resilience, not hype

10

1 Comment

Sustained ~252–291 GiB/s of HBM memory throughput on H100s under decode load — essentially hitting the hardware roofline. This matters because HBM throughput, not FLOPs, is the real bottleneck in LLM inference. By keeping memory nearly fully saturated, I’ve unlocked far higher efficiency and throughput than standard engines. The result: 0.9–1.36M tokens/sec with ~0.1 ms first-token latency  #AI #LLM #GPURouter #H100 #Inference #CostEfficiency #Innovation #DeepLearning #AIInfrastructure #HighPerformanceComputing NVIDIA AMD OpenAI Google Shilpa Kolhatkar Keith Strier a16z speedrun

5

Trying something new this week - a studio industry roundup with my take on what matters. The studio model is no longer an experiment being run in a handful of places. It is infrastructure being built across geographies, sectors, and institutional types. 5️⃣ stories from the past seven days:  1️⃣ OSS Ventures hit a EUR 40M first close on a EUR 75M follow-on fund -- capital raised specifically to back the 30 companies their studio created since 2019. The "build then fund" lifecycle is maturing. B'More Venture Studio with a similar thesis: studio operations first, then dedicated capital to scale what's working.  2️⃣ Delta40 raised $20M from 54 investors across 13 countries to launch Africa's first integrated venture studio and fund. The LP base is almost entirely DFIs and foundations. Impact capital is discovering the studio model as a deployment vehicle, something I seeing across geographies (conversations with Michael Bob Starr and Damias McDonald this week alone) 3️⃣ Start Holdings launched Start.vc out of NYC and Barcelona after a $200M EV infrastructure exit. Another operator-turned-studio-founder. This is the pattern we see quite often in our work at 9point8 -- experienced operators like Jason Goldsmith who've built and exited deciding the studio model is how they want to build next.  4️⃣ UBC HATCH admitted 10 new deep-tech spinouts for Winter 2026 -- mostly climate tech. University venture builders are quietly becoming one of the fastest-growing segments. Follow Evan Allen and Jaimie Testai to keep track of its rapid evolution.  5️⃣ The INVEST Act is advancing through the Senate. If passed, it expands qualifying VC fund size from $10M to $50M and loosens general solicitation rules for accelerator and university events. For studio-funds -- and for folks in the The Venture Studio Forum community tracking policy -- this could be the most significant capital formation legislation since the JOBS Act. If this roundup format is useful, I'll keep it going. What other noteworthy news hit your desk this week, and what stood out to you?

13

1 Comment

Why was the Fivetran-dbt merger all but inevitable? Fivetran & dbt Labs announced their merger yesterday. The all-stock deal combines two companies into an entity approaching $600 million in ARR. The beauty of the modern data stack was the explosion in choice. As the cloud exploded onto the scene, the legacy data warehouse was replaced by a collection of fast-moving platforms. In that era, specialization won. The pendulum is now swinging back towards consolidation. Why? The answer lies in compute economics & revenue scale asymmetry. The table below shows why. There are three different categories of software within this subset of the ecosystem: 1. Ingestion takes data from software & moves it into a cloud data warehouse. Snowflake acquired Datavolo, which commercializes the open source product Apache NiFi, calling it Openflow. Databricks acquired Arcion for ingestion through change data capture, calling it LakeFlow Connect. Fivetran focuses exclusively on this layer. 2. Transformation means reformatting the data within the cloud data warehouse. Snowflake launched native dbt Projects on Snowflake. Databricks offers Delta Live Tables, native SQL, & Python, plus supports hosted dbt through Databricks Workflows. dbt Core/Cloud is the leading independent transformation tool. 3. Compute revenue is generated when we ask questions of our data. Snowflake remains one of the leaders in structured data analysis with their cloud data warehouse. Databricks’ compute is their own as well. Here’s the asymmetry in one number. Compute represents 72% of the overall market ($7.6B of $10.6B). As a result of their massive operations, Snowflake & Databricks exert significant gravity within the ecosystem. They have expanded beyond the compute market to impose their presence & capture marginal revenue within customers, pressuring the competitive ecosystem. That’s not to say these components are independent. George Fraser analyzed Snowflake workloads in September 2024, finding transformation represents 40-45% of total Snowflake compute, which means even at smaller scales, startups can have significant impact on these behemoth businesses. The Fivetran-dbt merger is an inevitable evolution of a maturing market. Two unicorns must partner to compete against two decacorns. They solve two of the three customer problems. But not yet compute. One could surmise this consolidation signals the end of the modern data stack. I view it differently. The MDS has succeeded beyond our expectations. The stakes are higher now. Broad platforms, fast growth, & AI-native architectures define the next phase. Expect more consolidation. Category revenue estimates based on public disclosures & company filings. Ingestion: Informatica ($1.64B FY2024), Fivetran ($300M est.), Talend ($350M), others ($200M est.). Transformation: dbt Labs ($300M est.), others ($200M est.). Compute: Snowflake ($3.6B FY2025), Databricks ($4.0B ARR Aug 2025).

41

9 Comments

Harpoon Ventures + Black Flag is teaming up with Microsoft for Startups to provide founders with access to up to $250,000 in Azure credits, empowering builders to scale faster, train models more efficiently, and access world-class AI infrastructure. What’s included: ⚡ Up to $250K in Azure credits (usable across Azure OpenAI, Grok 3, Llama 4, and more) 💻 On-demand GPU VMs for model training and fine-tuning 🧠 Access to GitHub Enterprise + Microsoft 365 Business Premium 🤝 Dedicated Microsoft technical + GTM support This partnership is designed to give deep-tech and dual-use founders the tools and resources they need to win. Because when they win, we all win. 🏴‍☠️

108

8 Comments

Founders NEED to be smarter about structuring POCs. A pilot is not “prove the tech.” It’s 30 days to co-create a business case (ROI model) with your buyer. Do this: 1. Charge for the pilot. It filters tire-kickers. Be explicit: pilot fees ≠ production pricing. 2. If pushed on price before value: -Anchor to ROI: “For peers we unlock ≈ $10M; we target 1:10 ROI.” -Give ranges, not a point: “Final pricing will land between $500k–$1M depending on realized value.” 3. Build the ROI model with the buyer (during the pilot): Incremental gains: revenue lift, churn reduction, throughput, win-rates. Cost savings: headcount/time saved, license/tool consolidation. Opportunity cost: what higher-value work replaces the saved time? The Outcome? You qualify faster, negotiate on value (not features), and set up a clean path to outcome-based pricing. Save this for your next pilot. #AI #SaaS #Pricing #Monetization #GTM #VerticalAI #Founders #Sales

302

28 Comments

Google DeepMind has launched Gemini 2.5 Deep Think, now available to Ultra subscribers in the Gemini app. Official announcement and more details: https://lnkd.in/egvkTt_z This is a good opportunity to better understand what do we mean by "thinking models" and why do we need them. 🤓 💡 If you have 15 minutes to spare, I strongly recommend this excellent short talk from Jack Rae (Google DeepMind's thinking tech lead): https://lnkd.in/eKjKxQVG tl;dr; *AI intelligence progress is all about removing bottlenecks* Historically, major advances in AI come from identifying core “bottlenecks” in current models, and then innovating to remove them (e.g., deep learning, attention, transformers). *Bottleneck today: test-time compute* While LLMs like Gemini and ChatGPT are powerful, they are constrained by a fixed amount of computation per user request (“test-time compute”). This limits the depth and quality of reasoning and solution generation for complex queries. *Solution: dynamic, iterative ‘thinking’ stage* Hence the “thinking stage,” where the model can iteratively loop, perform extra reasoning, and dynamically decide how much additional compute to spend based on the problem’s difficulty. This allows the model to give simple answers quickly and use “deep think” mode to spend much more time and compute on hard or valuable tasks. *Implementation: reinforcement learning for thinking* Models are trained to use this thinking stage via reinforcement learning—rewarding successful problem-solving even with vague, high-level feedback. This shapes the model’s ability to reason, self-correct, decompose problems, try alternative approaches, and use tools when needed. This generic method for training "how to think" led to emergent behaviors such as forming hypotheses, testing them, discarding wrong solutions, breaking down complex tasks, and combining different solution strategies. *Ongoing challenges: efficiency and overthinking* Research continues on making the thinking stage more efficient (avoiding overthinking, unnecessary computation) so that models automatically and adaptively choose the right level of effort for a given task, as well as data efficiency - LLMs that think incredibly deeply and data-efficiently, from a small set of input knowledge. Happy thoughts! #AI #Gemini #DeepThink #GoogleDeepMind #ArtificialIntelligence Antonio Gulli Vladimir Vuskovic

16

Exciting to see the incredible progress from Peptone, a company we’ve backed since the early days at Walden Catalyst Ventures. Peptone just announced Peptron-O, developed in collaboration with NVIDIA — a groundbreaking AI engine that brings ensemble-first structure prediction to the disordered proteome. This is a major scientific and technological milestone, targeting one of the most complex and previously inaccessible areas of biology: intrinsically disordered proteins (IDPs). By shifting the paradigm from static models to dynamic ensembles, Peptron-O opens new doors for drug discovery — especially in areas like oncology and neurodegeneration, where IDPs play critical roles. We’re proud to have supported Peptone on their mission to build the infrastructure for next-gen molecular physics. Their work is a great example of how deep tech can transform life sciences. Read more about the announcement: 👉 https://lnkd.in/eygnKQ67 #AI #Biotech #DrugDiscovery #DeepTech #LifeSciences #WaldenCatalyst #Peptone #NVIDIA #PeptronO with: Kamil Tamiola, Ph.D. Francis Ho Shankar Chandran Nicolas Autret Andy Kau Roni Hefetz Victoria Slivkoff

56

1 Comment

Jazzed for this keynote by Jeff Erhardt at our 12th annual LDV Capital Vision Summit: The Materials Innovation Gap: Why AI Predictions Aren't Enough — And What It Will Take to Transform a $6 Trillion Industry RSVP: https://lnkd.in/eeUHN_m AI is transforming industry after industry, and generating enormous excitement in the world of materials science. Machine learning models can now predict millions of new material candidates with unprecedented speed and accuracy. At the same time, programmable material classes, including those recognized by the 2025 The Nobel Prize in Chemistry, could unlock breakthroughs in some of the most pressing challenges facing civilization, from environmental remediation to the energy transition. But their impact is constrained by something prediction alone cannot resolve: new materials must be physically produced, scaled, integrated, and qualified within the larger systems into which they are deployed, a journey that remains extraordinarily long, expensive, and unpredictable. Jeff will argue that closing this materials innovation gap requires treating AI and physical experimentation as equal partners, and building a new model of collaborative development that connects materials innovators and industrial partners far earlier in the discovery process.

12

SF AI-Native Operator Takeaway #2: In AI-native PLG, the hard part isn’t conversion... it’s discovery. Many AI-native teams are still talking about PLG using a classic SaaS mental model, but based on operator conversations in SF, that model is starting to break down in fairly obvious ways. The biggest bottleneck right now isn’t conversion. It’s discovery. In traditional PLG, users generally understood the category before they ever signed up. The problem was obvious, the product’s value was legible from the homepage, and the “aha” moment tended to show up quickly in first use. In that world, PLG meant optimizing onboarding, reducing friction, and improving free-to-paid conversion because user intent already existed. AI changes that assumption. In AI-native products, users are often curious but unclear. They don’t yet know what’s possible, value depends heavily on workflow, context, data, and role, and the product can feel abstract until it’s applied directly to their job. As a result, many users stall not because the product isn’t valuable, but because they haven’t discovered how it fits into their world and how they can't live without it. This is the real distinction people kept coming back to. PLG conversion answers, “Is this worth paying for?” PLG discovery answers, “What problem does this solve for me, right now?” What’s working best in practice is less about funnel polish and more about clarity up front: role- or workflow-specific entry points, guided examples instead of blank states, and opinionated first actions that show users a concrete outcome before asking them to explore. This also explains a broader pattern across AI-native companies. Forward-deployed teams and services-heavy delivery aren’t just implementation tools; they’re discovery mechanisms. They translate abstract AI capability into concrete workflow value, observe real use cases users wouldn’t self-discover, and feed those learnings back into what eventually becomes productized. PLG isn’t going away, but in AI-native companies it’s being redefined. Self-serve no longer means self-explanatory. Education becomes part of the product, and discovery has to come before optimization. The teams making progress aren’t obsessing over conversion rates yet. They’re focused on whether users see themselves in the product, how quickly they reach a meaningful outcome, and whether the product helps users get to a meaningful outcome for themselves quickly, without too much guesswork. Bottom line: in AI, PLG is less about removing conversion friction early and much more about creating understanding first. Once they understand, they may be hooked. Tomorrow is my last SF AI operator takeaway focusing on everyone's favorite topic du jour: 996 work schedules.

15

1 Comment

🎙️ Latest MAD pod: Evan Kaplan (CEO, InfluxData) breaks down InfluxDB 3.0: • Rust rewrite ⚙️ • Object-storage first (S3, Iceberg) ☁️ • Lessons learned building an open source business 🤕 • From hobbyists to 1 M+ daily installs 🛠️ • Why hyperscalers partner instead of compete 🤝 IoT, robotics & real-time analytics FTW! 🎧/📺 → YouTube: https://lnkd.in/eFb9g7CJ Spotify: https://lnkd.in/eWPkSaMG Apple Podcasts: https://lnkd.in/egk9nTzJ #IoT #TimeSeries #Databases #OpenSource #DataEngineering

87

4 Comments

We’re thrilled to share that Usul (YC S24) has raised $3.3M in Seed funding led by Scout Ventures with participation from BVVC, Teamworthy Ventures, True Ventures, Y Combinator and more. Usul is the modern platform for defense contract capture and federal growth. Our customers have won hundreds of millions in defense contracts for capabilities across Air, Land, Maritime, AI/ML and more. When Oliver Gomez, Joonghyun Lee and I joined forces during Stanford’s Hacking 4 Defense class, we interviewed 300+ defense companies to intimately understand how difficult it is for national security solutions to get deployed quickly at scale. In normal capital markets this is called dead weight loss. In warfighting, this costs lives. We're building an elite team of hackers and defense enthusiasts in San Francisco. If you’re interested in powering the next 1000 defense primes, email us at founders@usul.com or dm me. Read more about our journey in Erin O'Brien's Tectonic article below in the comments. Lastly, thank you to our amazing cap table for joining the ship. Bradley C. Harrison, Cody H., Nate George, Zach Beecher, Michael Keane, Joe Musselman, Gene Ebersole, Nick Nowlin, Thomas Lehrman, Stephen Schmalhofer, Michael Montano, Gustaf Alströmer 🇺🇦, Jared Friedman, Steve Blank, Jacqueline Tame, Jack Shanahan, Peter Newell, Steve Weinstein, Jeff Decker, Ivan Kirigin, Jeff Heitzman, Keith W. Gibson, CFCM, SCPM, John Tincoff, Michael Malis, Austen Allred, Tom Hsieh, Donald Donckers

448

77 Comments

A lot of VCs use founder age as a proxy for how well they’ll leverage LLMs and the latest AI. In my experience, the stronger signal is how technical they are. Technical founders understand how LLMs work and can exploit them fully; non-technical founders see magic, not mechanisms.

49

11 Comments

They raised $10m, moved billions of dollars through stablecoin rails, and didn't tell anyone. Head to Spotify for an inside look at how Zach Walsh, Mohamed Afifi, and the HIFI team are enabling companies with infrastructure for programmable money

71

2 Comments

Seyvik caught up with Brent Willess from Innospark Ventures at the Industrial AI Builders event a few weeks ago.. -Brent shares Innospark's focus on Physical AI, at the intersection of bits & atoms, and how founders lead the way into the future. Grow the Boston ecosystem. Build great companies here. Tell a friend!

11

1 Comment

🚀 Small Teams, Big Exits: The latest data from CB Insights shows that in Q1 2025, tech companies acquired for $100M+ had a median headcount of just 100 employees. This underscores a shift toward leaner operations and capital efficiency in the tech industry. Notably, the valuation per employee in these deals has been rising, reflecting the high value placed on compact, innovative teams. This trend challenges the traditional notion that bigger teams equate to higher valuations.​ https://lnkd.in/g7zf-fir How do you see this influencing startup strategies and investor priorities in the coming years?​ #TechMergers #LeanStartups #Innovation

6