Open source is impacting all layers of the AI stack. @DavidSHershey & I highlight how open source is the biggest driver advancing AI-native infrastructure, why it's happening & what it means. Our market map shows #OSS taking on proprietary products in every major area. #GenAI

Thanks for your insights @AstasiaMyers, @natalievais & @ejn! Founders @KubeCon_ got to benefit from a ton of awesome #VC brainpower on the stage today! Appreciate Kelsey Hightower & @meganreyno for organizing #CNCF StartupFest!

If you'll be in Salt Lake City next week for @CloudNativeFdn's @linuxfoundation's @kubecon_, join me, @natalievais, @AstasiaMyers & @ejn on Tuesday for a VC panel on the $100B opportunity in cloud-native #AI! We'll be discussing the market opportunities, trends & business models in AI that VCs are most excited about right now. #CloudNativeCon

Super excited for @KubeCon_ / #CloudNativeStartupFest coming up on November 12 in Salt Lake City - I'll be speaking / hosting a VC panel with @AstasiaMyers, @natalievais & @ejn! Join us to hear an investor perspective on the biggest opportunities for founders building in #AI & #cloudnative.

Nearly every great product company is a great infrastructure software company. Infrastructure enabled Google to deliver the most relevant search results and the fastest browser, Facebook to provide the most personalized social feed, Netflix to stream with the best viewer experience (minimal lag/interruption), Uber to determine the best real-time ride matches, etc. Infrastructure was key to helping these companies (and many others) win their markets. It’ll be the same in AI. The companies that leverage AI infrastructure in the smartest ways - figuring out the right breakdown and sequence of agentic tasks, utilizing data curation and RAG to get best-in-class model performance, managing multi-model and multi-modal complexity - will gain a strategic / competitive advantage. AI founders: invest in your infrastructure early.

Huge congrats to @feross and the entire @SocketSecurity team on their $40M Series B! The level and pace of their innovation in developer security is really remarkable. The @Unusual_VC team and I are proud to have supported Feross from the very beginning. Thrilled to see them change open source security for everyone!

Pricing AI products is a hot topic right now. How can early-stage AI founders think about it? One way is to start with a tried-and-true framework, which generally categorizes pricing into one of three approaches: 1. Value-based 2. Cost-based 3. Competition-based With #1, the price of your product is primarily driven by the "value" it provides. Typically that might be measured or justified by productivity gains, cost savings, or some other unit. New AI add-ons from Microsoft, ServiceNow, and others are priced this way on top of their existing products based on certain assumptions about productivity increases. With AI, there's also an opportunity to potentially price based on the "work" or outcomes that AI can generate, especially if AI is replacing existing services to which there is a dollar amount attached. That provides support for a high price ceiling - imagine if an AI agent does the work of 10 humans. But in new AI categories (e.g., new infrastructure tools), customer understanding of value-based pricing is likely weak right now, requiring a lot more education. With #2, your COGS is the main input into your pricing, especially when you want to maintain a certain margin profile. This could be relevant for businesses that rely significantly on hosted compute / inference since a customer likely has a reference price for things like GPUs in mind, which may create downward pricing pressure. Note that usage-based pricing has elements of both #1 and #2 but makes total spend unpredictable for customers. What if something the AI does involves a ton of model API calls or GPU cycles? With #3, competitor pricing is the biggest consideration for your pricing, as a means to gaining (or maintaining) market share. With this approach, your pricing relative to others can impact customer perception about whether you're selling a "premium" product (and can command stronger pricing power) or a commodity. It's often a balancing act between being aggressive without being caught in a race to the bottom. Companies in several categories of applied AI are experiencing this dynamic today. These approaches above aren't mutually exclusive and should be considered together to get to both an initial pricing model and level that are reasonable. For early-stage AI founders, it's important to recognize that your early customers want simplicity, transparency, and predictability when it comes to pricing. A rule of thumb is if you already have to educate prospects on a new product, try to minimize the amount you need to explain your pricing early on. Don't focus on extracting as many dollars as you can and instead, optimize for customer adoption and making your early users super successful - that'll give you flexibility in figuring out the best long-term approach to pricing. If you're an AI founder navigating pricing strategy right now, I'd love to hear how you are tackling it!

A common failure mode of technical founders is thinking that if you build a cool product, everyone will want it, i.e. distribution and GTM will take care of itself.  A great product earns you the right to compete and is a necessary but (by itself) insufficient requisite for building a successful company. Why? Because you can have an amazing product but if people aren’t aware of it, they won’t use it. Distribution trumps product. This is counter-intuitive for many founders.  It’s essential to think early on about how you'll drive distribution. This is especially critical in markets where there is a strong incumbent. Entrenched vendors have the high ground and will leverage it when they need to. Just consider this anecdote from a customer in the cloud security market as one of many examples.

What will gross margins for foundation model companies look like in the coming years? Cloud providers serve as an illustrative example and the answer is: probably higher than one would expect.  With OpenAI’s price drops and significant spend on compute, some may have thought that theirs is a low-margin, high-volume business model. But that’s an incorrect assumption.  OpenAI’s gross margin is estimated to be 41% today and expected to grow to 48% next year and 67% in 2028. Microsoft’s OpenAI service is at around 40% today. In 2006, soon after it started, AWS’s gross margin hovered around 47%. In subsequent years through 2014, it was estimated at 48%, 48%, 49%, 49%, 50%, 50.5%, 51%, 53%. This despite more than 40 price drops (presumably to maintain/grow market share in light of competition with an emerging Microsoft Azure and Google Cloud at the time), some of which were over 50% reductions at a time. It’s reasonable to expect a similar trajectory for foundation model companies. Model APIs are analogous to compute instances;  different models are equivalent to different instance types. Model API margins will improve as inference gets cheaper the same way cloud servers got cheaper at economies of scale. OpenAI and others will become more sophisticated at managing capacity/utilization and forecasting demand. Eventually, different pricing tiers for model usage could emerge, similar to how cloud providers offer reserved instance and spot pricing. All of these will help improve margins, even in light of aggressive price reductions. Of course, OpenAI and others are also already moving up the stack to offer services built on top of the API just as the cloud providers did on top of EC2. ChatGPT, potential new agents, possible robotics offerings, etc. are just some examples. This further drives them towards a higher margin profile and increases the opportunity for customer lock-in. So while there’s a lot of attention on OpenAI’s top line and “CapEx” for model training, the margins are key and those look quite promising for them and other foundation model companies. Sources: 

Enjoyed this conversation with @gracegongGG about #AI, #opensource & early-stage investing / company building!

Excited to partner with @spirosx & @maynkag again!

RT @Unusual_VC: Unusual Academy applications are due TONIGHT! Apply by midnight PT to be considered: Building a co…

Every founder's journey is unique. @Unusual_VC was founded by @jvrionis and @jyotibansalsf with the core belief that we could provide meaningful help to early-stage founders - in a tailored, customized way. Unusual Academy brings together an elite group of founders seeking to build world-class businesses and helps those founders connect best practices for company building to actual execution. We partner closely on figuring out how to align product thinking with go-to-market approaches and validate early hypotheses with design partners and customers. Our Academy program includes: - Up to $2M in a SAFE - A series of personalized workshops on customer discovery and go-to-market - Access to a supportive and inclusive community of founder alumni - Expert-led discussions on tackling new challenges with building AI-native products If you're an early-stage founder who wants to accelerate your path to product-market fit, then Unusual Academy is for you. Applications are open and close on 9/9! Learn more at 

Super excited to moderate the VC panel @KubeCon_ @CloudNativeFdn #StartupFest with awesome investors @CrystalHuang (@GVteam), @AstasiaMyers (@felicis) & @natalievais (@sparkcapital)! We'll be talking about the $100B opportunity across cloud-native #AI, #OSS & more. Check out the schedule & speakers for #CloudNativeStartupFest - see you all in November!

RT @jyotibansalsf: Are you a founder looking for $100K - $2M in investment? Unusual Academy is back and we're looking for companies to buil…

I’m super excited for this upcoming cycle of Unusual Academy, which always brings together an elite group of founders who build and learn alongside each other. Applications are now open!

RT @JuliaANeagu: Wayfair is using AI to augment and improve customer service and sales interactions. As AI-enhanced capabilities become int…

Calling all founders! Our team @Unusual_VC put together a very short survey to hear what's important when you're deciding to work with investors. We would appreciate 2 minutes of your time to share your input - thank you!