M12, Microsoft's Venture Fund

AI compute isn’t just a software problem. It’s a physics problem. As demand accelerates, traditional lithography is approaching both physical and economic limits. Incremental gains won’t be enough to support the next generation of AI systems. We’re excited to back Lace Lithography in its $40M Series A as they take a fundamentally different approach, using atom-beam lithography instead of light to unlock a new path for chip scaling. This is the kind of deep, physics-driven innovation that can reshape what’s possible for AI infrastructure. At M12, we believe progress in AI depends on advances across the full stack, from models to systems to the underlying hardware. Lace sits at the center of that shift, combining novel physics with AI-driven design to enable more efficient, scalable compute. More in Reuters: https://lnkd.in/eBhV24D5

Access to the federal market shouldn’t be a bottleneck for innovation. We’re excited to back Knox Systems, Inc. in their $25M Series A as they work to remove one of the biggest constraints: the time and cost of FedRAMP authorization. Today, bringing modern AI and SaaS into government environments can take years and significant investment — limiting what actually makes it into production. Knox is changing that with an AI-managed cloud and pre-authorized compliance boundary, enabling companies to reach FedRAMP in a fraction of the time and cost. We believe the next wave of AI will depend on infrastructure that meets the strictest security requirements without slowing progress — especially in regulated environments like the public sector. Congrats to Irina Denisenko and team! https://lnkd.in/eQXDrVn3

Grateful to join a thoughtful group of investors and operators at The Montgomery Summit this week. At the 2026 Private Capital Outlook panel with Madrona’s Matt McIlwain and Sapphire Ventures’ Jai Das, moderated by Ben Bergman of Business Insider few themes stood out: 1️⃣ Capital is concentrating in big winners but room for very capital efficient companies. AI category leaders are scaling at unprecedented speed, while lean, capital‑efficient companies are quietly reaching meaningful scale. The old SaaS benchmarks don’t apply —with speed, growth, adaptability becoming real differentiators. 2️⃣ Durability in AI is hard to earn. As Matt pointed out, in six years of Madrona’s AI 40 list, only Databricks has shown up every year. Durable flywheels—data, learning, and reasoning—are existential at the application layer as models commoditize and the model “blast radius” expands. 3️⃣ What’s up ahead? Are neo‑labs overcapitalized? What breaks when agents make trillions of decisions? And what new software and compute paradigms will bend the token cost curve? Huge thanks to the Montgomery Summit team for convening such a wide range of perspectives. I’m excited to see how these shifts show up across M12, Microsoft's Venture Fund’s portfolio—and the broader AI ecosystem—in the year ahead.

We’re excited to welcome several new faces to the M12 team as we continue investing in the next generation of AI and deep tech innovators: Alessandro Levi, Partner Alessandro invests in AI, deep tech, and next-gen computing platforms. Prior to M12, he invested in breakthrough science companies at Longview Innovation and Prime Movers Lab and previously founded deep tech startup Semplus Corp., bringing both investor and founder perspectives to supporting ambitious technical founders. Puneet T., Senior Portfolio Development Manager Puneet supports M12 portfolio companies across AI applications and next-generation LLM systems. He brings operating experience from Amazon, Redfin, and T-Mobile, where he built and scaled products across voice AI, marketplaces, and digital infrastructure. Tracy Liu, Senior Associate Tracy focuses on early-stage investments across the AI application and infrastructure stack. Prior to M12, she invested in growth-stage fintech and enterprise software at WestCap and previously worked on AI-driven recruiting products at Gem Software. Yuhang Zhang, Associate Yuhang's area of focus is AI applications and cybersecurity investments. Prior to M12, he was an investor at Amex Ventures, contributing to investments including Replit, Decagon, and Moderne, and previously worked at HGGC and McKinsey. We’re thrilled to have them on board and look forward to the experience and perspective they'll bring as we partner with founders shaping the future of AI. Welcome! Learn more about our team here: https://m12.vc/team/

Big news today from our portfolio company, Typeface! 👏 The team introduced Typeface Arc, a Marketing Orchestration Engine designed to help enterprises operationalize AI across the full marketing lifecycle. Many organizations now have powerful AI tools for content creation. But the bigger challenge is coordinating the dozens of teams, workflows, and systems involved in launching a campaign. Marketing, IT, and creative teams often operate in silos, making it hard to scale AI beyond experimentation. Typeface Arc tackles that orchestration layer — connecting brand intelligence, AI agents, and workflows so campaigns can move from strategy to creation to activation in a governed system. We’re excited to support companies building the infrastructure that helps enterprises move from AI pilots to real operational adoption.

Looking forward to joining leaders across venture, enterprise innovation, and AI next week at The Montgomery Summit, where we’ll explore the future of innovation, technology, and capital. I’ll be speaking on the “2026 Private Capital Outlook” panel, discussing questions top of mind for investors this year: -How do you underwrite AI companies when the underlying tech stack changes every 6–12 months? -With "mega rounds" and $100M seed bets now common, how does capital concentration affect returns, resilience, and portfolio construction? -How does GTM and business models and differ for AI Native from SaaS? Excited to join an outstanding group of leaders bringing perspectives from venture, growth equity, corporate, and sovereign capital, including Navneet Govil, Jai Das, and moderator Ben Bergman.

At the Inworld AI Consumer AI Accelerator, M12 Managing Partner Michael Stewart joined Speedinvest Partner Sameer Singh and Inworld AI Director of Partnerships Cheryl Fichter for a candid conversation on what actually matters to VCs evaluating consumer AI startups in 2026. With a cohort of 32 startups generating $50M+ in ARR, the discussion cut through the hype to focus on how investors assess risk, traction, and durability in a fast-moving market. Key takeaways: 1️⃣ Raise now, then conserve. If you’re fundraising, aim to do it in H1 2026 while the market is receptive, then shift to self-sustainability and disciplined burn to stay resilient if conditions tighten. 2️⃣ Authentic traction beats hype. Build for users, not investors. Prioritize engagement and retention over vanity metrics, especially since valuations can be decoupled from fundamentals. Be brutally honest about whether users are truly attached. 3️⃣ Fundraising is a numbers game. Repeat founders or hyper-growth can run a targeted process. Everyone else should expect to run a wide net of 50+ VCs because investors filter quickly using signals like profile, narrative, or growth velocity. 4️⃣ Mind inference costs early. Products that require inference on every interaction can be vulnerable as costs shift. Build engagement modes that don’t always require inference, and explore fallback options like local inference where it fits. 5️⃣ Seed → Series A = Product → Scale. The bar becomes a repeatable distribution engine and a clear path to monetization, plus the ability to “graduate” from the startup bubble into a public-facing brand where the public’s reaction is the real test. 📍 On March 3, these 32 startups will pitch at Inworld's Demo Day, The cohort has raised $30M+ to date, with founding teams from NVIDIA, Microsoft, Stanford, YC, and more. If you’re an operator or investor in consumer AI, request an invite to join here: https://lnkd.in/dESya3AV?

Speed, cost, and quality have traditionally been at odds in reasoning models. Today, Inception launched Mercury 2, the first diffusion-based reasoning LLM designed to break that tradeoff. Mercury 2 delivers ~1,000 tokens/second using a diffusion-forward approach that drives faster inference, lower cost, and more predictable production performance. That unlocks: → Reasoning without painful latency → Agent loops that stay fast instead of compounding delays → Voice and coding experiences that finally feel real-time This kind of real-time reasoning infrastructure aligns directly with our thesis on the systems that will enable agentic workloads at scale. We’re proud to support Inception AI as they bring diffusion-first reasoning into production. Huge congratulations to Stefano Ermon, Aditya Grover, Volodymyr Kuleshov, and the entire Inception team. Read more about the launch in Bloomberg: https://lnkd.in/eHKJY2kJ

Supply chains have become one of the most strategic, and strained, functions in global business. Tariffs, geopolitical volatility, and rising complexity have exposed how much critical work is still handled manually. That’s why we’re excited to back Didero in their $30M Series A. Didero is applying AI agents to one of the clearest real-world use cases for generative AI: enterprise procurement. Today, procurement teams spend the majority of their time on manual data entry, supplier follow-ups, and shipment tracking across hundreds of suppliers. Didero lets AI handle that heavy lifting – so humans can focus on forecasting, sourcing strategy, and dealmaking. We look forward to seeing how AI-native supply chain infrastructure redefines global trade. Congratulations to Tim Spencer, Tom Petit, Lorenz Pallhuber and the Didero team on this achievement. Read more: https://lnkd.in/etz9ypaM