Thrilled to share that I’ve joined @felicis. I’ve long admired how the firm works: moving fast, believing early, and truly supporting founders as people.
It’s an exciting time to build with Felicis, and to back the next generation of defining AI companies.
Power scarcity will define the next phase of AI development. For the past decade, we've assumed compute was the scarce input. Now, efficiency will be the fastest source of new capacity.
How do we get more tokens per watt? There are opportunities at every layer of the stack, as well as for companies like @CrusoeAI that reimagine it entirely. In my latest piece for @Felicis, we examine the approximate, directional gains that we think each of these technologies could have against our current baseline: felicis.link/WprlG72
My first reaction will probably be yours: it's clear why so many founders and investors are talking about memory. But at scale and layered atop one another, the technologies at the bottom of this chart also present a huge opportunity (and less of the wholesale change required by inference chips).
The AI buildout will be constrained by physical infrastructure for the rest of this decade. If you're building anywhere along this chain, we'd like to talk.
We’re delighted to announce that Fractile has raised $720M in a Series C led by @thrivecapital@foundersfund@lightspeedvp and @redpoint, with further participation by brilliant new and existing investors.
1/ We asked founders which seed-stage companies they’re most excited about, and they told us they have AI infrastructure on their minds. This isn’t surprising: Agents have left the chatbox and are now doing longer, more important work. At the same time, we’re in one of the most compute-scarce periods we’ve seen, with capacity sold out into 2028. The companies that nail memory, optimization, routing, and security aren’t just the ones that will win: they’re critical to the entire ecosystem.
The entire Emergent list from @NewcomerMedia and @Felicis is a great glimpse into the mind of founders, and the state of the early ecosystem: felicis.link/UGNhNgT
Chip designers have relied on formal verification for decades. @[Atalanta] is bringing it to AI and critical infrastructure, so “Can I trust this?” gets answered with proof.
Proud to be true believers of @anjaninna, Jonathan Ring & team since seed. Congrats on two years! 🧡
x.com/anjaninna/status/21180…
Two years ago to the day, Jonathan Ring and I made the decision to start @AtalantaTech. Building on our work at the White House that brought formal verification into the mainstream discourse, we wanted to keep answering the hardest questions facing our society: Can I trust this AI? Is this nuclear reactor safe? Will this satellite survive an attack? What happens when something goes wrong?
Today, we are proud to say that Atalanta is in the real world, where our product Argo is on the front lines answering those very questions. Hear directly from our customers and partners at the world’s most important institutions about how Argo is helping them win.
AtalantaAssembly.com
Concurrency limits in voice AI come down to lead time. The slower a GPU starts, the more spares you keep warm. Safety stock, basically.
@Phonic_Co cut model startup from 5.6 min to 26 sec. Same 10% buffer: ~800 stranded calls at peak vs zero.
Moin Nadeem @moinnadeem Nov 3 Article Concurrency Limits: The Hidden Tax in Voice AI (and how we solved it!) By Kendrick Shen and Moin Nadeem Interested in reading this in an interactive form? This blog post is also available on our website TL;DR Voice traffic spikes without warning, and that's exactly when
Voice agents usually trade intelligence for naturalness. Phonic-1 doesn't. A 10-person team rebuilt the stack, from data to kernels, around problems that only show up in real calls. Congrats @phonic_co! We at @felicis are proud to support the team.
Introducing Phonic-1, the most advanced speech-to-speech model for agentic tasks — from @phonic_co, the speech-to-speech model company.
We’ve had a breakout year and raised a $27.5M Series A led by @firstround with participation from @Lux_Capital, @felicis, and Definition.
The physical world is getting a digital workforce.
Magentic is building the digital workforce for manufacturing, supply chain, and finance. Its AI workers, called Mages, diagnose problems, gather evidence, plan the next step, and act across fragmented enterprise systems, while people stay in command.
Companies including Coca-Cola Europacific Partners and Siemens are already working with Magentic.
We at @Felicis are proud to lead Magentic’s Series A. Congratulations to Robin Van Aeken, Odhran O’Donoghue, and the entire Magentic team. The physical world is ready for agents. cc: @speechu@FHaskaramanfelicis.link/89PPsGd
I spent years advising companies on supply chain and saw how much critical work depended on fragmented data and manual effort. With Magentic, the physical world gets a digital workforce. We at @Felicis are proud to lead the Series A.
felicis.link/3Qvmvj1
Introducing the @Felicis Forecast: our view on what’s coming next. They are not neat categories, but these intersections start to pop up everywhere once you notice them.
1/ Today we launch the @Felicis Forecast. It's our map of the changes that defy neat categorization but matter most to what’s next. Where AI’s potential cuts across industries to collide with consumer, organizational, and socio-economic fault lines. Where the old structures have ruptured, and new ones are just coming into view.
Because even to people like us – generalists who’ve spent the past 20 years studying change – this moment feels different. Everything’s up for grabs. Rules, tidy bell curves, and comfortable patterns no longer apply.
So instead of following patterns, we forecast: 9 themes we just can’t get out of our heads, that seem to pop up everywhere once you notice them.
ordnl.link/Vyjrzep
I’m excited to announce that @Felicis has led Sooth’s Seed round!
The world is being asked to make more, faster, higher-stakes decisions under greater uncertainty. The tools most leaders use to navigate that uncertainty — gut, spreadsheets, the occasional consultant deck — were built for a slower world. To keep up, foresight must become a piece of infrastructure.
Yaser Sheikh and @rsalakhu are building that foresight engine, powered by a time series model that deciphers and weighs real events, signals, and probabilities. This is the next leg in the race to AGI: beyond LLMs and their focus on next-token predictions, beyond coding and chatboxes.
Read more from @asenkut and I here: ordnl.link/l9UQYf7
Small, elite cohort. Big ideas. Applications are now open for Felicis AI Fellows ! If you’re a student with ambition, technical depth, and a point of view on the future of AI, we would love for you to apply !!
felicis.com/fellows
We’re thrilled to share that @felicis partner @FHaskaraman has been named one of Venture Capital Journal’s 2026 Rising Stars in VC, a list that celebrates the industry’s most dynamic talent! ✨