Series A activity: Week of September…

1. Key Themes

AI infrastructure is shifting from model quality to operating economics (compute, power, monitoring)

The week's largest and most-discussed deals aren't foundation model companies — they're the picks-and-shovels layer around inference cost, energy, and reliability. As the article puts it on Euclyd: "inference costs are becoming the bottleneck for shipping AI features, and enterprises are starting to care as much about watts and latency as raw benchmark scores." Rune's pitch reinforces this: "'power-first' compute is starting to look like a prerequisite rather than a nice-to-have."

AI agents are moving from novelty to production, creating a new observability/governance stack

Raindrop's raise signals that autonomous agents now carry real operational risk, requiring dedicated tooling. "As teams move from chatbots to autonomous workflows, the failure modes shift from 'bad answer' to costly real-world actions, and the observability layer is becoming mandatory." Relatedly, AIUC's $40M raise for "AI underwriting and insurance for secure enterprise AI adoption" shows risk-transfer products emerging alongside monitoring tools for the same trend.

Europe (and beyond-Silicon-Valley geography generally) is a serious hub for deep-tech/semiconductor Series As

Several large or notable rounds came from outside the US — Euclyd (Eindhoven), Chift (Brussels), SYNALOGiK (Hereford, UK), Veridion (Bucharest), Treble (Reykjavík), TUSK IC (Antwerp) — underscoring that hardware-adjacent AI infra isn't a US-only story. On Euclyd specifically: "Being based in Eindhoven puts it in the middle of Europe's deep semiconductor supply chain, which can shorten the path from architecture to real hardware."

Large outlier deals are skewing the market's averages

With 35 deals raising $1.3B, the mean ($38.5M) is nearly double the median ($20.0M), indicating a handful of mega-rounds (Euclyd's $231.1M, barq's $329.5M) are pulling the average up while typical Series As remain closer to $20M — a reminder that headline averages can mask normal deal sizing.

2. Contrarian Perspectives

Energy volatility as a feature, not a bug

Most infrastructure narratives treat grid constraints and renewable intermittency as a problem to be engineered around. Rune inverts this: "Instead of treating energy volatility as a problem, the system is designed to chase it - so compute loads can soak up excess wind/solar and back off when the grid tightens." This reframes curtailed/cheap power as a monetizable resource rather than a limitation, positioning compute demand as a grid-balancing tool rather than a grid-straining one.

3. Companies Identified

Euclyd — AI silicon and systems platform for efficient foundation models, based in Eindhoven, Netherlands. Mentioned as one of the week's "Three Most Interesting Deals" for its scale and end-to-end ambition. "The $231.1M Series A led by Innovation Industries with participation from Samsung and others suggests this is more than a chip project - it's a serious attempt to deliver an end-to-end compute option for foundation-model workloads."
Raindrop — Monitoring platform for AI agents, based in San Jose, CA. Highlighted as emblematic of the shift toward agent observability. "Raindrop raised $50M in a CRV-led round, a notable check for infrastructure that's quickly becoming a standard part of the agent stack."
Rune — Computation technology monetizing surplus clean power, based in Mountain View, CA. Cited as a leading example of power-aware compute infrastructure. "Rune's $40M round led by Spark Capital underscores the growing appetite for infrastructure that ties compute economics directly to clean energy availability."
barq — Saudi digital payments and money-transfer app licensed by SAMA, based in Riyadh. Notable for raising the largest round of the week ($329.5M) at a $1.9B valuation, signaling strength in Gulf fintech.
AIUC — AI underwriting and insurance for secure enterprise AI adoption, based in San Francisco. Notable as a novel category (AI risk/insurance) emerging alongside agent infrastructure investment.
Watney — Autonomous robotics for data center infrastructure management, based in San Francisco. Notable for an unusually large ($80M) raise in the physical/robotics layer of AI infrastructure.

4. People Identified

Aaron Harris — Co-author of the newsletter/report. Credited alongside Jacob Dennis for compiling and analyzing the week's Series A data.
Jacob Dennis — Co-author of the newsletter/report. Credited alongside Aaron Harris for the analysis, including selection of the "Three Most Interesting Deals."

5. Operating Insights

Bundle hardware with software stack, don't just sell chips: Euclyd's approach — "pairing custom chips with the software stack that actually makes them usable in production" — suggests founders in deep-tech/silicon should design for end-to-end usability from day one rather than treating go-to-market as an afterthought to the chip.
Build observability before agents reach production-critical scale: Raindrop's traction implies that founders shipping agentic products should invest early in guardrails and debugging tools — "debugging multi-step agent runs, catching regressions, and setting guardrails before agents touch production systems" — since this is "becoming mandatory," not optional.
Geographic proximity to supply chain can be a structural advantage: For hardware-adjacent startups, location near a manufacturing/supply ecosystem (as with Euclyd in Eindhoven) can materially compress the timeline "from architecture to real hardware," a factor entrepreneurs should weigh in HQ decisions.

6. Overlooked Insights

Vertical AI applications are quietly proliferating across unsexy industries: Deals like Kastle (mortgage lending workflow automation), Magentic (procurement optimization), Treble (acoustic simulation for AEC), and Synapse Analytics (MLOps/video analytics for Cairo-based businesses) show AI Series A activity extending well beyond flashy infra plays into specific, often overlooked verticals — a sign that AI application-layer investing is broadening geographically and by sector even as headlines focus on infra megadeals.
Crypto/fintech infrastructure tying banks to blockchains is drawing traditional financial backers: Velocity's raise — "connecting banks and blockchains for capital management" — drew investors like Visa, Ripple, and Circle Ventures, suggesting incumbent financial players are placing early, direct bets on crypto-bank interoperability rather than waiting on the sidelines.