Identity Just Became the Hardest Problem in Enterprise AI
Nuclear Power Is Now an AI Infrastructure Story.
Identity Just Became the Hardest Problem in Enterprise AI
Israeli startup Oak came out of stealth on July 15 with $60 million in seed funding — co-led by Accel, Greylock, and CRV — to build what it calls an "AI-native Identity Operating System" [1][2]. The pitch: enterprises now have to manage identity not just for humans, but for machine services and, increasingly, autonomous AI agents acting on their behalf. Legacy IAM tools were never built for that.
This is a bet worth taking seriously. As companies move from "AI assistant answers a question" to "AI agent has credentials, permissions, and the ability to take action across systems," the old perimeter-based identity model breaks completely. Oak's already GA with enterprise customers, not just a deck — founders Shai Morag and Tal Marom clearly saw this coming early [3].
The size of the seed round is itself a signal. VCs aren't waiting for proof this problem is real; they're pricing in the assumption that agent sprawl inside the enterprise is inevitable within 12-18 months, and whoever controls the identity layer controls a chokepoint that's hard to rip out later.
Nuclear Power Is Now an AI Infrastructure Story
Valar Atomics is reportedly in talks to raise roughly $1 billion at a $6 billion valuation, with Sequoia expected to lead [1][2]. That's a 3x valuation jump from its $450 million raise in March, when it was valued at $2 billion. The three-year-old, El Segundo-based company builds small modular reactors aimed squarely at data center and industrial power demand, and has already demonstrated powering Nvidia AI chips [3].
This isn't really a nuclear story — it's an AI compute story wearing a nuclear costume. The bottleneck on frontier AI scaling stopped being chips or capital a while ago; it's power. Sequoia leading a $1B round on an SMR company tells you where sophisticated capital thinks the next multi-year constraint sits.
Expect more of this. Every hyperscaler is quietly signing power deals, and startups solving "how do we get gigawatts to a data center in 3 years instead of 10" are becoming as fundable as the model labs themselves.
What This Means For Your Business
Three stories, one thread: the infrastructure of AI is maturing faster than the governance and identity layers meant to control it. Governments are scrambling to vet models before release, startups are racing to solve who — or what — is allowed to act on a company's behalf, and investors are pouring billions into the physical plumbing (power) needed to keep training and running these systems at all. None of this is about writing better code. It's about building and governing systems that act autonomously at scale.
For companies making AI decisions right now, the practical implication is this: identity and access management for agents is not a "someday" problem — it's a "this quarter" problem if you're deploying any agentic workflows. Oak's raise is a leading indicator, not a curiosity. Similarly, if you're building products that depend on frontier model access, pay attention to the NIST-vs-NSA fight in Washington — the outcome will shape release timelines and possibly access tiers for who gets frontier capability first.
The bigger shift underneath all three stories: value is moving away from the artifact (the model, the code) and toward the orchestration layer — who manages identity, who manages risk review, who manages the power and infra pipeline. That's exactly the shift we've been tracking. Code is commodity. Judgment — about what to vet, what to trust, and what to build around — is the scarce resource.
Key takeaway: As AI models and agents proliferate, the real competitive advantage is shifting from writing the software to governing and orchestrating it — identity, oversight, and infrastructure are becoming the new moats.
Sources
- https://www.nytimes.com/2026/05/04/technology/trump-ai-models.html
- https://www.cnbc.com/2026/07/17/white-house-ai-access-anthropic-openai.html
- https://www.pbs.org/newshour/nation/trump-signs-executive-order-that-allows-voluntary-federal-vetting-of-top-ai-models-for-national-security-risks
- https://www.prnewswire.com/news-releases/oak-raises-60m-in-seed-funding-to-build-the-ai-native-identity-operating-system-302826349.html
- https://thenextweb.com/news/oak-60m-seed-ai-native-identity-platform
- https://regtechanalyst.com/oak-exits-stealth-with-60m-to-rebuild-identity-security/
- https://techcrunch.com/2026/07/17/nuclear-startup-valar-atomics-in-talks-to-raise-new-funding-at-6b-valuation/
- https://aiweekly.co/alerts/valar-atomics-eyes-6b-valuation-with-sequoia-led-1b-round
- https://www.theinformation.com/articles/nuclear-startup-valar-atomics-talks-6-billion-valuation-power-milestone
Stay ahead of AI
No spam. Unsubscribe anytime.
Want to go deeper?
Reading the news is one thing. Exploring the frontier is another. See what we're building.