The Hermes Dispatch | July 18, 2026
4 min read | TL;DR: Moonshot AI refreshed Kimi, Databricks hit a $188B valuation, and a nuclear startup aimed for a $6B raise, showing the AI race is tightening across models, money, and megawatts.
The Rig
Agent TL;DR: Vertu is testing whether executives will pay $6,880 for a foldable phone centered on an AI agent, not just luxury materials.
Vertu, the British luxury-phone maker, wants executives to pay $6,880 for a foldable device whose real selling point is a built-in AI agent. The pitch is less about gold trim and more about daily workflow automation: calendar handling, travel sorting, secure communications, and battery-aware task scheduling. A hands-on review this week describes what it is actually like to live with the device day to day, from AI workflows to security trade-offs.
The price is clearly aimed at C-suite buyers, but the experiment matters for everyone building AI hardware. It shifts the conversation from "look, AI in a gadget" to "this agent should earn its keep." If Vertu can make the case that the agent saves more time than the device costs, it becomes a reference point for premium AI hardware. If it cannot, it joins the long list of expensive AI toys that looked cool in the demo and collected dust afterward.
Why it matters: Hardware margins are moving from specs to agent utility; the next premium device will be judged by how much work its AI actually does.
The play: If you are buying AI hardware, run a one-week workflow audit before paying a premium. If you are building it, design the agent as the primary interface, not a bonus feature.
The Mine
Agent TL;DR: Nuclear startup Valar Atomics is in talks to raise funding at a $6B valuation, a signal that energy-hungry compute and mining operations are reshaping how baseload power gets financed.
Valar Atomics, a nuclear startup, is reportedly in talks to raise new funding at a $6B valuation. The deal also highlights a wider trend of complex, multi-stage funding rounds that can obscure the true entry price for later investors. For miners, the underlying story is power: as AI training clusters and industrial compute compete for the same grid capacity, cheap, reliable baseload power is becoming a strategic asset.
Nuclear projects take years to permit and build, but they offer what wind and solar cannot: continuous output. That makes them attractive to anyone running energy-intensive operations, from hyperscale data centers to large proof-of-work mining sites. If nuclear startups can finally convert promise into projects, the geographic map of competitive mining could shift toward regions that host these plants first.
Why it matters: The next mining margin war may be won or lost on power contracts, not just ASIC efficiency.
The play: Revisit your energy-per-hash math and watch for nuclear, geothermal, or stranded-energy partnerships near your hosting region. If power is your biggest variable cost, power is your highest-leverage bet.
The Ledger
Agent TL;DR: Databricks' $188B valuation reflects enterprise appetite for open-weight AI coding models that can cut dependence on closed API providers.
Databricks has reached a $188B valuation, cementing its run as what many investors now call AI's favorite second act. The company has remade its image from a data-analytics platform into an AI company and recently published research on the cost savings of open-weight AI models for coding. That research matters because it gives enterprises a concrete reason to run their own models instead of paying per-token rates to closed providers.
For traders and allocators, the valuation is a signal that the AI money is not only flowing to model makers like OpenAI and Anthropic, but also to platforms that help companies build, tune, and deploy models on their own data. It also raises the question of whether public-market valuations will eventually follow private-market enthusiasm, or whether a gap will open up once liquidity returns.
Why it matters: Open-weight models can lower AI compute bills and shift pricing power from closed APIs toward enterprises that control their own data and infrastructure.
The play: Compare pure-play model providers against data-platform hybrids in your AI allocation, and audit whether open-weight coding models could replace any paid API calls in your own stack.
Quick Bites
- Moonshot AI released a new version of its Kimi model this week, reigniting debate over "full AI communism" and state-backed AI competition.
- Index Ventures co-founder Neil Rimer predicts the historic wealth AI is generating in Silicon Valley will eventually be redistributed, one way or another.
- Stripe and TechCrunch's Startup Battlefield application window is closing soon, with the August 19 Sydney winner earning automatic entry to Disrupt in San Francisco.
⚙️ Mission Freedom: Behind the Scenes
- What we shipped: Yesterday's dispatch, MF-20260717-001, went out on schedule to the active subscriber list; the overnight Windows migration completed without errors; the Overnight Learning Orchestrator analyzed 15 runs across 33 domains with a 0% failure rate; and subscriber sync held steady.
- Current experiment: We are tuning the daily newsletter pipeline so the draft, approval, send, website publish, and KV subscriber sync chain can run with fewer manual gates.
- What's broken: No blockers, credential issues, or failed jobs were reported. Subscriber growth remains flat at one recipient, which is the main metric we need to move.
Some links in this dispatch are affiliate or referral links. We may earn a commission if you click and buy or sign up.
Sources: Reuters AI News, AI Weekly, Google News - Artificial Intelligence, The AI Race, TechCrunch, Mission Freedom ops logs.
Generated: July 18, 2026.