Tesla Faces New Federal Crash Probe

1. Key Themes

Defense Tech Is a Generational Investment Opportunity — With Unprecedented Valuations

The newsletter highlights multiple defense-tech companies raising at eye-popping valuations, with investors treating the sector with the urgency typically reserved for consumer internet. Mach Industries closed a $300M Series C at a $1.8B valuation, while Castelion — only three years old — is reportedly seeking a $12B+ valuation.

"Three years later, his company, Mach Industries, is running six weapons programs and earlier this month closed a $300 million Series C round at a $1.8 billion valuation. The startup has now raised roughly $485 million altogether."

"Castelion, a three-year-old startup based in Torrance, CA, that makes hypersonic missiles, is reportedly in the market to raise a round at a $12+ billion valuation."

AI Talent Wars Are Becoming a Systemic Risk for Big Tech

Two senior Google DeepMind researchers left within a single week — Nobel Prize winner John Jumper to Anthropic and Gemini co-lead Noam Shazeer to OpenAI — triggering a 7% drop in Alphabet's share price. This signals that talent concentration risk is now being priced into public market valuations.

"After the market learned that John Jumper, the Nobel Prize-winning Google DeepMind researcher best known for AlphaFold, was leaving for Anthropic (marking Google's second major AI researcher departure in a week after Gemini co-lead Noam Shazeer left for OpenAI), it sent Alphabet's shares tumbling as much as 7%."

AI Is Reshaping the Entire M&A Due Diligence Stack

Private equity firms are now using AI-generated product replicas to stress-test whether a target company's product has a defensible moat — a direct threat to software businesses whose value rests on complexity rather than network effects or data.

"Private equity investors are using AI-generated replicas of software products to test whether takeover targets have a defensible edge, with Bain consultants 'vibecoding' hundreds of rough prototypes to gauge how easily a company's product could be copied or reshaped by AI."

AI Infrastructure — From Chips to Networking — Is Still in Full Capital Deployment Mode

Multiple infrastructure-layer companies are raising enormous rounds: Groq ($650M for AI inference), Upscale AI ($190M Series A at $2B for AI cluster networking), and Nearfield Instruments ($380M for semiconductor inspection tooling used in AI chip production. This suggests the picks-and-shovels layer of AI remains a primary investment target.

"Upscale AI... builds networking infrastructure that connects accelerators, memory, and storage across large AI clusters to reduce latency and improve hardware utilization for cloud providers, raised a $190 million Series A round at a $2 billion post-money valuation."

"Groq... raised a $650 million round co-led by Disruptive and Infinitum... The company was last valued at $6.9 billion and has recently pivoted toward its neocloud business after Nvidia licensed its language processing technology."

Government Policy Is Actively Accelerating Quantum and Autonomous Systems Investment Timelines

Executive orders from the Trump administration set hard deadlines — 2028 for a research-capable quantum computer and 2031 for quantum-resistant security — creating government-backed demand signals that will shape both venture and procurement decisions.

"President Trump signed two executive orders aimed at accelerating U.S. quantum-computing development, including a 2028 target for a quantum computer powerful enough to conduct scientific research, while also moving up the government's deadline for quantum-resistant security systems to 2031."

2. Contrarian Perspectives

Spreading Across Six Product Lines Is a Strategic Advantage in Defense, Not a Weakness

Conventional startup wisdom demands focus. Mach Industries is explicitly doing the opposite — running six simultaneous weapons programs — and its founder argues that in defense, narrow focus is actually a losing strategy because national security requires a broad portfolio of capabilities deployed against a dynamic adversary.

"'It is a chess game you're playing with an adversary,' he said, 'with hundreds of different products that need to be shipped if we want security.' Pick just one, he suggested, and you've already lost the game."

"Thornton is aware that Mach's diffuse focus creates some lingering questions for outsiders. 'It's very hard,' he volunteered Thursday night."

AI Coding Is Not About Writing Code — It's About Supervising Other Agents

The popular narrative frames AI coding tools as replacements for individual developers writing lines of code. Claude Code's creator Boris Cherny argues the real shift is further upstream: agents will soon supervise other agents in continuous improvement loops — a qualitatively different and more powerful paradigm than current "agentic coding."

"Claude Code creator Boris Cherny says AI coding is moving from agents that write code to 'loops' in which agents continuously prompt and supervise other agents to improve code bases, a shift he argues could be as significant as the move from hand-written source code to agentic programming."

Microsoft Is Betting Against Frontier Model Lock-In — While Everyone Else Chases It

At a moment when most of the industry is racing toward proprietary frontier models, Microsoft CEO Satya Nadella is publicly advocating for an open, multi-model architecture where customers avoid dependency on any single frontier provider. This is a structurally contrarian position from one of the industry's largest players.

"Microsoft CEO Satya Nadella is pushing for a more open, cheaper AI model in which customers can choose among lower-cost systems, run autonomous agents such as Copilot Cowork inside environments they control, and avoid becoming dependent on a handful of frontier-model companies."

3. Companies Identified

Mach Industries | Defense tech startup, Torrance, CA | Featured as primary profile case study for a founder betting on multi-product defense expansion; closed $300M Series C at $1.8B valuation |

"His company, Mach Industries, is running six weapons programs and earlier this month closed a $300 million Series C round at a $1.8 billion valuation."

Castelion | Hypersonic missile startup, Torrance, CA | Mentioned as a signal of extraordinary defense-tech valuations at an early stage |

"A three-year-old startup... that makes hypersonic missiles, is reportedly in the market to raise a round at a $12+ billion valuation."

Groq | AI inference cloud infrastructure, San Jose, CA | Raised $650M; notable pivot away from chip design toward neocloud after Nvidia licensed its core tech |

"The company was last valued at $6.9 billion and has recently pivoted toward its neocloud business after Nvidia licensed its language processing technology and hired away several senior leaders."

Upscale AI | AI cluster networking infrastructure, Santa Clara, CA | Raised $190M Series A at $2B; backed by Nvidia and Salesforce Ventures — signal of strategic validation |

"Builds networking infrastructure that connects accelerators, memory, and storage across large AI clusters to reduce latency and improve hardware utilization."

Nearfield Instruments | Semiconductor inspection equipment, Rotterdam | Raised $380M Series D at $1.6B; critical for yield improvement on advanced AI chips |

"Makes chip inspection equipment that helps semiconductor manufacturers measure tiny 3D structures on wafers, find defects during production, and improve yields for advanced AI chips."

A24 | Indie film studio, New York | Received $75M from Google DeepMind as part of an AI filmmaking tools partnership — signals AI's entry into creative industries |

"Received a $75 million investment from Google DeepMind as part of a partnership to develop AI tools for filmmaking."

AppsFlyer | Mobile marketing analytics, San Francisco | Raised $1B+ Series E at $2.7B; investors include Google, Meta, and Unity — notable that direct competitors are co-investing |

"Secured a $1+ billion Series E round at a $2.7 billion post-money valuation... Investors reportedly include Moloco, Google, Meta, and Unity."

CRED | Indian fintech (credit, lending, wealth), 8 years old | Raised $900M Meta-led round at $4.5B; founder Kunal Shah departing to lead WhatsApp — signals deep strategic integration with Meta |

"Secured a $900 million Meta-led financing... Founder Kunal Shah will step down as CEO to lead WhatsApp."

Isometric | AI-powered carbon removal verification, London | Raised $40M Series A; compresses certification timelines from months to hours — infrastructure for carbon markets |

"Uses AI to check sensor data, satellite images, supply-chain records, and lab results, reducing review times from months to hours."

Bending Spoons | Software acquisition and revamp firm, Milan | Preparing U.S. IPO targeting up to $19B valuation; owns Vimeo, WeTransfer, AOL, Eventbrite |

"Preparing to file for a U.S. IPO that could reportedly raise up to $1.62 billion and value the owner of Vimeo, WeTransfer, AOL, and Eventbrite at as much as $19 billion."

Fomo | Retail crypto trading app, New York | Raised $75M Series B at $550M valuation in just one year, backed by Index, USV, and Benchmark — fast institutional validation in crypto retail |

"A one-year-old New York startup... raised a $75 million Series B round at a $550 million valuation."

Getty Images | Stock photography/media licensing | Deal with OpenAI to display licensed photos in ChatGPT sent stock up 90%; training rights remain undisclosed — a template for IP licensing in the AI era |

"Getty Images struck a multiyear display deal to bring its licensed photos into ChatGPT's search and discovery experiences, sending its beaten-down stock shooting up 90%, though the agreement does not publicly say whether Getty's images can be used to train future OpenAI models."

Prosper AI | Healthcare voice AI agents, New York | Raised $30M Series A led by a16z; automates scheduling, billing, and intake for health systems — a16z's continued healthcare AI commitment |

"Deploys voice-based AI agents to handle patient scheduling, insurance verification, billing, and intake workflows for healthcare providers and health systems."

Seedcamp | Early-stage VC firm, London | Raised $320M across two funds; early backer of Revolut, Wise, UiPath — European seed institution scaling follow-on capacity |

"The 19-year-old London seed firm that was an early backer of Revolut, Wise, UiPath, Synthesia, and Fluidstack, raised $320 million."

Lucid Motors | EV manufacturer | Cutting 18% of workforce (~1,500 employees) and eliminating a production shift — ongoing demand/supply misalignment in premium EV market |

"Cutting 18% of its workforce – or about 1,500 employees – and eliminating a second production shift at its Arizona factory."

Oracle | Cloud computing | Cut 13% of workforce (~21,000 employees) in fiscal 2026 while adopting more AI — large-scale AI-driven labor restructuring at enterprise scale |

"Oracle reduced its workforce in fiscal 2026 by 13% – or about 21,000 employees – as the cloud computing company restructured its business and adopted more AI across its operations."

4. People Identified

Ethan Thornton | Founder & CEO, Mach Industries | Profiled for building a multi-program defense startup from age 19, raising $485M total by 22 |

"Ethan Thornton dropped out of MIT at 19 to build weapons... Three years later, his company, Mach Industries, is running six weapons programs."

John Jumper | Nobel Prize-winning researcher, formerly Google DeepMind | Departing to Anthropic; his exit (alongside Noam Shazeer's) triggered a 7% drop in Alphabet shares |

"John Jumper, the Nobel Prize-winning Google DeepMind researcher best known for AlphaFold, was leaving for Anthropic."

Noam Shazeer | Gemini co-lead, formerly Google DeepMind | Left for OpenAI the same week as Jumper; together, the two departures represent a significant talent hemorrhage for Google |

"Marking Google's second major AI researcher departure in a week after Gemini co-lead Noam Shazeer left for OpenAI."

Satya Nadella | CEO, Microsoft | Publicly advocating for open, multi-model AI to prevent frontier model dependency — a strategically significant position |

"Pushing for a more open, cheaper AI model in which customers can choose among lower-cost systems... and avoid becoming dependent on a handful of frontier-model companies."

Boris Cherny | Creator, Claude Code (Anthropic) | Articulating a next-generation vision for AI coding: agent-supervising-agent "loops" as the paradigm shift beyond current agentic coding |

"AI coding is moving from agents that write code to 'loops' in which agents continuously prompt and supervise other agents to improve code bases."

Kunal Shah | Founder, CRED | Stepping down as CEO to lead WhatsApp after Meta's $900M investment — signals deep executive-level integration between Meta and Indian fintech ecosystem |

"Founder Kunal Shah will step down as CEO to lead WhatsApp, with Miten Sampat taking over as interim CEO."

John Doerr | Venture investor | Personally investing in Isometric's Series A alongside institutional funds — signals conviction in AI-powered carbon market infrastructure |

"AVP was the deal lead, with John Doerr and Walter Kortschak as well as previous investors Plural and Lowercarbon Capital also stepping up."

5. Operating Insights

Defense Portfolio Strategy: Breadth as a Competitive Moat

For founders and investors in defense tech, the Mach Industries case challenges the standard "do one thing well" playbook. In adversarial, government-procurement markets, covering multiple capability areas simultaneously may be necessary to remain relevant as requirements shift — and to build the kind of systemic relationship with the DoD that leads to rate-production contracts.

"He doesn't think defense rewards the kind of single-minded focus that rocket launch, say, demands. 'It is a chess game you're playing with an adversary,' he said, 'with hundreds of different products that need to be shipped if we want security.'"

AI Is Now a Due Diligence Tool — Founders Must Stress-Test Their Own Moats

Any software company seeking PE or strategic investment should proactively vibe-code a replica of their own product before investors do it first. If a rough AI-generated prototype can approximate your core functionality, your defensibility narrative needs to be rebuilt around data, distribution, or network effects — not product complexity alone.

"Private equity investors are using AI-generated replicas of software products to test whether takeover targets have a defensible edge, with Bain consultants 'vibecoding' hundreds of rough prototypes to gauge how easily a company's product could be copied or reshaped by AI."

AI-Driven Workforce Restructuring Is Now Quantifiable at Enterprise Scale

Oracle's 13% headcount reduction (21,000 employees) explicitly attributed to AI adoption provides a concrete benchmark for operators modeling AI-driven efficiency gains. This is no longer theoretical — it is being disclosed in regulatory filings.

"New filings reveal Oracle reduced its workforce in fiscal 2026 by 13% – or about 21,000 employees – as the cloud computing company restructured its business and adopted more AI across its operations."

6. Overlooked Insights

Quantum-Resistant Security Has a Hard Government Deadline: 2031

Buried in the quantum computing executive order is a 2031 mandate for quantum-resistant security systems across the government. This is an investable procurement timeline — companies building post-quantum cryptography infrastructure now have a clear government buying signal with a defined deadline, creating a potential wave of enterprise and government contracts over the next five years.

"Moving up the government's deadline for quantum-resistant security systems to 2031."

Five Eyes Intelligence Agencies Warn AI Cyber Risk Is a Months-Not-Years Problem

The joint statement from cybersecurity agencies across the U.S., U.K., Canada, Australia, and New Zealand is unusually urgent in its framing — "within months, not years" — and positions AI-enabled cyber threats as a near-term strategic emergency, not a future scenario. This is a strong demand signal for AI-native cybersecurity companies.

"Frontier AI models could transform offensive and defensive cyber capabilities within months, not years, urging businesses and governments to treat AI-enabled cyber risk as an urgent strategic threat."