OpenAI Files (Not So) Confidentially to...

1. Key Themes

AI Lab IPO Race Is Heating Up

OpenAI filed confidentially to go public, setting up a direct public-market competition with Anthropic and SpaceX. The filing comes despite significant headwinds: "OpenAI faces questions about its massive compute spending, missed growth targets, governance, lawsuits, and when it can generate more cash than it burns."

VC Valuation Manipulation Is a Systemic Problem, Not an Isolated One

The "dual-pricing" or "dual-tranche" structure — where lead VCs invest at a lower private valuation while announcing a much higher headline number — is gaining public scrutiny. Mercor's Brendan Foody stated: "in the last 6 [months] ive seen a half dozen rounds where sequoia invests in 2 tranches. everyone pretends they only did the higher valuation. founders misrepresent this to their employees & then shop it to angels too." A concrete example: Serval's announced $1B Series B valuation masked a prior Sequoia entry at under $400M just days earlier.

AI Infrastructure Cost Optimization Is Becoming a Fundable Category

The funding of PointFive — "a two-year-old Tel Aviv and New York startup that helps companies uncover and reduce wasteful cloud and AI infrastructure spending" — at a $500M post-money valuation signals investor appetite for the picks-and-shovels layer of AI cost management, not just capability building.

Physical-World AI Simulation Is Attracting Top-Tier Capital

PhysicsX raised a $300M Series C at ~$2.4B valuation for "machine learning models that simulate physical systems to help engineers design, test, and optimize products across aerospace, semiconductors, energy, and manufacturing workflows." The investor syndicate — including Nvidia, Siemens, Applied Materials, General Catalyst, and Atomico — signals this is becoming a consensus infrastructure bet across both strategic and financial capital.

AI Cybersecurity Is a Double-Edged Sword

Anthropic's Mythos Preview model "can now turn newly disclosed software vulnerabilities into working exploits within hours," including generating "a proof-of-concept exploit for a Windows kernel flaw in 31 minutes." This dramatically shrinks the patch window for defenders and validates investment in continuous offensive security testing — exactly what A Security, which raised $37M, is building.

2. Contrarian Perspectives

Apple's AI Slowness Was a Feature, Not a Bug

Conventional wisdom held that Apple was dangerously behind in AI. The article pushes back: "For years, Apple has been accused of being one of the biggest stragglers in the AI arms race... But it's slow-and-steady AI bet is starting to look pretty smart." As rivals face model cost overruns, safety controversies, and governance chaos, Apple's methodical approach may prove more durable — and WWDC 2026 with a new Siri AI may be the payoff moment.

SpaceX's IPO May Be a Capital Trap, Not a Rocket Ship

Renowned short-seller Steve Eisman warns that SpaceX is "far more capital intensive as it pushes into AI infrastructure," with capex rising from "42% of revenue in 2023 to 215% in the most recent quarter." His argument: "SpaceX's future now depends more on AI than on rockets or Starlink" — meaning investors may be buying an AI infrastructure bet at a space-company multiple, with no clear path to capital efficiency.

Headline Valuations in AI Are Largely Fictional

The dual-tranche pricing story is a direct challenge to how the market perceives AI startup valuations. At Aaru, "lead investor Redpoint backed the company at a $450 million valuation despite an announced $1 billion headline price." If the actual lead entry price is routinely 50%+ below the announced valuation, the entire AI unicorn landscape requires significant re-pricing skepticism.

3. Companies Identified

OpenAI

Description: Leading AI lab, developer of ChatGPT
Why mentioned: Filed confidentially to go public; faces scrutiny over spending and governance
Quote: "OpenAI faces questions about its massive compute spending, missed growth targets, governance, lawsuits, and when it can generate more cash than it burns."

Anthropic

Description: AI safety-focused lab, OpenAI rival
Why mentioned: Also filing for IPO; its Mythos cybersecurity model can generate working exploits within hours
Quote: "Anthropic says its Mythos Preview cybersecurity model can now turn newly disclosed software vulnerabilities into working exploits within hours... it generated a proof-of-concept exploit for a Windows kernel flaw in 31 minutes."

Mercor

Description: AI talent platform, last valued at $10 billion
Why mentioned: Co-founder Brendan Foody publicly called out Sequoia's dual-tranche pricing as systemic deception

Sequoia Capital

Description: Elite venture capital firm
Why mentioned: Named in dual-pricing controversy; partner Shaun Maguire offered a partial defense

Serval

Description: AI-driven IT helpdesk startup
Why mentioned: Case study in dual-tranche pricing — announced $1B Series B while Sequoia had entered days earlier at under $400M
Quote: "The company had been valued at less than $400 million as part of a Series A extension in which Sequoia participated — less than half the headline figure."

Aaru

Description: AI startup simulating user behavior for market research
Why mentioned: Second case study in valuation inflation — Redpoint entered at $450M despite a $1B headline
Quote: "Lead investor Redpoint backed the company at a $450 million valuation despite an announced $1 billion headline price."

Databricks

Description: 12-year-old data analytics software company for enterprise AI
Why mentioned: In talks to raise a Series M at $165B–$175B valuation — an extraordinary late-stage private round

PhysicsX

Description: London startup using ML to simulate physical systems for engineering
Why mentioned: Raised $300M Series C at ~$2.4B from a blue-chip syndicate including Nvidia and Siemens

PointFive

Description: Startup helping companies reduce cloud and AI infrastructure waste
Why mentioned: Raised $60M growth round at $500M valuation; represents emerging AI cost-optimization category

Kindred Ventures

Description: San Francisco early-stage and growth VC
Why mentioned: Raised $355M in new funds, largest ever; 2022 Fund III at 5.1x TVPI — well above 90th-percentile benchmark

Bending Spoons

Description: Milan app studio that acquires and operates digital businesses (Eventbrite, Vimeo, WeTransfer, Evernote, AOL)
Why mentioned: Filed to go public; reported $1.31B in 2025 revenue, 500M+ monthly active users, and Q1 profit of $27.4M

Tools for Humanity

Description: Eye-scanning identity startup co-founded by Sam Altman, supporting Worldcoin
Why mentioned: Laying off staff despite $2.5B valuation; "struggles to generate revenue"

SpaceX

Description: Elon Musk's aerospace and now AI infrastructure company
Why mentioned: Subject of Steve Eisman's bearish IPO warning; capex at 215% of revenue

Moonshot AI

Description: Beijing startup developing the Kimi chatbot
Why mentioned: Seeking $2B at $30B valuation — valuation rose from $4B+ in December to $20B (current round) to a targeted $30B

A Security

Description: NYC startup using AI to continuously attack customers' systems to find vulnerabilities
Why mentioned: Raised $37M from Lightspeed and Cyberstarts; backed by Wiz and Cyera CEOs as angels

4. People Identified

Brendan Foody

Description: Co-founder of Mercor (AI talent platform, $10B valuation)
Why mentioned: Publicly called out Sequoia's dual-tranche pricing as systemic deception affecting founders and employees
Quote: "in the last 6 [months] ive seen a half dozen rounds where sequoia invests in 2 tranches. everyone pretends they only did the higher valuation. founders misrepresent this to their employees & then shop it to angels too."

Shaun Maguire

Description: Partner at Sequoia Capital
Why mentioned: Pushed back against the "Sequoia scam" characterization, offering a market-conditions-based explanation
Quote: "What happens is other investors are willing to pay a high price for a hot company — usually AI — at multiples above what we're willing to pay. So we try to decouple the company-building relationship with our partner from the capital, and this leads to two tranches at different valuations in close succession."

Leopold Aschenbrenner

Description: 24-year-old former OpenAI researcher; founder of hedge fund Situational Awareness
Why mentioned: Grew his fund to $20B+ AUM with gains of 1,000%+ since inception; backed by Jane Street with large bets on Anthropic and SK Hynix
Quote: "Leopold Aschenbrenner... has grown his hedge fund, Situational Awareness, to more than $20 billion in assets after gains of more than 1,000% since inception."

Steve Eisman

Description: Investor famous for "The Big Short" trade against subprime mortgages
Why mentioned: Issued a public warning on SpaceX's IPO, flagging a dangerous capex surge tied to AI ambitions
Quote: "Capex has risen from 42% of revenue in 2023 to 215% in the most recent quarter."

Steve Jang & Kanyi Maqubela

Description: Co-founders of Kindred Ventures
Why mentioned: Raised $355M across new funds; their 2022 Fund III achieved 5.1x TVPI, "well above the 90th-percentile benchmark for its vintage"

Sam Bankman-Fried

Description: FTX co-founder, convicted of fraud, sentenced to 25 years
Why mentioned: Formally applied for a presidential pardon from President Trump

5. Operating Insights

Dual-Tranche Structures Create Legal and Reputational Risk for Founders

Founders who participate in or passively accept dual-tranche pricing — and then use the higher headline number to recruit employees and raise angel capital — are potentially misrepresenting material facts. Foody's call-out makes clear this is now a reputational liability, not just a negotiating tactic: "founders misrepresent this to their employees & then shop it to angels too." Founders should demand transparency on entry valuations from all investors in a round before announcing.

AI Cybersecurity Must Shift to Continuous Offensive Testing

Anthropic's Mythos model generating a working Windows kernel exploit in 31 minutes signals that the reactive patch-and-pray security model is dead. The funding of A Security — which "continuously attacks customers' systems to find and help fix security weaknesses before hackers exploit them" — reflects a structural shift: the time-to-exploit window is now measured in minutes, not weeks, making continuous red-teaming an operational necessity rather than a luxury.

Build a Real Data Layer Before Feeding AI Your Portfolio Documents

The PortfolioIQ sponsor note articulates a broadly applicable principle for operators deploying AI on internal data: "feeding a chatbot raw documents is neither safe nor effective." The insight generalizes — any enterprise AI deployment needs a verified, standardized, structured data layer beneath it to be both safe and actionable.

6. Overlooked Insights

Moonshot AI's Valuation Trajectory Is One of the Most Aggressive in the Market

Moonshot AI's valuation went from "$4+ billion" in December 2025 to a $20B lead round (still closing) to now targeting $30B — all within roughly six months. This near-8x step-up in under a year for a three-year-old Beijing chatbot company, happening concurrently with multiple open rounds, deserves scrutiny as a potential signal of froth in Chinese AI private markets.

H-1B Fee Block Has Immediate Implications for AI Talent Cost Structures

A federal judge blocked Trump's proposed $100,000 H-1B visa fee, "ruling that the charge amounted to a tax that Congress hadn't authorized the executive branch to impose." This is a significant but briefly mentioned legal win: for AI companies heavily dependent on H-1B talent pipelines, this ruling — if it holds — removes what would have been a material cost and access barrier to international engineering talent.