OpenAI Loses Top Exec Fidji Simo
- 01The $3 Trillion Revenue Gap in AI Infrastructure
- 02AI Dominates Venture Capital at Historic Scale
- 03AI Governance Vacuum Creates Systemic Risk
- 04Defense Tech and Maritime Autonomy Are Active Investment Categories
- 05OpenAI's Organizational Fragility Is Growing Pre-IPO
1. Key Themes
The $3 Trillion Revenue Gap in AI Infrastructure
The central tension in AI investing is whether the industry can generate enough revenue to justify its infrastructure bets. Sequoia's David Cahn calculates that, after three years of hyperscaling, the AI industry will need to earn $3 trillion to justify all chips and data center expenditures — and that number is likely an underestimate. As the article notes: "The rising costs of memory and the increasing use of exotic or inference-specific chips will drive that number up. 'Recently,' he writes, 'the required revenue per GW of CapEx has sharply increased due to these bottleneck dynamics and rising costs of construction.'"
AI Dominates Venture Capital at Historic Scale
AI is no longer one category among many — it is venture capital. "U.S. venture funding hit $412.7 billion in the first half, already nearly 30% above all of 2025, but PitchBook says the boom is overwhelmingly an AI story, with AI companies capturing 86% of dollars and a handful of giant rounds and SpaceX deals driving the market." The coming IPOs of SpaceX, Anthropic, and OpenAI could individually eclipse all historical VC-backed exits: "SpaceX's IPO and possible public offerings from Anthropic and OpenAI could create more value than all U.S. VC-backed exits since 2000 combined."
AI Governance Vacuum Creates Systemic Risk
There is no credible regulatory framework for frontier AI models, creating political and market risk. "As OpenAI rolls out Sol, experts say the Trump administration still has no clear process for deciding which frontier AI models are safe to release, raising concerns that an ad hoc system could favor companies with the best political access." Microsoft's Brad Smith adds institutional weight to the concern: "The Trump administration's AI policy has become 'regulation without transparent or complete rules,' arguing that recent restrictions on Anthropic's Fable 5 and OpenAI's GPT-5.6 show Washington needs clearer tools for reviewing frontier models without undermining trust in U.S. tech supply."
Defense Tech and Maritime Autonomy Are Active Investment Categories
Non-AI hard tech is attracting significant capital. Kraken Technology, which "develops uncrewed surface vessels and payload systems for maritime defense missions by NATO and allied military customers, raised a $175 million Series B round at a $1+ billion valuation" — backed by NATO Innovation Fund and Rheinmetall, signaling institutional defense capital flowing into autonomous maritime systems.
OpenAI's Organizational Fragility Is Growing Pre-IPO
OpenAI is losing senior leadership at a critical moment. "OpenAI's top product-and-business executive, Fidji Simo, is stepping down from her full-time role after an extended medical leave, leaving Sam Altman to fill a key position as the company prepares to go public and tries to catch Anthropic with enterprise customers." This is compounded by legal exposure: "The New York Times and The Daily News claim OpenAI hid evidence in their copyright case by saying it could not search training data or ChatGPT logs, even though an engineer allegedly testified that OpenAI had already searched for copyrighted journalism and built tools to track regurgitated outputs."
2. Contrarian Perspectives
Anthropic Is Beating OpenAI on Revenue — by a Wide Margin
The popular narrative positions OpenAI as the dominant commercial AI player, but the revenue data tells a different story. "Anthropic is thought to have hit $60 billion in ARR, while OpenAI reportedly earned $13 billion in 2025 (although in November 2025, it said it was at $20 billion ARR)." If accurate, Anthropic's ARR is 3–4x that of OpenAI — a striking reversal of mindshare expectations.
Nvidia May Be Hitting a Ceiling It Helped Create
The prevailing narrative treats Nvidia as the unchallengeable winner of the AI infrastructure era, but the article identifies a structural shift underway: "Nvidia is becoming a victim of the compute market it created: its GPUs made AI infrastructure the hottest trade in tech, but as compute supply broadens and GPU prices fall, investors are shifting toward the less glamorous memory-chip companies now sitting at the data-center bottleneck." The new bottleneck — and therefore the new investment opportunity — is memory, not compute.
IPOing at Peak Hype Is a Trap, Not a Prize
While many AI and defense startups race toward the public markets, Anduril's CEO is explicitly pumping the brakes: "Anduril CEO Brian Schimpf says the defense-tech company is in no rush to go public, warning that IPOing 'in the middle of a hype cycle' can backfire as defense and AI startups draw 'crazy high valuations' on expectations of future growth." This patience-as-strategy stance runs counter to the prevailing founder and investor pressure to capture current market sentiment.
3. Companies Identified
OpenAI | AI lab | Central to multiple stories: losing a top exec (Fidji Simo) before its IPO, facing copyright litigation, and trailing Anthropic on enterprise ARR. "OpenAI's top product-and-business executive, Fidji Simo, is stepping down from her full-time role after an extended medical leave, leaving Sam Altman to fill a key position as the company prepares to go public."
Anthropic | AI lab | Positioned as the enterprise leader over OpenAI. "Anthropic is thought to have hit $60 billion in ARR."
Mercor | AI talent marketplace for domain expert training data | Seeking to double its valuation in months. "Mercor…is reportedly in early talks to raise funding at a $20 billion valuation…The company last raised a $350 million Series C in October at a $10 billion valuation."
SpaceX | Aerospace/defense | Its IPO is described as potentially historic in scale. "SpaceX's IPO and possible public offerings from Anthropic and OpenAI could create more value than all U.S. VC-backed exits since 2000 combined."
Gradium | Low-latency voice AI for conversational agents | Remarkable seed round for a 10-month-old company. "Gradium…raised a $100 million seed round backed by Nvidia, FirstMark Capital, Eurazeo, DST Global Partners, Eric Schmidt, and Xavier Niel."
Databento | Market data infrastructure for financial and AI customers | Series B signals AI-driven demand for financial data pipelines. "Databento…raised a $97 million Series B round led by NEA."
Kraken Technology | Uncrewed maritime defense vessels | Unicorn valuation in defense autonomy. "Raised a $175 million Series B round at a $1+ billion valuation…backed by NATO Innovation Fund [and] Rheinmetall."
Ollama | Local/hybrid AI model runtime for developers | Benchmark and Y Combinator participation signals infrastructure-layer conviction. "Helps developers run open-weight AI models on personal computers and access larger hosted models when local hardware is not enough, raised a $65 million Series B."
Lyzr | Enterprise AI agent builder | Notable for using its own AI agent to fundraise. "Says it is on track to raise $100 million at a roughly $500 million valuation after using one of its own AI agents to run investor outreach and help draft investment memos."
Oxylabs | Web data and proxy infrastructure for AI training | Large Warburg Pincus bet on AI data supply chain. "Raised a $130 million round at an approximate post-money valuation of $3.6 billion."
TJM | AI agents for pharmacy automation | Rapid capital raise into a high-friction healthcare workflow. "Raised a $75 million Series B…The company has raised a total of $100 million" in just two years.
Anduril | Defense technology | Deliberately delaying IPO. "CEO Brian Schimpf says the defense-tech company is in no rush to go public."
Wonder | Food tech (Grubhub, Blue Apron parent) | Founder doubling down personally. "Arranging hundreds of millions of dollars in new funding at a roughly $9 billion valuation, with founder and CEO Marc Lore potentially contributing up to $200 million."
FanDuel | Sports betting | Legal liability signal for the sector. "Being sued for allegedly using VIP perks to deepen gambling addiction."
Sonos | Consumer audio | Cautionary tale on post-crisis restructuring. "Cut several senior design and product executives…raising internal concerns about whether the speaker maker has weakened the teams needed to create its next generation of products."
4. People Identified
David Cahn | Partner, Sequoia Capital | Originated and updated the AI infrastructure revenue gap analysis. "He calculates that the AI industry will have to earn $3 trillion to justify all those chips and other data center expenditures."
Fidji Simo | Former CEO of Apps, OpenAI | Her departure creates leadership and IPO risk. "OpenAI's top product-and-business executive…is stepping down from her full-time role after an extended medical leave."
Torsten Slok | Chief Economist, Apollo | Tracking hyperscaler payback timelines. "He points out that the hyperscalers — Google, Meta, Microsoft, and Amazon — are all predicting massive accelerations in their free-cash flow in 2028."
Brian Schimpf | CEO, Anduril | Articulating a disciplined anti-hype IPO philosophy. "Warning that IPOing 'in the middle of a hype cycle' can backfire."
Brad Smith | President, Microsoft | Calling out the governance vacuum in frontier AI. "The Trump administration's AI policy has become 'regulation without transparent or complete rules.'"
Marc Andreessen | Co-founder, a16z | Gaining direct policy influence at the Federal Reserve. "Will co-lead a new Federal Reserve task force on productivity and jobs, giving the Trump-allied venture capitalist and AI booster a role in shaping how the central bank assesses AI's impact on growth, inflation, and the future of work."
Marc Lore | Founder/CEO, Wonder | Putting substantial personal capital into his company. "Founder and CEO Marc Lore potentially contributing up to $200 million."
Elon Musk | CEO, SpaceX/Tesla/X | SEC settlement highlights regulatory double standards. "A federal judge approved Elon Musk's $1.5 million SEC settlement over his delayed disclosure of his Twitter stake…the SEC said the delay 'ultimately saved him a whopping $150 million.'"
Eric Schmidt | Former Google CEO | Active backer of early-stage AI. Participating in Gradium's $100M seed round.
Xavier Niel | Founder, Iliad | Co-invested in Gradium alongside Schmidt and Nvidia.
5. Operating Insights
Use AI Agents to Run Your Own Fundraise
Lyzr offers a concrete proof-of-concept for AI-assisted go-to-market: the company "used one of its own AI agents to run investor outreach and help draft investment memos" — and is now reportedly on track to close a $100M round at a $500M valuation. For founders building in AI, dogfooding your product in the fundraise itself is a powerful signal to investors and a forcing function for product refinement.
The Hyperscaler Payback Clock Is Ticking to 2028 — Build Toward That Timeline
Enterprises that can plug into hyperscaler revenue and free-cash-flow acceleration have a structural tailwind. Torsten Slok's observation that "the hyperscalers — Google, Meta, Microsoft, and Amazon — are all predicting massive accelerations in their free-cash flow in 2028" means the window to build products that monetize their compute buildout is roughly 18–24 months. Products and partnerships that align with that payback cycle are more fundable today.
Don't Hollow Out Product and Design Functions During a Turnaround
Sonos is a cautionary tale: after an app crisis, the company "cut several senior design and product executives — including much of its UX research team…raising internal concerns about whether the speaker maker has weakened the teams needed to create its next generation of products." Cutting customer-facing capability to hit short-term cost targets often destroys the future moat.
6. Overlooked Insights
The Shift from Compute to Memory as the AI Bottleneck Is an Investable Thesis
The article briefly notes that memory-chip companies are replacing GPU makers as the critical data-center constraint. This is not just a stock-picking observation — it has implications for enterprise infrastructure strategy. Cahn's analysis reinforces this: "The rising costs of memory and the increasing use of exotic or inference-specific chips will drive that number up." Memory, interconnects, and inference-specific silicon are the next layer worth watching.
Frontier AI Model Releases Are Now a Political Access Game
The regulatory observation about OpenAI's Sol is buried but significant: there is currently "no clear process for deciding which frontier AI models are safe to release, raising concerns that an ad hoc system could favor companies with the best political access." For investors, this means regulatory risk for AI companies is no longer just about compliance — it's about proximity to power in Washington, which creates moat advantages for incumbents with political relationships and headwinds for challengers.