Google Doesn't Have to Sell Its Ad Tech Business
1. Key Themes
AI Mega-Rounds Continue at Unprecedented Valuations
Capital is flooding into AI infrastructure and agent companies at eye-watering valuations, with multiple nine- and ten-figure rounds closing in a single week.
- "Cognition, a three-year-old San Francisco startup that develops Devin, an AI engineering agent that autonomously handles software development tasks, is reportedly set to raise around $1 billion at a $47 billion post-money valuation."
- "Wonderful, a one-year-old Amsterdam startup that helps large enterprises deploy AI agents across business workflows, raised a $550 million Series C at a $5 billion post-money valuation."
- "Tripo AI... raised a $446 million Series B round."
AI Chain-of-Thought Monitorability Is Becoming a Flashpoint for Safety
As reasoning models get more powerful, techniques that obscure their "thinking" are alarming safety researchers who see interpretability as a critical safeguard.
- "This technique, also called 'opaque recurrence,' will likely make the model's chain of thought more difficult to monitor — and that has AI safety experts rattled."
- "'I don't know whether Astra is much less CoT monitorable than previous models. But if OpenAI pushes this technique further, they'll have the option to massively increase the recurrence and totally destroys CoT monitorability,'" wrote Redwood CEO Buck Shlegeris.
- "'The technique is playing with fire, risking a taboo that OpenAI and Anthropic have fought to establish that we work hard to maintain Chain of Thought faithfulness and monitorability for as long as we can,'" Mowshowitz wrote.
Regulatory and Legal Battles Are Reshaping Big Tech's Risk Landscape
Government actions — courts, the administration, and litigation — are becoming decisive forces in how AI and ad-tech giants operate, with mixed outcomes for incumbents.
- "Google avoided a breakup of its ad-tech business after a federal judge rejected the Justice Department's demand that it sell key assets, instead ordering changes to its business practices whose details remain under seal."
- "The Trump administration has inserted itself into The New York Times' copyright lawsuit against OpenAI, urging a federal court to find that training AI models on copyrighted material generally qualifies as fair use."
- "OpenAI faces 30 additional lawsuits over the Tumbler Ridge school shooting, with plaintiffs for the first time accusing the company of aiding and abetting the attack."
Autonomous/Physical AI Infrastructure Is Attracting Serious Capital
Beyond software agents, investors are backing the physical and infrastructure layer of AI and robotics — from photonics to robot perception to defense drones.
- "iPronics... develops silicon-photonics optical switches for connecting GPUs and other computing infrastructure in AI data centers, raised a $125 million Series B."
- "Lyte... develops custom silicon, sensors, and spatial software to help robots perceive and navigate their surroundings, raised a $165 million Series C at a $1.6 billion post-money valuation."
- "Nscale, a UK-based Nvidia-backed AI cloud infrastructure provider, is telling prospective IPO investors it has about $103 billion in contracted revenue, with contracts averaging 5.7 years in duration."
2. Contrarian Perspectives
"Driverless" cars still depend heavily on unglamorous human labor. The Waymo narrative of full autonomy obscures a large, grim backend workforce — suggesting the economics and scalability of robotaxis are less "autonomous" than marketed.
- "An anonymous Waymo depot contractor describes the decidedly human labor behind the company's 'driverless' fleet: workers charge cars, clean sensors, reboot malfunctioning vehicles, offload trip data, untangle robotaxi traffic jams, and clean up vomit, feces, discarded joints, and even the occasional pig head."
Courts may be more protective of Big Tech than regulators expect. Despite aggressive DOJ antitrust action, the judiciary chose incremental behavioral remedies over structural breakup — a signal that antitrust enforcement against tech giants may be less existential than headlines suggest.
- "Google avoided a breakup of its ad-tech business after a federal judge rejected the Justice Department's demand that it sell key assets, instead ordering changes to its business practices whose details remain under seal."
3. Companies Identified
- Google — Search/ad-tech giant. Mentioned as the subject of a major antitrust ruling. "Google avoided a breakup of its ad-tech business after a federal judge rejected the Justice Department's demand that it sell key assets."
- OpenAI — AI lab behind ChatGPT/Astra. Mentioned for its controversial new reasoning technique and mounting legal exposure. "OpenAI's new Astra model will use a reasoning technique called 'recurrent depth'..."
- Cognition — AI coding agent (Devin) startup. Mentioned for a massive pending mega-round. "...is reportedly set to raise around $1 billion at a $47 billion post-money valuation."
- Wonderful — Enterprise AI agent deployment startup. Mentioned for a huge Series C plus secondary sale. "raised a $550 million Series C at a $5 billion post-money valuation... Investors also bought $170 million of shares from employees and early angel investors in a secondary transaction."
- Tripo AI — AI 3D asset generation startup. Mentioned for a huge, heavily Chinese-backed Series B. "raised a $446 million Series B round led by MPCi."
- Nscale — Nvidia-backed AI cloud infrastructure provider. Mentioned for extraordinary IPO-pitch metrics. "has about $103 billion in contracted revenue, with contracts averaging 5.7 years in duration."
- Uber — Ridesharing/delivery giant. Mentioned for major layoffs and strategic refocus. "is laying off about 3,300 employees or 10% of its workforce as CEO Dara Khosrowshahi cuts management layers... redirects spending toward ridesharing, delivery, and robotaxis."
- Grindr — Dating app. Mentioned as a case study in diversification and growth under CEO George Arison. "the company has grown 25% every year he's been in charge, dating back to 2022."
- Waymo — Autonomous vehicle company. Mentioned as a case study in the hidden human labor behind "driverless" tech. "workers charge cars, clean sensors, reboot malfunctioning vehicles..."
- Anthropic — AI lab (Claude). Mentioned regarding its standoff and reconciliation with the Trump administration over military use restrictions. "Anthropic is 'back on the right side' with the Trump administration, days after a federal judge blocked the Pentagon from blacklisting the AI company."
- idscan.net — Identity-verification provider. Mentioned as the apparent source of a massive data breach. "A dark-web identity-theft service is selling scans of more than 153 million U.S. and Canadian driver's licenses, apparently sourced from identity-verification provider idscan.net."
- Palo Alto Networks — Cybersecurity company. Mentioned for a larger-than-disclosed acquisition. "Palo Alto Networks may have paid as much as $500 million in cash and stock to acquire the company [Console]."
- Adobe — Mentioned for an acqui-hire of Rilo, a small Indian marketing-intelligence startup. "bringing its six-person team and some of its technology into Adobe."
4. People Identified
- George Arison — CEO of Grindr. Featured in the newsletter's podcast for transforming Grindr into a diversified platform. "his overarching mission is to transform the gay-dating app into a one-stop shop, including healthcare and travel concierge services."
- Buck Shlegeris — CEO of Redwood Research, AI safety organization. Cited for raising alarm over OpenAI's new reasoning technique. "I am extremely concerned by the reporting that Astra uses opaque recurrence."
- Zvi Mowshowitz — Longtime AI safety advocate/writer. Cited for warning about a "race to the bottom" among AI labs. "laws might be necessary to prevent a 'race to the bottom' among AI labs."
- Howard Lutnick — Commerce Secretary. Mentioned for commenting on Anthropic's political standing. "Anthropic is 'back on the right side' with the Trump administration."
- Dara Khosrowshahi — CEO of Uber. Mentioned for driving major layoffs and strategic reorganization. "cuts management layers, consolidates teams, and redirects spending toward ridesharing, delivery, and robotaxis."
- Charlie Dolman — Burning Man "fixer." Mentioned in a colorful profile as the event's problem-solver. "If you do meet me in Black Rock City, you're either in jail, or dead, or in some other ungodly mess..."
- Rahul Meka & Daniel Yum — Newly promoted partners at Playground Global. Mentioned for career advancement in deep-tech VC. "Playground Global promoted Rahul Meka and Daniel Yum to partner, elevating two principals on the Palo Alto deep-tech venture firm's investment team."
5. Operating Insights
- Diversify revenue lines around a core community/product to sustain growth: Grindr's expansion into healthcare and travel concierge services alongside its core product drove consistent compounding growth — "the company has grown 25% every year he's been in charge, dating back to 2022" — a tactic for founders looking to extend product lifecycle and lifetime value beyond a single use case.
- Expect "autonomous" claims to require significant hidden operational labor: Founders building autonomous/robotic products should budget for real-world operational overhead (maintenance, cleanup, incident response) that doesn't show up in the marketing narrative, as illustrated by Waymo's depot labor requirements.
- Secondary sales are becoming a standard feature of late-stage mega-rounds: Wonderful's Series C included a $170M secondary component for employees and early investors, suggesting operators should structure large raises to include liquidity for early stakeholders as a talent-retention and cap-table-hygiene tool: "Investors also bought $170 million of shares from employees and early angel investors in a secondary transaction."
6. Overlooked Insights
- Contracted revenue models are being used aggressively in AI infrastructure IPO pitches: Nscale's disclosure of "$103 billion in contracted revenue, with contracts averaging 5.7 years in duration" — implying ~$18B annualized — signals a potential new pattern of AI infra companies leaning on long-duration contract backlogs to justify valuations, which investors should scrutinize for revenue-recognition and counterparty-risk assumptions.
- Identity-verification infrastructure is a systemic single point of failure: The breach traced to "identity-verification provider idscan.net, whose customers include Hertz, Target, FedEx, and Caesars" reveals concentrated risk in third-party KYC/identity vendors used across major consumer brands — a quietly significant supply-chain security exposure.