Weekly Download #45

OpenAI Under Siege + The Data Center Arms Race + AI Rewrites the Rules of Cybersecurity

Author

Dev Chandra
April 13, 2026 • Estimated Reading Time: 8 minutes

In partnership with

Hi !

This week in Silicon Valley, startups & tech:

  • Three OpenAI Stargate leaders are leaving for Meta. Stargate UK is paused. Internal forecasts show OpenAI betting on $102B in ad revenue by 2030.

  • Meta committed an additional $21B to CoreWeave, totaling $35B+. Amazon is preassembling modular data centers under "Project Houdini."

  • Bessent and Powell warned bank CEOs about Anthropic's Mythos. Goldman, Citi, and others are testing the model internally.

  • TSMC posted Q1 revenue up 35% to $35.6B, beating estimates. War didn't dent demand for AI chips. Samsung is investing $4B in Vietnam.

  • SiFive raised $400M at a $3.65B valuation, pre-IPO. Alibaba led $293M into ShengShu. Hermeus raised $350M for unmanned hypersonic fighters.

Startup Intros has a simple mission: to help early-stage founders stay informed and navigate fundraising from compliance to capital.

🔥 Startup Intros Events Coming Up

🔥 OpenAI Under Siege

🏗️ The Stargate Exodus

Three senior executives who helped launch OpenAI's Stargate data center initiative are leaving for Meta. The departures follow The Information's report that OpenAI's data center strategy is undergoing yet another shakeup. Separately, OpenAI paused Stargate UK, citing energy costs and regulatory challenges; the UK initiative was announced alongside Nvidia and Nscale in September 2025.

The Timing: Meta just committed an additional $21B to CoreWeave for 2027-2032, is pulling top engineers into its Applied AI Engineering division, and shipped Muse Spark last week. OpenAI's infrastructure talent is walking into a company that's spending aggressively to catch up.

OpenAI called Musk's recent lawsuit amendments a "legal ambush" in a court filing, calling them "legally improper and factually unsupported." This followed Musk's move last week to redirect damages to OpenAI's nonprofit and push Altman off the board. The trial is set for later this month with $100B+ at stake.

The Policy: OpenAI is backing an Illinois bill that would shield AI labs from liability even for "critical harms" like 100+ deaths or $1B+ in damage, as long as they published safety reports. Florida's AG launched a probe into OpenAI and ChatGPT, citing data risks and a mass shooter's use of the product. And the White House is pushing back on GOP-led AI bills in Nebraska and Tennessee.

The Police: SFPD arrested a suspect for throwing a Molotov cocktail at Altman's home and making threats outside OpenAI's headquarters. Days later, a gun was fired at Altman's home from a car in a second attack; police arrested two suspects. Altman posted a blog reflecting on the attacks, personal regrets, and advocating for de-escalating rhetoric in the AI industry.

💰 The $102B Ad Bet

Beneath the chaos, internal documents show that OpenAI expects ads to generate ~$2.4B in 2026 revenue, quadruple to ~$11B in 2027, and reach ~$102B by 2030, or 36% of its total revenue. That last number would make OpenAI one of the largest ad businesses on Earth, rivaling Meta's current ad revenue.

New Features: OpenAI also launched a ChatGPT Pro tier at $100/month, aimed at Codex power users, offering 5x the usage of Plus. And Tubi became the first streaming service to launch a native app within ChatGPT, allowing viewers to discover 300K+ titles.

Translation: Three Stargate leaders left for Meta, while there is an internal ad revenue model that depends on reaching $102B by 2030, a number rivaling Meta. When your infrastructure team is defecting, and your financial model depends on an ad business that doesn't exist yet, how risky is your plan?

The IT strategy every team needs for 2026

2026 will redefine IT as a strategic driver of global growth. Automation, AI-driven support, unified platforms, and zero-trust security are becoming standard, especially for distributed teams. This toolkit helps IT and HR leaders assess readiness, define goals, and build a scalable, audit-ready IT strategy for the year ahead. Learn what’s changing and how to prepare.

🏗️ The Data Center Arms Race

💸 The CoreWeave Flywheel

Meta committed an additional $21B to CoreWeave for AI cloud infrastructure running from 2027 to 2032, on top of its prior $14.2B deal that extends through 2031. CoreWeave now has $35B+ in committed Meta revenue alone. Separately, CoreWeave signed a multiyear deal with Anthropic covering a variety of Nvidia chips across its now 43 active data centers.

The Demand: Blackstone filed for an IPO of a new data center acquisition vehicle to buy already-built and leased properties, planning to raise ~$2B. Amazon is working on "Project Houdini", which preassembles core server rooms into modules to cut construction time. The race is no longer about securing GPUs; it's about who can build the physical infrastructure fastest.

🔧 The Chip Supply Chain

The Infrastructure Buildout:

Translation: If you're building an AI startup, the good news is the compute is coming. The bad news is that three or four companies are deciding where it goes. Does that concentration help you or hurt you?

🔒 AI Rewrites the Rules of Cybersecurity

🏦 The Government and Banking Response

Bloomberg reports that Treasury Secretary Scott Bessent and Fed Chair Jerome Powell warned bank CEOs about the risks from Anthropic's Mythos at an urgent meeting on April 7. Goldman Sachs, Citigroup, and other banks are now testing the model internally; JPMorgan Chase is the only bank formally named in Project Glasswing.

A week earlier, CNBC reported that JD Vance and Bessent had questioned Amodei, Altman, and other AI leaders about model security and responding to cyberattacks. The White House's National Cyber Director Sean Cairncross is now leading an effort to identify security vulnerabilities in critical infrastructure that AI could exploit.

OpenAI is finalizing a product with advanced cybersecurity capabilities that it plans to release only to a small set of partners, similar to Anthropic's approach with Mythos. The cybersecurity AI arms race has officially begun.

🌐 The Threat Landscape

The Urgency:

The Defense? The Trump admin proposed cutting CISA's budget by $707M+, while Anthropic committed $100M to fill the gap through Project Glasswing. Researchers discovered that attackers have been exploiting an Adobe Reader zero-day since December 2025. And France announced it plans to move government computers from Windows to Linux to reduce reliance on US technology.

Translation: The Treasury Secretary and Fed Chair don't hold urgent meetings with bank CEOs over a press release. Mythos forced the entire financial system to reckon with what happens when development costs for exploits drop from weeks of human labor to $2,000 in API calls. When private companies are running the national cybersecurity response, who's accountable when the next breach hits?

📊 AI Adoption Hits an Inflection Point

🎤 The HumanX Signal

At HumanX in SF, where 6,500 executives, founders, and investors gathered, Anthropic dominated the conversation over OpenAI for the first time. Claude Code was "the tool on everyone's lips," generating $2.5B+ in annualized revenue as of February.

Synthesia CEO Victor Riparbelli captured the sentiment: "The guys at Anthropic were just like, 'We're not going to do anything about video, we're not going to care about voice models, we're just going to solve code gen,' and now we're here." Further, Ramp data shows 30.6% of US businesses paid for Anthropic tools in March, up from 24.4% in February; OpenAI remained nearly flat MoM at ~35%

The Other Major Theme: Chinese open-weight models. Executives noted that GLM-5.1, Kimi K2.5, and Qwen3.5 now dominate industry benchmarks, raising questions about whether closed-model companies can maintain their lead.

📈 The Numbers

Stanford HAI's 2026 AI Index Report found that generative AI reached 53% population adoption within three years, faster than the PC or the internet. The estimated value of GenAI to US consumers hit $172B annually by early 2026, with median value per user tripling between 2025 and 2026. The US leads in capital, infrastructure, and chips; China leads in patents, publications, and autonomous robotics. But the Foundation Model Transparency Index dropped from 58 to 40, as the most capable models became the least transparent.

Gallup's survey of 23,717 employed US adults found 50% now use AI at work, crossing the majority threshold for the first time. 30% are frequent users (daily or weekly). Leaders adopt faster: 67% use AI daily or weekly, compared with 46% of individual contributors. Among non-users with AI tools available, 46% say they simply prefer to do work the way they already do.

Translation: Half the workforce is using AI. Adoption is outpacing the PC and the Internet. The value per user tripled in a year. But the transparency of the models powering all this adoption is declining, not improving.

⚡ Startup Quick Hits

  • SiFive: $400M Series G at $3.65B for the RISC-V chip designer led by Atreides; CEO says it's the final round before an IPO

  • ShengShu: ~$293M led by Alibaba for the creator of Vidu AI video generator, two months after an ~$88M round; Baidu and Luminous also invested

  • KreditBee: $280M Series E at $1.5B for the Bengaluru-based digital lending platform, entering the unicorn club

  • Hermeus: $200M led by Khosla Ventures plus $150M in debt at a $1B valuation for unmanned hypersonic fighter aircraft

  • Aria Networks: $125M Series A from Sutter Hill and others for the AI-native network that works with any AI chip

  • Chapter: $100M Series E at $3B for the AI-powered Medicare enrollment platform, doubling its valuation; total funding now $285M

  • Modus: $5M seed plus $80M Series A led by Lightspeed for AI agents that automate audit workflows and invest in accounting firms

  • MillTech: $60M at $325M for the UK-based FX hedging platform; plans to expand in North America and build AI tools

  • Elorian: $55M at $300M for visual AI models with better reasoning for robotics; founded by ex-Google DeepMind researchers

  • Patlytics: $40M Series B led by SignalFire for AI-powered patent filing and litigation automation

  • True Footage: $40M Series C led by Cox Enterprises' Socium Ventures for AI-generated property valuations trained on years of appraisal data

  • AfterQuery: $30M Series A at $300M for coding and finance training data; founded by a 23-year-old and already at a $100M+ annual run rate

💰 Investor Quick Hits

  • SpaceX financials: SpaceX lost just under $5B in 2025 while generating $18.5B+ in revenue; the figures include xAI, which SpaceX acquired in February

  • Neo VC: The VC firm that made early bets on Kalshi and Cursor pushed its first two funds far above industry averages

  • ARR is the least trusted metric: Silicon Valley investors say annual recurring revenue has no SEC definition and can be massaged

💸 IPO & M&A Quick Hits

🌟 Editor's Note

At Startup Intros, our mission is to bring the latest founder-investor news straight to your inbox, keeping you ahead in the fast-paced world of Silicon Valley.

💭 Parting Thoughts

Three OpenAI Stargate leaders left for Meta. The Treasury Secretary and Fed Chair held an urgent meeting with bank CEOs over an AI model. Meta committed $35B+ to a single GPU cloud provider. TSMC beat estimates during a war. The FBI says cybercrime hit $21B. France is ditching Windows. OpenAI's internal revenue model depends on a $102B ad business that doesn't exist yet.

This was the week every institution, government, banks, regulators, and Big Tech, started reacting to AI simultaneously. Not debating it. Not studying it. Reacting. So you should build accordingly.

Till next time!

Dev Chandra
Founder & CEO @ Startup Intros
EiR @ Context VC
LinkedIn: /in/devchandra

Startup Intros Weekly Download: Your trusted source for founder-investor insights, delivered with clarity and focus.