PLUS: An AI-piloted F-16, big tech's AI cash burn, and a new second brain for agents
Happy reading
Payments giant Stripe is reportedly making a massive $10 billion move into AI infrastructure with the acquisition of model marketplace OpenRouter. The deal signals a significant push into the foundational 'plumbing' that powers AI development.
The potential acquisition highlights a growing focus on the systems that allow developers to access and switch between hundreds of different AI models. With such a high price tag, is the future of AI development less about a single dominant model and more about the marketplace that connects them all?
In today’s Next in AI:
Stripe’s massive $10B bet on AI models
An AI agent successfully pilots an F-16
Big tech’s colossal AI cash burn
A new ‘second brain’ for AI agents
Stripe's $10B AI Bet
Next in AI: Payments giant Stripe is in talks to acquire OpenRouter, an AI model marketplace, in a massive deal that could be valued at around $10 billion. This move signals a major push into the AI infrastructure space.
Explained:
OpenRouter’s platform helps developers easily access, compare, and switch between hundreds of different large language models from providers like OpenAI and Anthropic.
The potential $10 billion price tag represents a significant jump from the startup's recent $1.3 billion valuation, showing intense market interest.
The acquisition allows Stripe's customers to control spending and diversify away from relying on a single AI model provider, addressing a key need for businesses building with AI.
Why It Matters: This deal highlights that the AI infrastructure layer—the "plumbing" that lets developers choose the best model for any task—is becoming as critical as the models themselves. It reinforces the idea that the future of AI is a multi-model ecosystem, not a winner-take-all market.
AI in the Cockpit
Next in AI: DARPA and the U.S. Air Force have successfully flown a modified F-16 fighter jet controlled by an AI agent. This milestone marks a major step in converting standard operational aircraft into autonomous platforms for advanced combat scenarios.
Explained:
A specialized VENOM Autonomy Kit allows a human pilot to toggle control with the AI, creating a safe environment for live testing without altering the jet's core software.
Unlike one-off experimental jets, this test used a standard operational F-16, proving the technology can be scaled across the existing fleet.
These VENOM F-16s are now the cornerstone of the AIR program, designed to test multiple AI agents under the most realistic flight conditions to prepare for future combat.
Why It Matters: This success shifts autonomous military aviation from bespoke experiments to a scalable reality for the existing Air Force fleet. It lays the groundwork for a future where human pilots act as mission commanders, orchestrating teams of autonomous aircraft in complex engagements.
The AI Spending Spree
Next in AI: Tech giants Alphabet and Tesla are betting big on AI with massive spending increases, testing Wall Street's patience as colossal investments push their cash flow into the red.
Explained:
Alphabet raised its 2026 capital expenditure forecast to a staggering $195-$205 billion, while Tesla plans to spend more than $25 billion this year on its own AI and robotics initiatives.
The spending is fueling a historic buildout of AI infrastructure, from Google’s data centers packed with advanced chips to Tesla’s factories for its Cybercab and Optimus robots.
This aggressive investment pushed both companies into negative free cash flow, with Alphabet reporting a cash burn of $5.9 billion—its first since going public in 2004.
Why It Matters: This moment highlights the high-stakes race for AI dominance, where spending hundreds of billions is now the price of admission to stay competitive. For professionals, this signals an era of accelerated infrastructure growth and opportunity, even as the market questions the immediate return on these foundational investments.
Your AI's Second Brain
Next in AI: A new YC startup, Screenpipe, has launched a tool that locally records your screen and audio, creating a searchable memory that AI agents can use to learn how you work and automate your tasks.
Explained:
To avoid bogging down your computer, Screenpipe intelligently captures events like clicks and app switches using the OS accessibility tree, only falling back to OCR when needed.
All your activity is stored locally in a searchable database, and the tool includes privacy features like on-device PII redaction to keep your data secure.
A local API allows AI agents to securely access the recorded memory, enabling them to help with tasks like summarizing your workday or generating SOPs.
Why It Matters: Screenpipe represents a significant step toward creating truly personalized AI assistants that learn from your unique workflows, not just generic data. By giving agents a continuous, local memory of your actions, this approach could unlock a new level of proactive automation and contextual understanding.
AI Pulse
OpenAI launched the GPT-5.6 family of models, with its flagship Sol model setting a new state-of-the-art in coding and knowledge work benchmarks while claiming greater efficiency than rivals.
OpenAI introduced GPT-Live, a new generation of full-duplex voice models that can listen and speak simultaneously to power a more natural conversational experience in ChatGPT.
Oracle slashed its workforce by 21,000 employees, or 13%, to fund a massive $300B AI computing deal with OpenAI that is now reportedly in jeopardy due to a credit downgrade.
AMD partnered with Cerebras to create an AI inference solution combining AMD's Helios systems with Cerebras's wafer-scale chips to target ultra-low latency workloads.
Google released its first ATLAS report, a large-scale study on AI adoption which found that 68% of occupations are using AI, but typically for only about 21% of their tasks.