Layoff season
Even as core economic data suggests that AI is not having as big of an impact on the economy as previously predicted, it’s hard to ignore the headlines of Fortune 500 companies shedding tens of thousands of jobs.
The big picture:
Untangling the root cause here is messy: even if AI isn’t replacing those workers today, that doesn’t stop CEOs from doing layoffs in the hope that AI can retroactively replace them tomorrow (though we’re already seeing some cautionary tales here).
It’s also likely that many of these layoffs are the result of some combination of pandemic-era overhiring and an increasingly complex macroeconomic environment - though the political climate may make it challenging to say the latter part publicly.
But it is clear that we’re seeing companies attempt to adopt AI usage (if not outright AI automation) at a far faster speed than previous technology shifts - leaving less time for workers to retrain and labor markets to adjust, and potentially creating acute economic and social disruption.
Elsewhere in the FAANG free-for-all:
Amazon opens Project Rainier, an $11B AI data center on 1,200 acres in Indiana that trains and runs Anthropic’s AI models using 500K+ Amazon Trainium 2 chips.
GitHub plans to launch Agent HQ, a hub to let developers access third-party AI coding agents, including Claude and OpenAI’s Codex, in the coming months.
Adobe unveils Project Moonlight, an AI agent on its Firefly platform designed to act as a creative director for social media campaigns via text prompts.
And Grokipedia, a Grok-based encyclopedia, launches with 885K+ articles, drawing from Wikipedia content; Elon Musk called it “a massive improvement” over Wikipedia.
Elsewhere in AI anxiety:
Scale AI and CAIS’ Remote Labor Index finds the best AI performed less than 3% of tasks when measuring AI models’ ability to automate freelance work.
Employees are increasingly using AI to create ultra-realistic fake expense receipts, with AppZen and others attributing the rise to advances in image generation models.
AI startups are aiming to address loneliness among the elderly, with a study predicting the AI market for aging and elderly care will grow to over $43B in 2025.
An a16z-backed startup is selling thousands of ‘synthetic influencers’ to manipulate social media as a service.
An ex-OpenAI staffer of four years says it isn’t doing enough to protect users, especially those with mental health issues, and needs to offer more than words.
OpenAI is for profits now
OpenAI completed a major restructuring that has its nonprofit arm owning 26% (though still retaining control) of the for-profit company. The deal also restructured Microsoft’s ownership agreement, with the tech giant retaining 27% ownership, valued at $135 billion.
Between the lines:
The agreement reveals both companies hedging their AGI bets: Microsoft can now pursue AGI independently or with other partners, while OpenAI gains freedom to release open-weight models and shop around for compute infrastructure - though Microsoft does retain IP rights through 2032, and OpenAI has committed to $250B in Azure spend.
The political dimension proved crucial: OpenAI appears to have gotten regulatory approval for the deal by promising to remain in California and emphasizing how AI is currently “the economic heart” of the state’s economy.
Ultimately, the deal is preparing for a future where OpenAI operates as a standalone public company with multiple partners rather than as Microsoft’s captive AI lab - as reports suggest OpenAI is preparing for an IPO as early as late 2026, valuing the company at up to $1 trillion.
Elsewhere in OpenAI:
The Sora app adds new features, including “character cameos” to create AI deepfakes of pets and illustrations, as it begins selling usage credits (free users are limited to 30 daily generations).
OpenAI launches Aardvark, a GPT-5-powered autonomous cybersecurity research agent that can identify and help patch vulnerabilities, in private beta.
A new “company knowledge” ChatGPT update lets Business, Enterprise, and Education users search for data in connected apps like Slack and GitHub.
OpenAI is developing AI tools to generate music from text and audio prompts, including capabilities such as adding guitar accompaniment to vocal tracks.
And PayPal signs a deal with OpenAI to let PayPal buyers and sellers complete transactions through ChatGPT starting in 2026, using the Agentic Commerce Protocol.
Elsewhere in AI geopolitics:
OpenAI urges the US to commit to building 100GW of new energy capacity each year to stay ahead of China in AI, calling “electrons are the new oil.”
President Trump says he and Xi Jinping discussed Nvidia and other chipmakers’ access to China, but didn’t discuss approving sales of Blackwell chips.
The US AI data center boom is consuming capital, power, and people that the rest of the economy needs, crowding out Trump’s manufacturing revival plans.
Papers, patents, and tenders show China’s military is integrating DeepSeek and Qwen models in weapons like AI-powered drones and continues to use Nvidia chips.
And Saudi Arabia is pouring money into data centers and working with US and Chinese tech giants, landing its AI ambitions in the middle of a geopolitical tussle.
The five trillion dollar company
On the heels of its GTC in DC event, Nvidia has become the world’s first company to reach a $5 trillion market capitalization, cementing its position as the most valuable corporation in history.
Why it matters:
The concentration of such enormous market value in a single chipmaker has raised questions about market stability and monopolistic dynamics in the AI supply chain, particularly as Nvidia’s fortunes remain tied to sustained enterprise spending on AI systems whose long-term returns remain unproven.
Of course, Nvidia’s CEO pushed back against AI bubble concerns, announcing partnerships across industries and projecting half a trillion dollars in revenue from its latest chips - spanning humanoid robots (Foxconn), cybersecurity (CrowdStrike), defense/logistics (Palantir), and pharma (Eli Lilly), aiming to broaden its revenue base as competition intensifies.
And the conference’s patriotic messaging and Trump references reflect Nvidia’s delicate geopolitical balancing act: championing American manufacturing while pursuing “sovereign AI” deals in Europe and seeking White House help to restore access to the Chinese market that’s cost them billions in lost revenue.
Elsewhere in Nvidia:
Nvidia invested in 59 AI startups in 2025 so far, including 10 in the past two months, surpassing 55 investments in 2024 and 12 investments in 2022.
Jensen Huang says the US needs “finesse” to maintain its AI lead over China, and Nvidia is in an “awkward place” as President Trump meets Xi Jinping later this week.
The company unveiled BlueField-4 data processing unit, combining a 64-core Grace CPU and ConnectX-9 SuperNIC for 800 Gbps of network throughput, launching in 2026.
Nvidia’s Blackwell GPUs are now in full production in Arizona, after being solely manufactured in Taiwan previously.
And Nvidia announced NVQLink, a high-speed interconnect that is the “Rosetta Stone connecting quantum and classical supercomputers.”
Elsewhere in AI infrastructure:
Google, Meta, and Microsoft collectively spent nearly $80B on AI infrastructure in Q3, dividing the market on whether they can translate huge capex into income.
The AI buildout, surpassing railroads and the internet, moves Big Tech from being asset-light to asset-heavy, which has historically produced inferior returns.
Sam Altman says OpenAI has committed to spending about $1.4T on infrastructure so far, equating to roughly 30GW of data center capacity.
An analysis of 30+ US data center proposals in 14 states found that in most cases, local officials signed NDAs and worked with shell companies to hide details.
And Indonesia and other developing countries are becoming AI hubs amid the multitrillion-dollar spending boom, in part driven by an “AI decolonization” philosophy.
Things happen
OpenAI releases gpt-oss-safeguard, its open-weight reasoning models for safety classification tasks. Canva launches a foundational design model that generates editable layered designs. YouTube is using AI to upscale videos to “HD.” Grammarly rebrands as Superhuman. AI can code, but it can’t build software. UMG announces partnerships with Stability AI and Udio to create fully licensed AI-generated music. Qualcomm unveils two AI inference chips, the AI200 and AI250. The AI boom will spark a “wildfire” that wipes out many companines. I’m drowning in AI features I never asked for and I hate it. Using AI to negotiate a $195k hospital bill down to $33k. Anthropic expands Claude for Financial Services with Excel integration. IBM releases open-source Granite 4.0 Nano AI models. JPMC staff can officially use an LLM for year-end performance reviews. Andrew Cuomo uses AI MPREG Schoolhouse Rock to attack Mamdani. Character.ai limits under-18s to two hours of chat ahead of November ban. Cal State to become “AI-empowered” university via partnerships with Amazon and ChatGPT. China’s MiniMax releases MiniMax M2, an open-source model optimized for AI agents and coding. Extropic shares its first chip using probabilistic bits that it says can be 10,000x more energy efficient. It’s insulting to read AI-generated blog posts. The new calculus of AI-based coding.
Thanks for the good 😊