...

Artificial Intelligence Weekly

AI in the Markets

Amazon just dropped $50 billion on AI infrastructure for government customers, and Elon's xAI is about to close a $15 billion round at a $230 billion valuation. Meanwhile, the guy who called the 2008 crash—Michael Burry—is now shorting Nvidia and saying the real demand is "ridiculously small". AI spending is set to blast past $1.5 trillion in 2025, making it way bigger than the dot-com bubble, and it's all happening twice as fast. Even OpenAI's Sam Altman is out here admitting investors are "overexcited about AI." So yeah, it's getting weird.

Here's the thing that's freaking people out: Nvidia's making a $100 billion deal where they're basically funding OpenAI to buy... Nvidia's own chips. It's like your dealer giving you money to buy their product. But the bulls aren't backing down—JP Morgan says "there was a bubble, it burst. We don't think we're in one now for AI", and OpenAI's revenue jumped to $13 billion while Anthropic went from $87 million to $7 billion in a year—an 80x increase. Those are real numbers, not vaporware.

But underneath all the money flying around, companies are quietly cutting thousands of jobs. Tech layoffs in 2025 have already hit 182,963 people across 626 companies, with many literally saying they're replacing humans with AI. Salesforce axed 4,000 customer support roles in September, flat-out saying AI can do 50% of the work, and Klarna's CEO said they shrunk headcount by 40%, partly because of AI. So we've got this bizarre situation where billions are flooding into AI companies while the same technology is being used to justify mass layoffs. The full picture suggests we're either witnessing the birth of the most transformative technology ever, or setting up for a spectacular crash. Maybe both.

Artificial Intelligence Weekly

Tech Marvell to acquire Celestial AI for as much as $5.5 billion

The purchase price could increase to $5.5 billion if Celestial hits revenue milestones, Marvell said. The deal is an aggressive move for Marvell to acquire complimentary technology to its semiconductor networking business.

cnbc.com

Bubble Trouble: AI rally shows cracks as investors question risks

Soaring valuations in AI stocks this year have stoked worries that Wall Street may be inflating another speculative bubble, with some of these fears coming to the fore after a stellar earnings report from AI bellwether Nvidia failed to rally the stock and broader market.

reuters.com

6 AI stocks we’re keeping our eye on in December 2025

A few large companies, such as Palantir Technologies, have hogged much of the AI spotlight recently due to their massive spikes in value, and their valuations and performance have remained high even in the face of tariffs and broad market volatility. They continue to be represented in our list of the best AI stocks below, but there are many smaller, more obscure AI-related companies that have had substantial returns over the last year, too.

nerdwallet.com

The Power Crisis Meets Light-Speed Hardware

The US government officially lit the fuse on massive AI demand by launching the Genesis Mission [1]. This historic, national effort aims to use AI across 17 labs to double American scientific productivity within a decade, instantly cranking up the pressure on energy grids. The hardware world’s answer? A foundational deep-tech miracle: researchers demonstrated direct tensor processing with coherent light [2]. This means complex deep learning calculations can be performed passively using light instead of electricity, promising efficiency gains of up to 100 times over current electronic chips. Forget bigger data centers—the future of compute is happening at the speed of light.

OpenAI’s "Code Red" for the Efficiency Crown

This intense focus on efficiency is now defining the corporate battle for model supremacy. Facing serious competition from rivals, OpenAI reportedly declared an internal "code red" and is fast-tracking a secret project known as "Garlic [3].” The goal is not raw size but architectural optimization: squeezing GPT-4.5-level reasoning and coding ability into a much smaller, faster, and cheaper model to reclaim the top spot against Gemini 3 and Opus 4.5. This efficiency breakthrough is a game-changer, proving that performance-per-dollar is the new metric of success.

The Science Backs the Architectural Shift

The validity of this architectural focus was underscored by a critical research finding presented at NeurIPS 2025. The research delivered a critical "negative finding," confirming that the popular post-training technique of Reinforcement Learning with Verifiable Rewards (RLVR) only improves a model’s sampling efficiency. Crucially, they do not expand the model’s fundamental reasoning capacity.[4] This proves the industry consensus: genuine intelligence and advanced capabilities (like in 'Garlic') have to be engineered directly into the foundational architecture, reinforcing the strategic necessity of focusing on pre-training and efficiency.

Artificial Intelligence Weekly

AI Startup Mania: 25-Year-Olds Become Billionaires While Bezos Plots His Comeback

The AI coding wars just went nuclear. Cursor raised $2.3 billion at a $29.3 billion valuation in November—tripling its value in just five months. The company's 25-year-old founder Michael Truell turned down acquisition offers from OpenAI (reportedly in the billions) to keep building. Get this: Cursor hit $100 million in revenue in its first 12 months, the fastest any SaaS company has ever reached that milestone, then blasted past $1 billion in annual revenue. The coding assistant now processes over a billion lines of accepted code daily and companies like Uber, Spotify, and even OpenAI itself are using it. Accel and Coatue led the round, with Nvidia and Google jumping in to lock down strategic partnerships. Meanwhile, Jeff Bezos came out of retirement to co-CEO a mystery AI company called Project Prometheus that's building AI for manufacturing with $6.2 billion already in the bank.

The funding bonanza isn't slowing down. Black Forest Labs grabbed $300 million at a $3.25 billion valuation for its image generation tech, while AI banking automation startup Model ML landed $75 million in Series A to build agents that auto-generate pitch decks and financial models. Over in prediction markets, Kalshi closed a $1 billion Series E led by Sequoia to expand its platform that now trades over $1 billion weekly. The enterprise AI space is exploding too—Eon raised $300 million at a $4 billion valuation for its cross-cloud data backup platform, while wealth management startup Nevis secured $40 million from Sequoia to automate the busy work that financial advisors hate. November alone saw a stunning 43% of all venture funding go to just 14 mega-rounds over $500 million each—the highest concentration in three years.

But here's where it gets messy: Suno just raised $250 million at a $2.45 billion valuation despite getting sued for half a billion dollars by Sony, Universal, and Warner. The music generation startup is pulling in $200 million in annual revenue and users are cranking out 7 million songs per day—every two weeks Suno's output equals Spotify's entire catalog. Investors quintupled its valuation from $500 million in May, copyright lawsuits be damned. Meanwhile, voice AI company Gradium raised $70 million to build real-time voice agents that actually sound human. The whole AI startup ecosystem is basically running on FOMO at this point—investors are terrified of missing the next OpenAI or Anthropic, so they're throwing billions at anything that looks remotely promising. November's total venture funding hit $39.6 billion, up 28% year-over-year, with AI startups capturing the lion's share. Whether these valuations hold up when the music stops is anyone's guess.

Artificial Intelligence Weekly