INFRASTRUCTURE

Nvidia clutches its crown

Google isn’t only making OpenAI nervous. 

The search giant is reportedly in talks with Meta on a deal worth billions to supply the social media firm with Google’s Tensor Processing Units, or TPUs, for its data centers in 2027, as well as rent chips from Google Cloud, The Information reported. 

The news follows the banner release of Google’s Gemini 3, its latest state-of-the-art model that outperforms both its previous models and competitors. Gemini 3 was entirely trained on Google’s TPUs, designed in partnership with Broadcom. A Google spokesperson told CNBC that the company has seen accelerated demand for both its TPUs and Nvidia’s chips. “We are committed to supporting both, as we have for years.”

Google’s progress in the chip space might be making Nvidia antsy. On Tuesday, the company said in an X post that, while it’s "delighted” by Google’s success, its chips are “a generation ahead of the industry — it’s the only platform that runs every AI model and does it everywhere computing is done.” 

As it is, no other firm truly rivals Nvidia. The firm is the most valuable company in the world by market cap, and holds as much as 95% of the market share for AI chips, leaving competitors like AMD and Intel in the dust.

But Google isn’t the only company seeking to increase its chip self-reliance: OpenAI has long been working on developing its own GPUs, partnering with Broadcom to do so. Amazon offers its own line of custom chips for training and inference, and is injecting its hardware into government operations alongside a $50 billion investment. 

While Nvidia’s position is safe in the short term, this rise has started “to challenge Nvidia's reign through their own respective vertical integration capabilities,” Yaz Palanichamy, senior advisory analyst at Info-Tech Research Group, told The Deep View.

Tech firms seeking independence from Nvidia’s chips may also not be the chip giant’s only cause to worry. As frontier models reach new heights, the demands of the industry may be shifting from training AI to actually using it. While Nvidia’s chips are the best in the business for training, Palanichamy noted, custom chips like Google’s could be seen as a “cost-efficient alternative when dealing with model inference tasks, which could reveal an (as of yet) untapped market segment for tech giants such as Google and Amazon to capitalize on.”

TOGETHER WITH RADIANT

Tired of writing meeting follow-ups? Us too.

You know the feeling — the meeting ends, and suddenly you’ve got homework. Notes to tidy, action items to assign, emails to draft. Radiant handles all that for you.

This free AI meeting assistant from Supernormal quietly captures your meetings (without a bot in sight) and drafts your follow-up emails for you. Summaries, next steps, and action items land in one neat draft you can open in Gmail, review, and send in seconds.

Meetings in, momentum out. We like that equation.

Download Radiant on your Mac for free.

POLICY

Trump’s ‘Genesis Mission’ highlights China's AI battle

The U.S. is doubling down on its push to stay ahead of China in AI.

On Monday, the White House announced the “Genesis Mission,” an executive order aimed at accelerating national AI development, harnessing federal datasets to train models for scientific research and discovery. 

The order directs the Department of Energy to create a secure, unified platform for AI experimentation to generate frontier models. Michael Krastios, science advisor to President Donald Trump, told CBS News that the project will empower scientists to reach currently unreachable breakthroughs, shortening “discovery timelines from years to days or even hours." 

The initiative is just the latest in a string of moves by the Trump Administration to secure AI supremacy in the heated race with China, having signed the AI Action Plan earlier this year and fighting against regulation that seeks to put boundaries on AI development in the name of safety. In the administration’s press release, it noted that the race to claim AI dominance was “comparable in urgency and ambition to the Manhattan Project.” 

As it stands, the US has the advantage of a strong concentration of advanced models, a strong talent pool and hardware and infrastructure that’s largely restricted from being sent to China, Thomas Randall, research director at Info-Tech Research Group, told The Deep View. 

And these efforts stand to greatly benefit US-based AI companies. The department will partner with a number of private sector tech giants on the project, including Nvidia, Anthropic, OpenAI, Google, AMD, and Amazon. 

“Much of this progress comes from the private sector, while government efforts mainly focus on helping innovation move faster, even if that means the country has fewer formal AI safety frameworks in place,” said Randall.

Chinese firms, however, are making their own strides, particularly on open source and low-cost AI. In mid-November, Beijing-based startup Moonshot AI released its Kimi K2 Thinking model, a trillion-parameter open source model. Firms like DeepSeek and Alibaba-backed Z.ai each have released their own open source, affordable models this past year. And AI demand is quickly growing in the country, as evidenced by Alibaba’s cloud revenues hiking 34% this past quarter. 

“It is moving fast in open-source AI and is very effective at weaving AI into daily life,” Randall said. “Because so many digital services in China are centralized and widely adopted, new AI features can spread across the population quickly.”

TOGETHER WITH TEMPORAL

62% of teams lose time or revenue to reliability troubles

It doesn’t have to be this way, though.

It’s time to create AI applications that go beyond prototypes and toys to work in the real world.

Easily build ambient and human-in-the-loop agents, and ensure your systems survive reliability challenges with Temporal.

We handle the tricky parts of AI systems so you don't have to.

Get started with Temporal today — build reliable AI systems in minutes.

CULTURE

Suno, WMG settle lawsuit

AI has notched another win in the music industry. 

On Tuesday, AI music firm Suno announced a “first-of-its-kind partnership” with Warner Music Group, settling the lawsuit that the record label previously brought against the startup. As part of the deal, Suno also acquired SongKick, a live music discovery platform, from Warner.  

Suno also noted a number of upcoming changes to its platform coming in 2026, including launching more advanced and licensed models, currently available models being “deprecated,” and downloading audio requiring a paid account. 

• The company is also introducing a number of restrictions, including that free-tier songs are not downloadable and paid-tier users have limited monthly download caps. 

• In the announcement, the companies noted that artists “will have full control over whether and how their names, images, likenesses, voices, and compositions are used in new AI-generated music.” 

“AI becomes pro-artist when it adheres to our principles: committing to licensed models, reflecting the value of music on and off platform, and providing artists and songwriters with an opt-in for the use of their name, image, likeness, voice and compositions in new AI songs,” Robert Kyncl, CEO of Warner, said in the announcement. 

The announcement comes on the heels of a number of developments between AI and the music industry as record labels seemingly shift their perspective on the technology. Last week, Warner settled its lawsuit against AI music platform Udio. In October, Universal Music Group settled with Udio and forged a partnership, and announced an alliance with Stability AI. Suno also announced a $250 million funding round last week, bringing its valuation to $2.45 billion.

Though labels seem to be warming up to AI, many musicians are rallying against the tech, and the jury is still out on consumer opinions.

LINKS

• Pro-AI super PAC launched $10 million campaign pushing federal AI laws

• Amazon asks engineers to use in-house AI coding tool over Cursor

• Spotify will raise US subscription prices in first quarter of 2026

• Peloton AI-powered bikes see sluggish sales so far

• Pony.AI plans to triple its robotaxi fleet by end of 2026

• XAI will reportedly close a $15 billion funding round in December

• Ilya Sutskever talks superintelligence, alignment on Dwarkesh Patel podcast

• Perplexity Shopping: An upgrade shopping experience in Perplexity, pulling from memory and offering instant checkout 

• Flux.2: Black Forest Labs latest model for image and test rendering

• ChatGPT Voice: OpenAI’s voice model is now available inside of its flagship chatbot, no separate mode needed. 

• Replit Data Connectors: Replit now offers data warehouse connectors for Snowflake, Databricks, and BigQuery. 

• Dragoneye: Build video detection models in minutes for Analyzing security camera footage, ensuring compliance or monitoring production lines.

• Perplexity: AI Inference Engineer

• Glean: Software Engineer, AI and Security

• Harvey: Staff Applied AI Engineer

• Character.AI: Research Engineer, Post-Training

GAMES

A QUICK POLL BEFORE YOU GO

With tech giants investing over $1 trillion in U.S. AI infrastructure, who do you think benefits most?

• Tech company shareholders (34%)

• China (15%)

• Defense and intelligence agencies (14%)

• Future Americans (19%)

• Something else (18%)