WORKFORCE

AI's impact on jobs is getting more complicated

The world expects AI to impact the job market, and cracks may already be appearing. 

On Tuesday, the Bureau of Labor Statistics reported that the US lost 105,000 in October and added 64,000 in November. The unemployment rate hit 4.6%, its highest in more than four years. 

Several developments to watch:

• One area in particular that’s seen an uptick, however, is construction. According to the Bureau, construction jobs rose by 28,000, with 19,000 of those roles being “nonresidential specialty trade contractors.” This rise coincides with the substantial push to rapidly build and deploy data centers in the US — sometimes called "AI factories."

• Physical AI may be a burgeoning area of interest for the tech industry, but it’s still in its infancy, far from ready to safely replace human workers. AI software, however, is more than ready to soak up office work. 

• Recent research from MIT and Oak Ridge National Laboratory, using a labor-simulation tool called the Iceberg Index, found that AI is already capable of replacing 11.7% of jobs in the US, amounting to $1.2 trillion in wages. 

• This follows a Microsoft study published over the summer that found positions such as sales representatives, technical writers and data scientists were highly exposed to AI, while hands-on jobs like mechanics, repair technicians and maintenance workers were the least exposed to the tech. 

As AI creates shaky ground for white collar jobs, the tech “is the one thing that will bring blue-collar jobs back at scale and close the labor shortage that's costing the US economy $1 trillion by 2030,” Daniel Walsh, founder of Veroskills, told The Deep View. 

It’s a sentiment echoed by Nvidia CEO Jensen Huang, who, in a recent episode of the Joe Rogan Experience, expressed a desire for more Americans to take on more manufacturing roles again, noting that he wants to “re-industrialize the United States.” 

“We need to be back in manufacturing,” he said on the podcast. “Every successful person doesn’t need to have a PhD. Every successful person doesn’t have to have gone to Stanford or MIT.”

The grand vision of AI is that it will take over the work humans don’t want to do. It was supposed to do the dangerous, dirty, hands-on work that puts people at risk, or even simply do the dishes and laundry that take up time we could be spending on intellectual pursuits. AI has, instead, been doing the opposite, handling the brainwork with enough proficiency that it makes companies like Salesforce, Accenture and Klarna comfortable enough with its capabilities to slash their white-collar workforces by thousands. With CEOs like Huang pushing this narrative as progress, the question we should be asking isn’t what AI can do, but who AI is actually helping? And with what? 

TOGETHER WITH AUTH0

Secure your AI agents with Auth0 for AI agents

Enable your AI agents to more securely access your tools, workflows, and users’ data. Outsource the various complexities (like auth) so you can focus on what really matters – building the app. That’s why we’re such fans of Auth0 for AI Agents.

Auth0 for AI Agents handles all the tricky aspects of adopting AI for your business with just a few lines of code. Handle and secure access tokens, process authorization requests, and implement AI with more confidence… without sacrificing any security. It’s exactly the kind of support builders like those that read The Deep View need to take their idea to the next level. 

Try Auth0 for AI Agents, now generally available—the complete auth solution for AI agents.

HOW TO AI

Turn any question into a cited research report

Perplexity's Research mode is like having a research analyst on call. You ask one question; it runs dozens of searches, reads hundreds of sources, resolves contradictions and delivers a structured report with inline citations in under 4 minutes. The free tier provides 5 reports per day (150/month versus 5/month for ChatGPT's competing Deep Research feature). Use it for market research, competitor analysis, literature reviews, product comparisons, or any question that would normally take hours of Googling.

Steps:

1. Go to perplexity.ai and sign in (free account works)

2. Click the mode selector in the search box → Select "Research"

3. Write a specific prompt with scope, timeframe, and desired format:

   • ❌ "AI trends"

   • ✅ "Analyze AI adoption in healthcare diagnostics 2023-2025. Include market size, key players, FDA-approved tools, and barriers to adoption."

4. Wait 2-4 minutes—you can watch it search, read, and reason in real-time

5. Export: Click the Download Icon and get it as a PDF or Document, or convert to a shareable Page

Takeaways:

Your prompt is everything. Don't say "AI trends"—brief it like a junior analyst: timeframe, industry, format, angles to cover.

Always click through citations before acting on anything important. It occasionally pulls from thin sources.

It can't access paywalled content or breaking news, and struggles with super niche topics. For higher-stakes research where you're comfortable waiting 5-30 minutes, OpenAI's Deep Research goes deeper—but again, you only get 5 free per month.

MARKETS

US sentiment turns negative on AI factories

Frontier labs need them. Investors love them. Most people only have a vague sense of what they are. 

This is the role that data centers have tended to play in the AI boom. There is big money in installing advanced computer chips in large rooms to train AI models. But more recently, the tide of sentiment appears to be turning against the relentless trend of data center building.

The AI industry operates on the belief in a scaling law: the more computing power used to train models, the more advanced the models become. In the high-stakes race to build frontier AI models, companies have poured large sums into data center development. Oracle and OpenAI struck a deal pouring $300 billion into building data centers over the next five years, for example. 

In the face of relentless demand, even some businesses tangential to data centers have made a pivot. Aviation startup Boom Supersonic is using its jet-engine technology to deliver energy to data centers, and Ford is using its electric-vehicle batteries to power them.

But the data center Midas touch may be growing less golden. CoreWeave, which builds and leases large data centers to companies like OpenAI, has been hammered by investors in recent weeks over concerns about its high debt load and the high cost of building new data centers. Likewise, investors have slammed Oracle for spending heavily and taking on debt while financing new data center projects.

That’s all on top of a potential backlash to the gravity of data centers in the economy, especially if other types of infrastructure see their resources sapped. US Senator Bernie Sanders even announced he would push for a moratorium on the construction of AI data centers in the US to "give democracy a chance to catch up."

Make no mistake: data centers will continue to be massively lucrative. Their importance will likely only be bolstered by US fears of falling behind China in the AI race. The US House of Representatives just voted to advance a bill that would make the permitting process easier for AI infrastructure projects. However, as AI bubble fears intensify and companies face greater scrutiny of their balance sheets, perhaps data center build-outs will cease to be the kind of blank-check projects they have been for most of 2025.

TOGETHER WITH YOU.COM

Stop Guessing. Prove AI ROI.

AI spend is rising, but are you measuring return on investment? We love this guide from You.com, which gives leaders a step-by-step framework to measure, model, and maximize AI impact. 

What you’ll get:

• A practical framework for measuring and proving AI’s business value

• Four essential ways to calculate ROI, plus when and how to use each metric

• A You.com-tested LLM prompt for building your own interactive ROI calculator

Turn “we think” into “we know.” Download the AI ROI Guide

BIG TECH

Databricks raises $4B to power AI agents

There may be growing skepticism about how AI companies will make money, but one area of the AI ecosystem where there's no question about revenue growth is data. 

We saw another example this week: Databricks raised $4B in a Series L funding round, valuing the company at $134B. Databricks primarily needs the extra capital to fuel the growing demand for its technology to power AI agents. Databricks has both Lakebase, a database for AI agents, and Agent Bricks, an AI agent platform. In both cases, the company provides a foundation for businesses to deploy powerful agents while locking down privacy, security, and data sovereignty. 

Let's put Databricks' fundraising round — and its growth — in perspective:

• It's the third time it has done a funding raise in less than a year

• Databricks had just raised $1B in August at a $100B valuation, so the company has jumped in value by 34% in three months

• The company's annual revenue run rate (ARR) hit $4.8B in Q3, and the company said 20% of that now comes directly from its AI products

• Arch-rival Snowflake also recently hit a $4.8B ARR, but it's a public company and is valued at a $74B market cap — that's quite a disparity from Databricks' $134B

The big question is how many more times Databricks will return to the private market for funding before going public. "I wouldn't rule out that we would be public by the end of next year," said Databricks CEO Ali Ghodsi on TBPN. "We're not trying to exactly time the market. We're trying to win it."

Larry Dignan, Editor in Chief of Constellation Research, told The Deep View, "Databricks has the growth that's fueling its data-meets-AI strategy. What's unclear is whether that growth makes Databricks worth the valuation premium to Snowflake. It will be fun to watch these two fierce rivals battle when both are publicly traded."

While various parts of the AI boom — particularly consumer apps — are facing tough questions about profitability and long-term viability, enterprises continue to plow ahead with plans to deploy AI to drive efficiency, productivity, and automation. That's especially true for AI agents and the need to build a strong data foundation for successful deployments. That trend continues to power rapid growth for both Databricks and its rival Snowflake.

LINKS

• Enterprise agent platform Sequence raises $20 million in Series A funding.

• Meta updates smart glasses to isolate sound in noisy rooms

• DoorDash rolls out an AI social app for finding restaurants called Zesty

• Bill to cut down red tape for AI data center buildout advances in House 

• Lovable is reportedly valued at $6.6 billion in the latest round

• Waymo is in talks to raise at a more than $100 billion valuation

• Meta SAM Audio Model: A unified AI model that allows users to isolate and edit sound from complex audio. 

• Gemini 2.5 Flash Native Audio Model: An updated version of Google’s flagship model, now built for live voice agents. 

• Adobe Firefly Updates: Adobe’s generative AI tool now supports prompt-based video editing and includes new third-party models from Black Forest Labs and Topaz Labs.

• AI2 Bolmo: A new family of models from the Allen Institute for AI for “byte-level” language model training with better quality. 

• Alibaba Wan2.6: A new multimodal model for generating images and videos with high-quality instruction adherence, motion physics and aesthetic control.

• Google CC: An experimental productivity agent that delivers a personalized email briefing every morning.

• Nvidia: Generative AI Application Engineer

• Capital One: Applied Researcher I (AI Foundations)

• Snap: Research Scientist, Generative AI

• Amazon: Member of Technical Staff - Reinforcement Learning, AGI Autonomy

GAMES

POLL RESULTS

Should US states be able to regulate AI individually?

• Yes, states have to be able to set guidelines (29%)

• No, we need one policy to untether innovation (48%)

• Other (23%)