AI INFRASTRUCTURE

OpenAI’s $300B Oracle deal highlights AI’s scaling challenge

OpenAI’s $300 billion cloud computing deal with Oracle drove the market this week and highlights the pace of AI model development raising questions about whether we are at an inflection point where infrastructure spending is growing faster than revenues from AI products.

Chirag Dekate, a VP Analyst at Gartner covering AI, quantum and supercomputing, told The Deep View that while there's a lot to be excited about, it’s important to note that generative AI hasn’t quite lived up to its hype. When you factor in its elusive value returns, we reach a crossroads because greater infrastructure is needed to deliver continued gains, and this infrastructure is necessary to scale. 

Infrastructure spending is outpacing AI revenues, a concern Dekate says the market is ignoring. Much of the recent news reflects promises for future deals, not booked revenue, with many unknowns.

“I think what this deal does say is less about the pace of modern development,” he said. “I think it says a lot more about how the market is rewarding risk-taking as opposed to value delivery.”

On demand, Dekate noted that none of the model innovators have fulfilled their value promise yet. Current AI revenues fall far short of the massive investments being made, and the gap is widening as underlying problem spaces continue to grow. 

“These risks should not be ignored,” he said. “I think it's incumbent upon the modern innovators to showcase continued progress in terms of value scaling. And that has just not happened.”

OpenAI chose Oracle to help address its ongoing compute constraints, despite already having capacity with Microsoft, CoreWeave, Stargate and Google. Oracle provides reliable, extreme-scale GPU infrastructure, allowing OpenAI to scale both research and product development more effectively. At the same time, Microsoft retains exclusive access and the right of first refusal to OpenAI’s models, without assuming the risks associated with fast-depreciating GPU assets, making the partnership a strategic move that benefits all parties.

“It’s a win, win, win, sort of a partnership, and it enables risk mitigation for all parties involved, to a limited extent,” Dekate said.

The sheer scale of OpenAI’s Oracle deal highlights the significant compute requirements and growth trajectory of OpenAI. As these models become larger and more complex, especially frontier models, there is increased demand for infrastructure. The Oracle agreement highlights just how much infrastructure is necessary for frontier AI models, both for training and, increasingly, for inference as these models evolve to exhibit reasoning and agent-like behaviors. OpenAI is also forming a global network of partnerships with companies such as Microsoft, Google, Broadcom and SoftBank to support this scale.

Dekate summed it up best, saying that when you start putting it all together, OpenAI likely has access to the largest AI supercomputing capacity, not just in the U.S. “My inference out of this is that open AI is positioning for extreme-scale inference growth globally.”

Various users across X have questioned the logistics of the deal. Check it out here and let us know what you think in the poll at the end of today’s newsletter.

TOGETHER WITH AG1

AG1 Is Tackling The World’s Next Big Issue…

Sleep. So many of us struggle with it – not to mention the domino effect that it has on our energy levels, focus, physical and mental health, and beyond. So, 15 years after they first revolutionized your morning routine, AG1 is now optimizing your nightly ritual with AGZ.

What’s AGZ? Glad you asked. This nightly drink uses a blend of clinically-studied ingredients to support the 3 key stages of rest: Winding down, restful sleep, and waking refreshed. Featuring calming adaptogens, herbs, minerals (but no melatonin), all packaged up in delicious flavors like Chocolate Mint and Mixed Berry, it’s your one-stop pre-bed shop for a better night’s sleep.

Ready to turn down the stress and focus on the rest? You can try AGZ right here (and get a free Frother with your first purchase).

RESPONSIBLE AI

FTC investigates chatbots’ impact on kids and teens

The FTC wants to know what steps AI companies are taking to protect children and teens.

The agency asked seven AI companies with chatbots to turn over information on how they assess and monitor risks of the technology on children and teens. Those businesses include Alphabet, Character Technologies, Instagram, Meta Platforms, OpenAI, Snap, and xAI. All were asked to detail how they: 

• Mitigate harms, especially for children

• Monetize user engagement

• Track user engagement and process inputs

• Share data with third parties

• Generate and monitor outputs for potential harms before and after deployment

• Approve and develop characters (company- or user-created)

• Use disclosures, ads and other messaging to inform users about features, audiences, risks, and data practices

• Enforce compliance with rules, terms of service and policies

The FTC said it’s using the information to study the safety of chatbot usage among children and teens. The research is particularly important due to chatbots' ability to simulate human-like communication and potentially develop interpersonal relationships with their users. The FTC wants to gain insight into the possible harmful effects their interactions may have. 

We’ve covered numerous stories about Character.AI, Meta, and OpenAI’s recent troubles with conversations between minors and their chatbots.

“AI chatbots can effectively mimic human characteristics, emotions and intentions, and generally are designed to communicate like a friend or confidant, which may prompt some users, especially children and teens, to trust and form relationships with chatbots,” the FTC said in its announcement.

FTC Chairman Andrew Ferguson said protecting children online is a top priority for the FTC and so is fostering innovation in critical sectors of the economy. 

“As AI technologies evolve, it is important to consider the effects chatbots can have on children, while also ensuring that the United States maintains its role as a global leader in this new and exciting industry,” he said. “The study we’re launching today will help us better understand how AI firms are developing their products and the steps they are taking to protect children.”

TOGETHER WITH MODE MOBILE

iPhone 17 Is Coming. Here’s What Elon Thinks

“Apple used to really bring out products that would blow people’s minds.” Those are Elon Musk’s words. 

Today, Apple’s strategy looks less like innovation, and more like repetition. But a new smartphone company is stepping up to deliver the mind-blowing moments we've been missing.

Turning smartphones from an expense into an income stream, Mode Mobile has already helped users earn and save $325M+ and achieved a staggering 32,481% revenue growth rate over three years. Uber did it to taxis, Airbnb did it to hotels…And now, Mode Mobile is doing it to the $500 billion smartphone industry.

Now, with their Nasdaq ticker $MODE* secured, investors still have a chance to get in before a potential IPO.

Last round sold out at $0.30/share – don’t miss this one.

LINKS

• Albania appoints world’s first AI-made minister, will handle public procurement

• Google Cloud is making billions by monetizing its AI products

• Alibaba unveils trillion-parameter Qwen AI model

• Figure Technology shares close 24% above IPO price, raising $787.5 million

• F5 will acquire CalypsoAI for $180 million

• JFrog unveils AI platform upgrades for software supply chains

• Claude can now create and edit files

• BoostUp rebrands, launches new fleet of AI revenue agents

• Pennsylvania governor says state ready to lead AI revolution

• Proposed legislation may remove AI oversight for developers

• Latitude: Platform for building and running AI agents without code

• xpander.ai: Full stack development platform for AI engineers

• Genspark: An AI agent engine that designs anything with one prompt

• Modal Notebooks: High-performance, collaborative GPU notebooks in the cloud

• Gabriel, São Paulo: Senior ML engineer with 9 years designing scalable data infrastructure and 4 years implementing enterprise-grade recommendation systems (AWS, Kafka, Spark) — $47/h

• Mateo, Buenos Aires: AI specialist with 6 years in production-level computer vision systems and 3 years deploying models on secure edge devices for enterprise IoT use cases — $49/h

• Sofia, Medellín: NLP lead with 8 years delivering compliance-ready language models, including 5 years working on multilingual enterprise chatbots and document automation — $46/h

(Sponsored)

AI IN EDUCATION

AI classroom use jumps sixfold in two years

AI adoption in American classrooms has exploded from 10% of educators in 2023 to 67% today — a sixfold rise that's not only changing how teachers work but making them more likely to stay in the profession.

HMH Education's Educator Confidence Report, which surveyed 1,205 educators between May 6-22, reveals that generative AI is providing an unexpected morale boost for a profession facing widespread burnout and turnover.

The retention impact is significant: while 32% of teachers who don't use AI are considering leaving the profession within five years, only 21% of teachers who use AI are contemplating an exit. This comes as 16% of all teachers nationally intend to leave their jobs, down from 22% in 2024, but still concerning amid widespread teacher shortages affecting about 1 in 8 teaching positions.

Teachers are embracing AI for practical benefits:

• Time savings of 1-5 hours weekly (68% of users)

• Creating differentiated learning opportunities for students

• Strengthening student connections through personalized approaches

• 79% feel confident using AI in instructionally effective ways

However, concerns about the impact on students remain widespread. The biggest worry among both AI users (45%) and non-users (43%) is student over-dependence on the technology. Plagiarism concerns follow closely, cited by 44% of users and 40% of non-users.

Despite concerns, there's a broad consensus on responsible use: 87% of all educators agree students must learn to use AI ethically, while 79% believe AI tools should require teacher supervision.

Francie Alexander, HMH's SVP of Efficacy Research, noted that "teachers are pouring the time they're getting back right into their students," creating "more room for those meaningful moments."

AI is becoming less of a disruptive force and more of a retention tool for educators, potentially helping address chronic teacher shortages by making the profession more manageable and rewarding. With 3/4 districts struggling to fill positions for the 2024-25 school year, any tool that reduces the 11 percentage point gap in departure intentions could prove significant for staffing stability.

GAMES

A QUICK POLL BEFORE YOU GO