Good morning. The Sun emitted its largest solar flare on record since 2017 yesterday, and scientists are saying that a geomagnetic storm is highly likely.

For those who missed the Aurora last time it popped up all over the world, you might have another shot this weekend (though timing and location remain uncertain at this point).

Happy hunting!

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

AI for Good: NASA discovers a new triple-star system

A team of NASA professionals, working alongside amateur astronomers — and armed with artificial intelligence — recently discovered a new triple-star system. 

The details: The system contains one set of stars that orbit each other every 1.8 days, with a third star circling the first two every 25 days. This discovery broke a 1956 record for the shortest outer orbital period for such a star system. 

This is just the start for NASA; researchers will soon begin examining images from its Nancy Grace Roman Space Telescope, which will be far more detailed than the ones obtained by TESS. 

“We don’t know much about a lot of the stars in the center of the galaxy except for the brightest ones,” Brian Powell, a data scientist at Goddard, said in a statement. “Roman’s high-resolution view will help us measure light from stars that usually blur together, providing the best look yet at the nature of star systems in our galaxy.”

Design and Make Anything with Autodesk software

Fusion is more than just CAD—it's an all-in-one cloud platform that combines CAD, CAM, CAE, and PCB into a single solution. With Fusion, you can:

Trusted by over 4.6 million professionals, Fusion is the future of design and manufacturing.

Start with a free 30-day trial to explore all the features, and you can get 30% off a 1-year subscription from 10/3 to 10/31.

Learn more about Fusion today!

Report: GenAI coding assistants don’t really help 

Since its gimmicky popularization, tech companies have been searching for powerful applications for the massively expensive, environmentally costly generative AI that they’ve been developing. 

One of the earliest applications here involves the idea of a generative AI coding assistant. 

A new study casts a bit of doubt on the efficacy of that application. 

The details: The study, conducted by Uplevel, featured a sample of around 800 engineers across Uplevel’s customer population. 

The report also found that developers with Copilot access saw a 41% higher bug rate than those without. 

Uplevel said that engineers might benefit from deploying a more conservative Copilot adoption strategy until the tools get more powerful. 

GenAI Productionize 2.0

Don't miss the premier summit on productionizing generative AI within the enterprise!

Join AI experts from leading brands, startups, and research labs for insights and actionable strategies on generative AI governance, frameworks, and practical techniques for evaluation and observability.

Free Registration

If you want to get in front of an audience of 200,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

Microsoft’s Italian expansion 

Microsoft said this week that it will invest $4.8 billion toward expanding cloud and AI infrastructure in Italy over the next two years. Part of this investment will involve an effort to train a cohort of more than one million Italians in new digital skills and AI. 

The details: Microsoft said that this investment will make Northern Italy one of its largest data center regions in Europe. 

Some context: This expansion is just one of many that Microsoft is undertaking. The tech giant recently announced that it is building two new AI centers in Abu Dhabi; at around the same time, Microsoft said that it will invest $1.3 billion over the next three years in Mexico, to expand its infrastructure for cloud computing and AI. 

A component of this is that the infrastructure here is incredibly energy-intensive, a problem that has not been solved despite Big Tech’s increasing push for more and more capacity. 

AI companies need a lot of money. Investors are still forking it over

The business of artificial intelligence is one that has been dominated by fantastical promises and enormous expenses. 

The inherent obstacle here is the cost of GPUs, the chips that are needed to train and power generative AI. A single Nvidia H100 chip can cost upwards of $30,000, and these companies need far more than one chip.

In addition to the chips, there’s the high cost of the energy needed to power these GPU-lined data centers, as well as the cost of retaining top talent, which in the tech and AI sector, has proven to be enormous. 

All in, you have a venture that’s burning billions of dollars each year and is generating hardly any return. OpenAI, for instance, expects to lose $5 billion this year, despite being on track to rake in some $3.7 billion in revenue. 

But the state of its financials (and its organizational chaos) has not scared investors away. OpenAI recently secured $6.6 billion worth of new funding at a $157 billion valuation. The round included participation from Soft Bank, Microsoft, Nvidia and Thrive Capital. Apple was initially slated to participate, but backed out shortly before the round closed. 

A day after securing the funds, OpenAI said that it had established a $4 billion revolving line of credit with JPMorgan Chase, Citi, Goldman Sachs, Morgan Stanley, Santander, Wells Fargo, SMBC, UBS and HSBC. With the credit boost, OpenAI said that it now has access to “over $10 billion in liquidity, which gives us the flexibility to invest in new initiatives and operate with full agility as we scale.”

What is clear is that, right now and for the foreseeable future, AI is a cash-burn business. But investors, weirdly, don’t seem to care. 

A look at the data: New Crunchbase data found that, in another quarter of slowing venture funding around the world, AI is still the leader. Nearly a third of all venture dollars in the third quarter of the year went to AI startups, according to Crunchbase, amounting to nearly $19 billion. 

And despite the lack of clear payoff for AI companies, tough hurdles for adoption and Big Tech dominance, the venture mindset — looking at the numbers — is pretty optimistic. Crunchbase found that Q3 represented the “second-largest quarter for AI funding since the mainstream launch of OpenAI’s ChatGPT in November 2022, behind only Q2 2024.”

The quarter included a $5 billion raise from Waymo, a $1 billion raise from Ilya Sutskever’s Safe Superintelligence and a $640 million raise from Groq. 

Still, the number of deals has been steadily declining over the past year or so, while the size of the deals has been going up.

There’s a few things going on here. One is the unflagging optimism from investors, which I would attribute to faith in the heavily hyped and marketed transformative potential of AI. Sam Altman and the rest have done their job in selling an attractive idea of how powerful AI will be. 

I don’t see any evidence that, economically, there’s any justification — at these levels of investment — for a system that promises to increase efficiencies, yet does so in a profoundly inefficient way. 

There’s also OpenAI’s newfound $10 billion liquidity, which smells a bit desperate to me. Specifically, the fact that the $6.6 billion fundraise wasn’t enough — it needed to be coupled with an extra $4 billion line of credit … this is a business that cannot stand on its own. It is simply too costly. 

At some point, OpenAI and the rest will stop subsidizing the cost of using their generative AI models, and it will become significantly more expensive. At that point, I think the business might fall apart, as the integration of generative AI becomes less worthwhile the more expensive it becomes. 

Its big calling card is cheap efficiencies. Expensive efficiencies aren’t efficient. 

🤔 Your thought process:

Selected Image 1 (Left):

Selected Image 2 (Right):

💭 A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

Here’s your view on the German court’s decision:

32% of you said the court did not make the right decision; 30% said it did, as scientific exceptions must be protected. The rest said it’s not really clear.

Something else: