What is OpenAI head Sam Altman digging: an economic moat or a grave?

This fall was marked by a flurry of news about the premier artificial intelligence developer, OpenAI. Who would have thought that a small nonprofit company founded in 2015 would command hundreds of billions of dollars 10 years later? Yet that's exactly what has happened to OpenAI - its CEO Sam Altman has not only driven its value to $500 billion, making it the world's most expensive startup, but he's also making deals that impact the stocks of major tech companies. However, what is he doing, in a strategic sense?

The phrase "Move fast and break things" was the motto of Facebook (now Meta) and a popular Silicon Valley philosophy that proclaimed that speed of innovation was more important than caution. However, in 2014, Facebook officially abandoned it, and in 2019, Hemant Taneja, now CEO of venture capital fund General Catalyst, wrote in Harvard Business Review that this era is finally over - it's time for startups and tech investors to take a more responsible approach, assessing how their technology will impact society.

Sam Altman, however, doesn't seem to agree. Despite the poor release of the latest GPT-5 model in August, which he himself said later, "We had some serious screw-ups in the rollout," his company has since unleashed a flurry of new developments.

On September 29, it announced the introduction of Instant Checkout, a shopping feature directly from ChatGPT, for starters on Etsy and Shopify sites.

The very next day, Sora was released, an iPhone app that uses AI to allow users to easily create and voice short videos and create a TikTok-like feed of them. The "cameo" feature can also add a specific person's image and voice to a video, including your own - what Wired called "dipfaking yourself for entertainment."

On the third day after release, despite invitation-only access, Sora rocketed to the top spot on the AppStore's free app chart, surpassing even ChatGPT itself.

On October 6, OpenAI showed new tools at the OpenAI Developer Conference. In particular, how you can call other applications directly from ChatGPT and work with them using simple text prompts: create images in Canva, create playlists in Spotify, or find real estate with Zillow.

"Over the next six months, you're going to see the evolution of ChatGPT from a really very useful application to something more like an operating system," Reuters quoted ChatGPT CEO Nick Turley as saying. He also added that the company "never planned to create a chatbot. We wanted to create a super helper, but we got a little distracted."

Altman is also actively poaching Apple employees for his secret project with former Apple VP Johnny Ive to create some new AI gadget (or several). This year alone, about 20 people have been poached, and Apple even had to cancel a big meeting in China in August so that executives wouldn't have to leave the Cupertino headquarters and could counteract the frenzied headhunting by OpenAI on the spot, according to The Information.

What else has the head of OpenAI been busy with besides the rapid expansion of the team and product line? A very important thing: clouds.

Altman has never hidden that he believes in a simple formula - the more computing power, the smarter the AI trained and working on it. He wrote about this in his manifesto "The Age of Intelligence" a year ago.

Over the course of this year, his appetites have only increased. "Now we have to make these terrible compromises. We have more efficient models, but we just can't offer them because we don't have the capacity. We have other new products and services we'd like to offer," he told The Verge in an interview in August, adding that OpenAI will spend "trillions" (yes, that's right, plural) of dollars on data centers in the near future.

That is, it is no longer enough for him the $500 billion that investors led by Softbank head Ma Maesi Son in January promised to invest in a joint project with OpenAI to create data centers Stargate.

As it soon turned out, Altman was not kidding about the trillions. His deals to create and equip the latest data centers with Oracle and Nvidia were valued by the Financial Times at $300 billion and $500 billion respectively. And on Monday, October 6, another mega deal with AI processor maker AMD was announced. Shares of the chipmaker after the announcement soared by 24%, increasing its capitalization by more than $60 billion to $331 billion, Bloomberg writes. The FT also estimates this deal at $300 billion, plus the September agreement with CoreWeave, to which OpenAI has already promised a total of $22.4 billion.

And it turns out that OpenAI is going to spend more than $1.1 trillion on cloud computing power alone over the next 10 years.

Quite ambitious for a company that, according to the FT, will generate $12bn in revenue this year while making a loss of around $10bn. Questions about profitability are simply brushed aside by Altman, saying they are not even "in his top ten concerns".

"Of course, we have to become very profitable someday, and we are confident and patient - we will get there... Right now we are in the investment and growth phase," he said in October.

The strategy of "growth first, profit later" is not new. For example, it was used by Amazon, which first broke even in the fourth quarter of 2001, seven years after its founding. The New York Times wrote about it in an article with the telling title "Amazon's Surprise: Its First Profit". However, the scale is not comparable - Jeff Bezos's company had by then burned through "only" $2.8 billion, that's about $5.1 billion in today's prices.

To commit more than $1 trillion to your company, you have to be ... it's hard to even find the words .... very confident in yourself and your technology, not to say "crazy." What is Altman hoping for?

"OpenAI and a growing number of its partners are betting that the use of AI will continue to grow exponentially. If growth stalls or even slows, the investor enthusiasm that has helped boost share prices through these deals could quickly fade," warns the FT.

How it will turn out is unknown, and extremely difficult to predict. However, are there any other factors that could contribute to the success (or failure) of OpenAI?

Warren Buffett, the famous investor and creator of Berkshire Hathaway, nicknamed "The Oracle of Omaha" for his foresight, invented and promoted the term "economic moat" - a company's competitive advantage that shields it, like a moat is a castle, from the encroachment of competitors and thus allows it to make good profits in the long term. These were the companies Buffett tried to choose for investment.

He cited Geiko, an automobile insurance company, as an example, which set up direct-to-consumer sales without middlemen, and that allowed it to cut costs and offer lower prices, writes CNBC, which has built up an entire library of Buffett's public speeches.

The opposite example is Coca-Cola, here the moat is its uniquely strong global brand, which allows it to sell its drinks more expensively than its competitors.

Until recently, OpenAI's economic moat was the patents on its AI developments, but that moat is "increasingly fragile," Fortune writes, citing analysts at JPMorgan. Competitors are catching up.

As prominent technology analyst Mary Meeker noted in June, the costs of building and training the latest AI models are rising while the prices for using them are falling, especially thanks to the entry of Chinese companies into the race, which, unlike the same OpenAI, rely on open-source models and offer access to them more cheaply.

Yes, ChatGPT now leads with 78% of the market and its six major competitors share the remainder, but can it hold on to the lead? That's the question Fortune asks itself. There may be a chance of that. As Robert Siegel, a management professor at Stanford University's Graduate School of Business, told the magazine, OpenAI is already well on its way to achieving a valuable quality - "stickiness": the longer customers use a particular AI, the less likely they are to switch to a competitor because the AI is learning. The more you use a particular AI engine, the more it learns about you and your needs. "If you keep asking questions on ChatGPT, which has learned your behavior better, and you like it, there's no reason to leave," he believes.

There is another kind of economic moat - infrastructural. Another Buffett investment is often cited as an example - BNSF (Burlington Northern Santa Fe Corporation), one of the largest U.S. freight railroad networks with 8,000 locomotives and 52,000 km of track. Given the enormous cost of this infrastructure, it is virtually impossible to create a competitor.

According to McKinsey, there is now 44 GW of AI data center capacity worldwide, and Sam Altman, with the help of partners, intends to build about 20 GW in the US alone in the next 10 years. If he succeeds, it will be the kind of moat that will also prove extremely difficult to bridge. Perhaps his almost manic obsession with deals to build more and more data centers can be explained by this.

In addition, since Altman has managed to involve such companies as Oracle, AMD and, most importantly, Nvidia - a true technological behemoth, the first and so far the only company in the world with a capitalization of over $4.5 trillion - they will now have to play along with him. Nvidia has already pledged to invest up to $100 billion in OpenAI in exchange for a stake in the startup, while AMD, on the other hand, may sell up to 10% of its shares to OpenAI at a symbolic price of 1 cent (current value $217) as part of the deal. In the long run, this, too, could help OpenAI finance its purchases of AMD's AI processors.

Well, it's do or die. Sam Altman seems to be digging either an economic moat or a grave, and if he fails, not only for himself.

This article was AI-translated and verified by a human editor